For years, data governance, the combination of roles, policies, and processes that make data trusted and enable responsible use, was the cornerstone of enterprise data strategy. Every data leader knew the playbook: define ownership, enforce standards, establish lineage, and implement controls to ensure data is accurate, secure, and used appropriately. However, organizations still experience mixed results from many [lingering BI issues](https://cube.dev/blog/the-harsh-lessons-of-modern-bi-and-why-scaling-ai-will-be-even-harder).

Now, AI has entered the chat with the same underlying concerns, only amplified. While AI has an endless supply of answers, many questions remain. Who owns the model? Can you trust its outputs? Is the data used to train or query the model governed? Is it biased? Are decisions traceable? Organizations must now focus on [AI literacy](https://cube.dev/blog/ai-literacy-is-the-new-data-literacy-why-a-semantic-layer-is-the-missing) for their data teams to build the right foundation, while bringing people, process, and technology together.

The rise of generative and agentic AI has pushed governance from a back-office IT concern to a boardroom imperative. But here’s the truth: you can’t have AI governance without strong data governance. And even more importantly, you can’t scale AI governance unless it’s built into the systems AI relies on. That’s where Cube Cloud’s universal semantic layer plays a defining role. It centralizes business logic, ensures consistency, and enables AI to generate responses that are not just fast, but trusted, auditable, and aligned with enterprise standards.

## What Is AI Governance?
AI governance refers to the framework of policies, processes, and controls that guide the responsible use of AI technologies. It ensures AI operates within acceptable boundaries, ethically, legally, and strategically. But unlike traditional data governance, AI governance adds layers of complexity:

- Model behavior and interpretability
- Prompt input and response output control
- Bias mitigation in generated content
- Versioning and change management of semantic and model logic
- Auditability of AI-driven decisions
- Security and compliance for data used in AI training or querying

While the goals of accuracy, consistency, and security are similar to data governance, the scope is broader and the consequences of failure more public. A flawed dashboard might cause confusion. A flawed AI-generated report could trigger a legal investigation.

## Why AI Governance Fails Without Data Governance
It’s important to remember that AI doesn’t reason as humans do. It synthesizes patterns in the data it’s trained on and queried with. If that data isn’t governed—if metrics are inconsistent, definitions vary, or calculations aren’t understood—then no amount of prompt engineering or model fine-tuning will save you. The result is fast answers no one trusts. Data governance gives you the raw materials of trust. But raw materials aren’t enough. AI systems need to interpret and apply that governance consistently in real time. That’s where most organizations hit a wall.

## Where AI Governance Meets Business Logic
Cube Cloud is the AI- and BI-ready universal semantic layer that translates governed data into business-ready answers. It ensures that every AI system from copilots and chat interfaces to LLM-powered analytics agents accesses a consistent, auditable, and contextualized representation of your data. Here’s how Cube Cloud enables AI governance in practice:

__1. Centralize Business Definitions Across All AI Outputs__

With Cube Cloud, you define metrics like “Customer Churn,” “Gross Margin,” or “MRR” once in plain, auditable, version-controlled YAML, then expose them to every tool, including AI interfaces. This means when an AI tool answers a question about churn, it’s using the same business logic your CFO signed off on, instead of its own fuzzy interpretation of the word. AI governance starts with semantic consistency, and Cube enforces it.

__2. Trace Every Answer Back to Its Source__

A cornerstone of governance is auditability. You need to know where the data came from, what transformations were applied, and why a specific answer was generated. With Cube Cloud’s semantic layer, every AI-generated response can be traced back to the exact data model, metric definition, and data source. Whether the AI output appears in a Slack thread, an executive summary, or a web app, it’s backed by governed, inspectable lineage. This kind of visibility turns AI from a risky black box into a trustworthy system of record.

__3. Govern and Enforce Data Access Controls__

The problem with traditional governance is fragmentation with rules and definitions varying by BI tool, spreadsheet, or team. Cube solves this with a single source of truth that feeds every downstream data experience, including AI. Instead of building custom AI guardrails for every department or model, you define governance rules in Cube once and enforce them everywhere. This reduces risk, eliminates inconsistencies, and ensures compliance at scale. Whether AI is summarizing a dataset, generating a dashboard, or responding to a prompt, it’s doing so within the boundaries you’ve defined.

__4. Support Human-AI Collaboration, Not Confusion__

AI governance isn’t just technical. It’s cultural just like data governance. Employees need to understand what AI is doing, how it works, and why they can trust it. When AI confidently generates a financial summary, it needs to align with what the business actually reports. By embedding semantic understanding into AI interactions, Cube makes AI more predictable, interpretable, and aligned with human expectations. That’s essential for adoption and critical for compliance.

## AI Governance Is a Data Problem, Solved by Cube
Many companies are sprinting to define new AI governance frameworks, policies, and oversight committees. That’s necessary. But without a solid data foundation, those efforts are cosmetic at best.

The real opportunity lies in operationalizing data and AI governance through architecture. That’s exactly what Cube Cloud delivers. With a universal semantic layer at the center of your AI, data, and analytics strategies, you can ensure:

- Consistent answers across every AI interaction
- Trust in outputs, grounded in governed definitions
- Full transparency for audits, compliance, and internal review
- Scalable enforcement of business logic and access controls

## Govern Your Data, Govern Your AI
As generative and agentic AI take hold across the enterprise, data leaders must shift from enabling curiosity to enforcing confidence. Governance can no longer be a gatekeeper. It must be an enabler, but it can only do that if it lives inside the systems AI touches.

AI governance isn’t separate from data governance. It’s the next evolution of it. Cube Cloud is the bridge between the two, making sure every AI response is trustworthy and transparent for your business.

If you’re serious about scaling AI safely and confidently, start where trust begins: with your data, definitions, and semantic context. [Contact sales](https://cube.dev/contact) to learn more about how Cube Cloud centralizes logic, ensures consistency, and makes every answer traceable for the enterprise.


Michael Hetrick

Artificial Intelligence

Why AI Governance Starts with Data Governance 

We're thrilled to announce a significant leap forward in our AI API capabilities, particularly for our valued users leveraging Snowflake. Last month, as part of the General Availability (GA) release of our powerful AI API, we proudly unveiled the "Bring Your Own Large Language Model" (BYOLLM) feature. This groundbreaking addition empowers organizations to seamlessly integrate their own compatible Large Language Models (LLMs) with our AI API, directly addressing critical concerns surrounding data security and cost optimization.

In today's data-driven landscape, organizations are increasingly seeking to leverage the immense potential of LLMs for various applications, from natural language processing to advanced analytics. However, with this increased adoption comes a heightened awareness of the need to protect sensitive data and manage costs effectively. This is where Cube's AI API with BYOLLM truly shines.

[Cube’s AI API](https://cube.dev/docs/product/apis-integrations/ai-api) allows Cube Cloud customers to use an LLM to interact with their data, via their trusted and governed universal semantic layer. Users can simply ask questions in natural language and the AI API will generate a Cube JSON Query to run or be run against our REST API, in order to answer their question. With BYOLLM an organization can choose to use a compatible LLM from their cloud provider and now from Snowflake the AI Data Cloud.

For organizations deeply invested in the robust Snowflake data cloud platform, maintaining LLM usage within the existing Snowflake security boundary and optimizing credit spend is of paramount importance. We recognize these priorities, which is why we're excited to highlight that you can now seamlessly utilize the cutting-edge Anthropic Claude 3.5+ Sonnet models offered on Snowflake Cortex directly with our AI API.

This powerful integration ensures that your highly sensitive data remains securely within your established Snowflake environment, where you have complete control over access and governance. At the same time, you can unlock the full potential of advanced AI capabilities, enabling you to derive deeper insights and drive innovation. Furthermore, this integration offers significant cost management benefits by consolidating your LLM usage within your existing Snowflake infrastructure and credit spend, eliminating the need for separate, potentially more expensive, AI services.

Getting started with this game-changing integration is remarkably straightforward. We've documented the process in our documentation, which you can access here. In essence, it simply involves enabling the BYOLLM feature on the AI API configuration page within Cube Cloud and securely saving your Snowflake Cortex credentials. This streamlined setup allows you to quickly and easily harness the power of Snowflake Cortex hosted LLMs with your Cube-powered applications.

We are acutely aware that a large and growing number of organizations are actively using open-source Cube or our fully managed Cube Cloud platform in conjunction with Snowflake. We deeply value this vibrant community and are particularly enthusiastic about collaborating with these organizations to facilitate their seamless adoption of the AI API using Snowflake Cortex hosted LLMs. We understand that each organization has unique needs and workflows, and we're committed to providing tailored support and guidance to ensure a smooth and successful integration.

We believe that the combination of Cube's AI API, BYOLLM, and Snowflake Cortex offers a powerful solution for organizations seeking to enhance their data security, optimize costs, and unlock the transformative potential of AI. By keeping your LLM usage within your Snowflake environment, you can maintain control over your data while leveraging cutting-edge AI models.

If you're interested in exploring this integration further and discovering how it can benefit your organization, please don't hesitate to reach out to us. Our team is readily available to answer your questions, provide technical guidance, and assist you in getting started. We are committed to helping you maximize the value of your Snowflake investment and unlock new possibilities with AI.

We're excited to embark on this journey with you and witness the innovative applications you'll build with Cube's AI API and Snowflake Cortex hosted LLMs. Let's work together to drive innovation, enhance security, and achieve new levels of efficiency. We look forward to hearing from you!

David Jayatillake

Bring your own Snowflake LLM 

Over the last decade, enterprises have poured time, money, and headcount into data literacy programs. The goal was admirable: empower every employee to understand data, make informed decisions, and reduce over-reliance on central data teams. But reality painted a different picture. Most programs struggled to scale, failed to deliver measurable impact, or collapsed under the weight of misaligned tools, training, and expectations.

Now, AI is having its moment, only now there’s more urgency, higher stakes, and greater hype. Organizations are scrambling to deploy generative AI across every function, promising faster insights and better productivity. Beneath the surface, the same cracks are forming. AI may be more powerful than dashboards ever were, but it still relies on the same foundation of good data. Just like with BI, people using AI without understanding or trusting the data behind it are set up to fail.

If data literacy was about helping humans understand data, AI literacy is about helping humans understand the answers AI gives them. And if we’ve learned anything from the past, it’s that literacy can’t be taught in isolation. It must be built into the systems and workflows people already use.

## AI Literacy Starts with Trusted Data
AI literacy isn’t just about prompt engineering or understanding hallucinations. It’s about ensuring that every AI-generated answer is aligned with how your business defines success. It’s about making sure AI is pulling from the right sources, using the right terminology, and reflecting the right business logic. 

Without guardrails, even the best AI can confidently generate the wrong answer. This is where the universal semantic layer becomes essential. Cube Cloud sits between your data sources and your AI applications, providing a consistent, governed layer of business logic that AI can understand and apply. Just as it does for BI tools, spreadsheets, and embedded analytics, Cube ensures your AI is grounded in business context, not just raw data. Let’s break that down:

__1. Enrich Your Data with Business Context__

Raw data is rarely meaningful on its own. A column labeled rev_qtrly_adj might mean something to a data engineer but not to an AI model or a business stakeholder trying to interpret AI-generated output. This is where semantic modeling come in.

Cube Cloud lets teams define dimensions, metrics, relationships, and hierarchies in plain business terms. For example, instead of exposing a sales rep or an AI model to messy SQL logic and table joins, you define a metric like “Net Revenue” once, using the exact business logic your finance team approved. Then, any AI agent can query that metric through Cube, no interpretation or translation needed.

This semantic model becomes the shared language between your data, your AI, and your users. It transforms your data from a black box into understandable, usable knowledge. And most importantly, it eliminates the risk of AI making assumptions that don’t match your business logic.

__2. Generate Responses Aligned with Vocabulary and Intent__

Generative AI is only as useful as the quality of the prompts and the clarity of the data it interacts with. But in the real world, prompts are vague, messy, and inconsistent, just like human-to-human communication. A sales leader might ask, “What’s our top-line growth in APAC last quarter?” while a product manager asks, “Are signups trending up?” Without shared context, these prompts can lead AI to produce wildly different—or just plain wrong—answers.

Cube Cloud helps generative AI systems bridge the gap between intent and execution. By connecting AI tools to governed semantic models, Cube ensures that vague human language is translated into precise, context-aware queries. AI no longer has to guess what “top-line” means or how to filter “last quarter”. It uses the definitions, filters, and calculations your business already trusts.

With Cube, AI responses speak your business’s language—literally. Whether you’re using AI to build dashboards, write executive summaries, or answer ad hoc questions in natural language, every result aligns with how your business thinks and talks.

__3. Ensure Traceable, Auditable Outputs__

One of the biggest challenges in AI literacy is trust. Can you trust the answers AI gives? Where did the data come from? What logic was applied? Can you verify the numbers? Without transparency, AI becomes a black box that users either blindly follow or entirely ignore. Neither option is safe.

Cube Cloud makes AI output traceable by design. Every response generated via Cube is backed by governed, version-controlled models. You can trace the result from the data source to the exact metric definition and downstream data consumers. Need to know how AI calculated revenue growth? You can inspect the semantic model. Need to verify the filter applied for “APAC”? It’s documented and auditable.

This level of transparency doesn’t only boost trust; it makes AI literacy achievable. When users understand where answers come from and why they’re correct, they’re more likely to embrace AI as a reliable partner, not a risky shortcut.

## From AI Hype to AI Habit
The AI boom is a multi-year megatrend that shows no signs of slowing down. But for AI to truly transform how organizations work, it has to become part of daily decision-making. That means people have to trust it, understand it, and be confident using it. That’s what AI literacy is really about.

Just as we learned in the data literacy era, literacy is not about training alone. It’s about putting the right foundations in place, making understanding natural, usage easy, and governance automatic, bringing people, process, and technology together.

Cube Cloud is that foundation. By enriching your data with semantic context, aligning AI output with business intent, and ensuring every answer is traceable and governed, Cube makes AI literacy both possible and sustainable. AI doesn’t just need data access. It needs the right data, used the right way, understood by the right people.

## Build AI Literacy on a Trusted Foundation
If your organization has struggled with data literacy, you’re not alone. Now, you get a second chance. The rise of AI gives every business a new opportunity to get it right with tools that are built for scale, consistency, and context.

Cube Cloud brings structure to your AI initiatives by connecting them to the universal semantic layer your business trusts for consistent data and confident decisions. It’s not about more dashboards or better prompts. It’s about a universal understanding of your data, across every tool, team, and user. [Contact sales](https://cube.dev/contact) to learn more about how Cube Cloud can help you build AI literacy on a trusted foundation.


AI Literacy Is the New Data Literacy: Why a Semantic Layer Is the Missing Link

For decades, Business Intelligence (BI) was hailed as the holy grail of data-driven decision-making. Companies invested heavily, pouring time, money, and talent into BI platforms, data warehouses, and analytics teams. The goal was simple: democratize data. Make it accessible so that everyone, from top executives to frontline employees, could make smarter, data-backed decisions.

But the reality is far less inspiring. Despite massive investments and technological leaps, BI adoption has stalled. Research shows that only 25-35% of employees actively use BI tools in their day-to-day work. Even in organizations that pride themselves on being data-driven, many employees still rely on outdated reports, gut feelings, or inconsistent spreadsheets to make critical decisions.

What went wrong? As companies now turn their eyes to Artificial Intelligence (AI) as the next big thing, there’s a risk of history repeating itself. If scaling BI was tough, scaling AI will be even harder because of its complexity, higher stakes, and stricter governance requirements. Let’s explore the lessons from BI’s struggles and why they should be a wake-up call for organizations looking to scale AI successfully.

## Lesson 1: Data Access Isn’t Enough—Context and Consistency Matter
BI promised self-service analytics with tools that allowed business users to explore data without constantly relying on IT or data teams. To some extent, it delivered. BI tools made data more accessible than ever before, but accessibility isn’t the real problem.

The deeper issue is a lack of context and consistency. Data in most organizations lives everywhere, from cloud data warehouses to SaaS apps and legacy systems. Each source has its own definitions, structures, and quirks. Sure, BI tools can connect to them, but they do not unify them. No common access point exists to ensure that everyone uses consistent, governed, and accurate data.

![](https://ucarecdn.com/b4fda624-f8f1-4021-9ff2-25c7ce78eeea/-/quality/smart/-/format/auto/)

Semantic definitions are fragmented. Different teams pull similar reports and end up with conflicting numbers, with each convinced they are right. Filters are applied differently. Joins don't match. Calculations vary. Trust becomes eroded, and once trust in data disappears, so does tool adoption.

With AI, this problem only intensifies. If AI accesses inconsistent or inaccurate data, the consequences can be catastrophic, such as flawed outputs and bad business decisions. Because AI often operates as a black box, these errors might not even be noticed until it’s too late.

The lesson here is simple but crucial: It’s not enough to make data accessible. Organizations need a trusted data foundation, built in a universal semantic layer to ensure consistent definitions, unified metrics, and the right context, whether it’s feeding a BI dashboard or used for AI outputs. 

## Lesson 2: Complexity Overwhelms Self-Service Initiatives
BI tools evolved to be more user-friendly over the years with drag-and-drop interfaces, sleek dashboards, and even natural language querying. But for many business users, all the bells and whistles still aren’t enough.

Data is just too complex: joining data from multiple tables, applying the right business logic, and building meaningful KPIs. These tasks often demand a level of technical know-how that most business users simply do not have. So, what happens next? They fall back on data teams, defeating the whole purpose of “self-service” BI.

![](https://ucarecdn.com/0f35fb28-0716-4e36-9e10-a05a16a82a40/-/quality/smart/-/format/auto/)

AI takes this complexity and cranks it up a notch. Successful AI models need more than proprietary data. They require cleaned, enriched, and transformed data that is engineered for the task at hand. Without a simplified, governed approach to managing this complexity, AI initiatives will end up like BI, bottlenecked, dependent on overburdened data teams, and inaccessible to the broader organization.

A universal semantic layer abstracts complexity, delivering clean, consistent, and context-rich data to both AI and BI. It frees data teams from endless wrangling and lets them focus on high-value tasks.

## Lesson 3: Trust Is Fragile—And AI Raises the Stakes
One of BI’s silent killers is a lack of trust. It starts small as a dashboard showing the wrong numbers or a report that pulls outdated data. But over time, these small inconsistencies chip away at user confidence. Eventually, users abandon BI tools altogether, opting for offline, ungoverned spreadsheets. Sound familiar?

Now, apply that same trust issue to AI and multiply the stakes. AI models aren’t just powering dashboards. They’re driving decisions in high-stakes areas like finance and healthcare. One flawed AI output can have real-world consequences. If stakeholders don’t trust that AI is using accurate, unbiased data, they won’t use it.

![](https://ucarecdn.com/daefe8e4-ccc2-4c24-8498-9614401462d5/-/quality/smart/-/format/auto/)

Explainability adds another layer to this challenge. In BI, you can usually trace how a dashboard was built or how a metric was calculated. But with AI, it’s often a black box. Understanding why the model generated a specific response can be difficult to untangle. 

Building trust in AI requires a multi-pronged approach:
- Consistent, high-quality data throughout the AI pipeline.
- Robust governance and monitoring to detect data drift, bias, or anomalies.
- Explainability tools that demystify complex models and help users understand the “why” behind AI-driven decisions.

Just like BI, AI will fail to gain meaningful adoption without trust. A universal semantic layer makes proprietary data AI-ready so that it can deliver trustworthy outputs across a wide range of requests.

## Lesson 4: People and Processes Lag Behind Technology
Here’s a truth many organizations learn the hard way: Technology alone doesn’t solve problems.

Plenty of BI failures can be traced back to this flawed thinking when companies invest heavily in tools but neglect the people and processes needed to make them successful. They didn’t train users properly or establish governance frameworks to foster a data-driven culture. As a result, tools sit unused. Data stays siloed. Gut feelings drive decisions, instead of facts.

![](https://ucarecdn.com/fd6d0be3-0664-4b60-9a1c-91d300a4a2ac/-/quality/smart/-/format/auto/)

Scaling AI requires even more attention to people and processes. AI initiatives demand cross-functional collaboration. Data scientists, engineers, domain experts, and business stakeholders must work together to prepare data, train models, validate results, deploy solutions, and continuously monitor quality and performance.

Organizations that neglect the people and processes behind BI will find themselves in even deeper trouble when trying to scale AI. The fix is a holistic approach that balances technology investments with the right people, training, and governance structures operationalized within a universal semantic layer.

## Lesson 5: Speed Without Strategy Leads to Waste
In the rush to become data-driven, many companies adopted BI tools without a clear strategy. The results were messy with duplicated dashboards, conflicting metrics, fragmented implementations, and wasted resources.

The same risk looms with AI, only bigger. AI projects often start as pilots or experiments. But without a clear roadmap, they can spiral into disjointed initiatives that don’t align with business goals. Most organizations are struggling to move from AI experimentation to execution. Siloed models, inconsistent data pipelines, and a lack of governance lead to wasted time and money.

![](https://ucarecdn.com/e5fc5df7-1255-4802-9afa-b242e4e45d93/-/quality/smart/-/format/auto/)

To avoid this pitfall, organizations need a strategic approach to AI adoption:
- Align AI initiatives with business objectives and KPIs.
- Establish a trusted data foundation to ensure consistency.
- Implement strong governance for data, models, and outputs.
- Foster collaboration between technical and business teams to bridge gaps.

A more pragmatic approach ensures that AI isn’t just another shiny tool. AI should be a driver of real business value, and at the center of these initiatives should be a universal semantic layer.

## The Path Forward: Build a Strong Data Foundation for AI Success
Although BI’s struggles have taught us some tough lessons, they provide us a roadmap for scaling AI the right way. Consistent, governed, and accessible data is non-negotiable for any successful data initiative, whether it’s AI or BI. The key to achieving  success is a universal semantic layer.

By unifying data from multiple sources, standardizing business logic, and providing governed access to both AI and BI, a universal semantic layer creates a trusted data foundation. It ensures that everyone, humans and machines alike, works from the same consistent, reliable information.

As AI adoption accelerates, organizations will need to ask themselves: Are we building on solid ground? Or are we about to repeat the mistakes of BI? Because if scaling BI was hard, scaling AI will be even harder, unless you get the data right. 

By solving for key barriers for scaling AI initiatives, you can avoid repeating the struggles of scaling BI initiatives, all from one trusted place. Consequently, you can also improve BI adoption across the organization with unified data. With a universal semantic layer, models and metrics can be reused across AI, BI, spreadsheets, and embedded analytics solutions. [Contact sales](https://cube.dev/contact?utm_source=cube&utm_medium=blog&utm_content=post-gartner%20blog) to learn more about how Cube Cloud can help advance your AI and BI initiatives. 

Deep Dives

The Harsh Lessons of Modern BI and Why Scaling AI Will Be Even Harder

This year’s Gartner Data and Analytics Conference in Orlando delivered a powerful message: in today’s fast-paced AI landscape, enterprises must balance speed with governance, embracing a modular, open, and collaborative approach to data management. The Cube team returned with rich insights that resonate deeply with our mission to provide a universal semantic layer that unifies, democratizes, and secures data across organizations.

## Trustworthy Data and Robust Governance
A recurring theme at the keynote was the need for trustworthy data. In an era where data fuels AI journeys, establishing robust trust models and governance frameworks is more critical than ever. The session highlighted how contextualizing data through semantic layers not only enhances data trustworthiness but also drives consistency across analytics tools and AI agents.

At Cube, our universal semantic layer is designed with these exact challenges in mind. By serving as a single source of truth for [models and metrics](https://cube.dev/product/data-modeling), our solution empowers enterprises to set up robust governance while ensuring that data is both accessible and reliable. This alignment allows organizations to implement “guardrails” that enable innovation and experimentation—providing the freedom to fail safely and learn quickly.

## Balancing Governance with Speed
One of the standout messages from Gartner was the idea of “going slower to go faster.” In other words, investing time in solid data governance can accelerate long-term innovation. Enterprises are encouraged to build trust through carefully managed semantic layers, which ultimately speeds up decision-making and AI deployments. 

Cube’s approach reflects this philosophy. By moving [data access controls and policies](https://cube.dev/product/data-access-control) upstream into the universal semantic layer, we help companies achieve the delicate balance between control and agility, ensuring that even in a rapidly changing AI environment, every data-driven decision is backed by well-governed, high-quality data.

## Embracing Variety and Modularity in AI Platforms
The keynote also emphasized the value of variety in AI platforms—supporting a best-of-breed approach to rapidly evolving technology landscapes. In an environment that demands modular and open architectures, reusability becomes a significant competitive advantage. Data products, active metadata, and collaboration between IT, data analysts, and AI agents are not just buzzwords—they are essential building blocks for next-generation data platforms.

Cube’s universal semantic layer is inherently modular and designed for reuse. By enabling seamless integration across disparate systems and tools with [Data APIs](https://cube.dev/product/data-apis), our software supports a variety of use cases from data ops to fin ops, ensuring that all teams—from IT to business analysts—work with consistent, governed data. This flexibility is especially critical in a world where the CDAO role is evolving to meet the demands of AI, and cross-functional collaboration is key to unlocking the highest ROI on AI investments.

## Overcoming Cultural Roadblocks and Driving Collaboration
Despite all the technological advancements, a lack of data culture remains the number one roadblock for many organizations. The conference underscored the urgent need for leaders to foster an environment where data is seen as a collaborative asset. As companies invest AI dollars in platforms that democratize data access, the transformation of roles like the CDAO is both inevitable and necessary.

Cube is committed to helping organizations overcome these cultural hurdles. Our universal semantic layer not only standardizes data across the enterprise but also bridges the gap between data engineering and analytics teams. By allowing them to choose between code and [visual modeling](https://cube.dev/blog/announcing-visual-model-editor-in-cube-cloud), Cube Cloud provides a shared workspace for trusted data. This collaboration is crucial for creating an environment where data is not siloed but is a shared resource driving innovation and operational excellence. 

## Looking Ahead
The insights from Gartner’s keynote offer a clear roadmap for the future of data and analytics: invest in trustworthy data, balance governance with agility, and leverage modular, open architectures to drive AI success. At Cube, we’re proud to be part of this transformative journey, providing a universal semantic layer that underpins these best practices and empowers enterprises to harness the full potential of their data.

As we move forward, our commitment remains steadfast—to help organizations navigate the complexities of the modern data landscape with confidence, ensuring that every data initiative is both robust and agile enough to meet the challenges of tomorrow. [Contact sales](https://cube.dev/contact?utm_source=cube&utm_medium=blog&utm_content=gartner%20bickell%20blog) to learn more about how Cube Cloud can advance your AI and BI initiatives with trusted data.


Brian Bickell

Gartner Data and Analytics Conference 2025: Building Trust and Agility in the AI Era

### New pricing for the AI API

The cost per token of using LLMs has fallen substantially since we initially launched the AI API. Our initial pricing of one request per CCU was necessary at the time, but now with much more efficient models like Claude 3.5 Sonnet and GPT-4o available, we can offer a much lower price of 5 requests per CCU.

As of December 2024, the AI API is available for all paying Cube Cloud customers, including those on starter plans. You do not need to reach out to your AE or intercom to get it enabled; it is already enabled.

Here is a link to our AI API docs on how to use it: [https://cube.dev/docs/product/apis-integrations/ai-api]

The AI API needs to be enabled in your deployment in order to use it. Toggle to enable in your configuration flags:

![](https://ucarecdn.com/fe4b7a81-09b2-4894-946b-6faec35b980d/)

## Cube's AI API with Value Search
Cube's AI API is already a powerful way to allow users to access your data with AI quickly, safely, and transparently. We've heard from customers that users often refer to values inside of a dimension instead of stating where the dimension equals a specific value. This usually means the request won't be successful, and the user won't get the data they're looking for.

We have released a feature called Value Search for use with the AI API, which addresses this problem. Value Search allows Cube Cloud customers to select specific dimensions to be indexed so their values can be referenced by a query. Our RAG system can then use relevant values to answer questions that contain these dimension values or similar terms. It will include the pair of the dimension name and the value to filter the dimension.

Here is an example of how you can make a dimension searchable in your Cube data model YAML:
```yaml
- name: order_status  
  sql: order_status  
  type: string  
  meta:  
    ai:  
      searchable: true 
```

Here is our documentation related to the feature: [https://cube.dev/docs/product/apis-integrations/ai-api#value-search](https://cube.dev/docs/product/apis-integrations/ai-api)

## BYOLLM
Committed Cube Cloud customers can now choose to bring their own LLM to use with the AI API and incur costs this way instead of paying one Cube Cloud CCU per 5 requests. The AI API is compatible with Claude 3.5 Sonnet, available from AWS Bedrock, GCP Vertex, Snowflake, and Anthropic, OR GPT4o, available from OpenAI or Microsoft Azure. This enables almost anyone to use the AI API with their own LLM, whether for cost, security, or both.

Here are [instructions](https://cube.dev/docs/product/apis-integrations/ai-api#other-llm-providers) on how you can change the LLM provider for AI API use.

![](https://ucarecdn.com/4e143148-8c55-44d3-afb7-7dc7af214d7c/)

We are also offering free workshops for customers looking to put AI API into production until the end of April.

## Launching new Slack app for Cube AI API
Many of our customers have requested this feature since we launched the AI API. They expressed interest in enabling their stakeholders to interact with their Cube semantic layer through Slack using AI, similar to our AI API. The AI API ensures that questions are answered rapidly, safely, and transparently.

![](https://ucarecdn.com/2ffadc11-9ba0-4a9c-9b7b-0de4a44de9ed/)

You can install the Cube AI API Slack application easily in the settings page for AI API:

![](https://ucarecdn.com/9070ccb3-3fa9-4e04-aa38-8c4474250bb4/)

Once installed, the Slack app will introduce itself to your workspace individually, starting a conversation with each user. Users can then direct message the application with their data questions, which will be turned into threads where follow-up questions can be asked and feedback can be given. You can choose to receive a log of all questions asked and answered in a separate private channel for your engineering team to observe and learn from.

Want to learn more? [Contact our sales team](https://cube.dev/contact?utm_source=cube&utm_medium=blog&utm_content=2025%20api%20updates) today.

Make your AI Smarter - turbocharging Cube's AI API with value search; + BYOLLM and Slack app

This year, [Gartner’s Data & Analytics Summit](https://www.gartner.com/en/conferences/na/data-analytics-us) is focused on a pivotal theme: “From Inspiring Exploration to Leading Execution.” As AI, data, and analytics strategies continue to evolve, enterprise data leaders must move beyond experimentation to operational excellence, balancing opportunity and risk while delivering tangible value to their organizations.

At Cube, we understand that AI-readiness isn’t only about cutting-edge models and on-demand insights; it’s about having the right data foundation to drive meaningful organizational impact. Cube’s universal semantic layer unifies, governs, optimizes, and integrates enterprise data for both AI and analytics, ensuring a smooth transition from exploration to execution with reusable data models and metrics across all data consumers.

### Aligning AI, Data, and Analytics Strategies with Business Goals
Data leaders are well-aware of the challenges of adopting and scaling BI across their organizations. Now, they face increasing pressure to turn AI potential into business value, ensuring alignment with strategic goals while managing stakeholder expectations. However, many enterprises struggle to bridge the gap between strategy and execution, often hindered by siloed, inconsistent data across their organization. 

With Cube Cloud, data leaders can confidently drive AI adoption while maintaining alignment with organizational goals and delivering measurable business impact. Cube’s universal semantic layer unifies siloed data and ensures business users, analysts, and AI applications work from a single source of truth. As a result, your data teams will realize faster time-to-value, and every level of the organization will have access to consistent, governed data for informed decision-making.

### Making Data AI-Ready for Your Organization
Enterprises are quickly learning that most AI initiatives fail to scale due to data readiness issues rather than AI model limitations, as many struggle to shift from testing to production. Context is key to AI success. Without structured, contextualized, and governed data, AI outputs are prone to inaccuracies, bias, and operational inefficiencies. 

Cube Cloud ensures your AI initiatives can move beyond proof-of-concept and scale confidently across the enterprise, providing reliable and trustworthy results. With AI-ready data, you can ensure AI can interpret and act on business data accurately. Data can be aggregated and optimized for real-time AI inference, improving efficiency and reducing operational costs. With Cube’s commitment to interoperability, integration with AI frameworks and other platforms powers AI-driven analytics and automation.

### Delivering Trusted Insights Across the Enterprise
Enterprises must navigate an increasingly complex AI and analytics landscape, ensuring data is not only available but also accurate, consistent, and actionable. You must be able to share and reuse the same data models and metrics across data consumers. Without a unified semantic layer, inconsistent metrics and siloed data sources lead to unreliable AI outputs. 

Cube empowers enterprises to deliver trusted insights at scale, driving informed decision-making across all business functions. Cube Cloud provides centralized metrics management, which allows you to define KPIs once and ensure consistency across data consumers. Additionally, API-driven data access delivers data to any application, from cutting-edge AI to modern OLAP in spreadsheets. Finally, broad integration capabilities enable seamless analytics experiences across platforms like Power BI, Tableau, and custom embedded analytics applications.

### Unifying and Optimizing Enterprise Data
Effective data management is critical for scaling analytics and AI initiatives. However, enterprises often grapple with fragmented data definitions and performance issues across multiple cloud data platforms and data consumers. These discrepancies shift attention away from improving metrics and outcomes to explaining how the metric was calculated, while slow query responses degrade user experience.

Cube Cloud acts as the bridge that connects raw cloud data to meaningful business insights that accelerate data-driven innovation. Cube Cloud consolidates fragmented data definitions into a centralized, governed layer accessible by all data consumers. It also implements intelligent caching and pre-aggregation strategies to enhance query performance and reduce cloud costs. Cube Cloud empowers data teams with simplified access to governed, fast, AI-ready data.

### Ensuring Security, Governance, and Compliance
As enterprises scale their AI and analytics efforts, ensuring data governance, security, and compliance is a top priority. Inconsistent governance policies can lead to regulatory risks, data breaches, and reduced trust in AI-driven decisions, which is not easily recovered once it is lost. 

Cube provides enterprises with the flexible governance they need to scale AI and analytics with confidence and compliance. Using robust security controls, Cube Cloud implements role-based data access policies, user-dependent query rewrites, and data masking to ensure sensitive data is protected. For auditability and compliance, it provides automated data lineage from source to consumer. Cube Cloud enforces governance rules to ensure every AI and BI tool accesses the same, trusted data.

### Meet Cube at Gartner’s Data & Analytics Summit
As enterprises transition from AI exploration to execution, Cube Cloud stands ready to deliver unified, governed, AI-ready data. By solving for key barriers for scaling AI initiatives, you can also solve for the key barriers for scaling BI initiatives too, all from one trusted place.

Let Cube help you turn your AI, data, and analytics strategies into tangible business outcomes. See you at the Gartner Data & Analytics Summit!

- Visit Booth #931 to take the “Is Your Data AI-Ready?” assessment.
- Attend our theater session: "Cube: Scaling AI with a Universal Semantic Layer" [public link](https://www.gartner.com/en/conferences/na/data-analytics-us/sessions/detail/3859199-Cube-AI-Teammates-for-Data-Analytics-Teams) | [app link](https://nav.gartner.com/#/en/navigator/BI23/whoshere/exhibitors/3859199/sessiondetail/speakers-tab?profilematch=false&sortedby=exhibitorname) on Wednesday, March 5, 2025 at 1:05pm in Theater 3.
- Learn more at [Cube's Gartner Data & Analytics Summit Hub](https://learn.cube.dev/gartner-2025).

Go From AI Exploration to Execution with Cube Cloud

After ChatGPT was released in late 2022, almost everyone in data and analytics rushed to leverage GenAI models for text-to-SQL solutions. After two years, it became clear that text-to-SQL alone was not simply enough.

Enterprises need intelligent AI teammates to amplify their data and analytics teams' productivity. This is part of a larger trend as enterprises adopt [agents](https://www.salesforce.com/agentforce/) and [digital teammates](https://www.mayfield.com/ai-teammates/) across different roles and functions in the organization. Future AI data engineers will work alongside human data teams to carry out tasks such as building data assets, investigating ongoing issues, and optimizing costs.

For an AI agent to be a helpful teammate, it is not enough to take a natural language query as input and output a SQL statement. AI agents for data and analytics need a comprehensive understanding of the existing underlying data assets, from the raw data through transformations and semantic modeling to reporting. More importantly, they need the ability to make changes to these assets as they reason and go through the chain-of-thought process.

For example, when the agent is tasked with adding a new report to the existing dashboard, it needs to evaluate the existing semantic model to see if the current model can support the requested report. If it lacks dimensions or measures for the required report, the agent needs to go into the transformation and raw data layers to see what is available and what needs to be built or modified. It would investigate, build, test, validate, and collect feedback from the human as it progresses through the task to finally deliver the requested report.

## Use cases
AI data engineers can perform a variety of tasks, but the deliverables typically come down to either the creation of a new data asset or the modification of an existing one. The agent can modify extract-load pipelines, transformations, semantic models, and, ultimately, dashboards and reports.

The value added to the business is increased productivity. AI will level up every data professional: senior data engineers will be able to do more within the same timeframe as before, and junior data engineers will be able to tackle more challenging tasks. It will also enable less technical team members to contribute to areas they couldn't before, e.g., dashboard designers contributing to data transformation pipelines.

Agents will be able to handle most of the use cases data teams are working on today – creating or modifying reports and dashboards, optimizing queries from time and cost perspectives, debugging data issues, refactoring codebases, etc.

## Architecture
We still need to build many basic blocks to enable this future. Some of them would require advancing the fundamental enabling technology - LLMs, while others would require infrastructure for AI agents to retrieve, understand, and modify data assets.

### Advanced reasoning 
In September last year, OpenAI [introduced o1](https://openai.com/index/introducing-openai-o1-preview/), a new series of models with a reasoning architecture that spends more time thinking about problems. This architecture and chain-of-thought approach are prerequisites for complicated AI agents that can complete sophisticated tasks, such as making changes to the organization's data assets. We are still in the very early stages of this next frontier in AI, but the development pace looks very promising.

AI data engineers will rely on reasoning and chain-of-thought processes as they receive input from humans, delve into semantic models, transformations, and the data itself to ultimately perform actions for the given task.

### Data Assets as Code
The data and analytics field has experienced a massive trend in recent years of applying software engineering practices to data management. This has led to the rise of code-first workflows as the primary way to manage data assets. Code-first workflows provide many benefits for humans, such as collaboration, version control, and CI/CD, but they are absolutely necessary for AI agents.

AI agents consume code as input and produce it as output. This will become the major interface layer between the agents and data tools for managing transformations, semantic modeling, and BI-as-code data assets.

AI agents will work on multiple tasks simultaneously. They will create branches within the version control system, spin up containers, make changes to the codebase, run CI, and present results to humans. Humans will review code changes and provide feedback through the code review process. This way, code-based workflows being used by data teams today can seamlessly integrate new AI teammates.

### Agents Workspace
As humans need laptops and access to data tools, AI agents would need containers to execute workflows and API access to data systems. AI data engineers will write code faster than humans do today. This would set a high bar for infrastructure.

On one end, it would need to scale up massively as agents can work on thousands of threads simultaneously. On the other end, it should scale down to zero once agents are done with their work. The speed of provisioning containers for AI agents would be crucial.

### Interoperability of the existing data tools
Human data teams use many different data tools in today's workflow: extract-load, transformation, warehousing, semantic modeling, cataloging, observability, visualization, etc. Agents would need a way to programmatically manipulate data asset states within these systems. I believe code will become the universal way to manage the state of data assets across the board, and data tools will integrate with existing version control and CI/CD systems to enable development-deployment cycles for both humans and agents on data teams.

At Cube, we are excited about the future of AI agents for data teams and the opportunity to contribute to making it happen. We're investing in building the fundamental blocks to enable AI agents to manage data assets across the data stack. Please [reach out](https://cube.dev/contact?utm_source=cube&utm_medium=blog&utm_content=ai%20data%20analyst%20blog) if you would like to discuss the future of data and analytics!

Artyom Keydunov

AI Data Analyst

Data-driven decision-making is crucial for business success, but organizations face a growing challenge of complexity and data governance. These challenges make it difficult to access data in a unified way.

In [Part 1](https://cube.dev/blog/exploring-the-semantic-layer-through-the-lens-of-mvc), we explored the semantic layer through the lens of MVC, and in [Part 2](https://cube.dev/blog/universal-semantic-layer-capabilities-integrations-and-enterprise-benefits), we outlined its benefits. In this final piece of the series, we examine the integration of a semantic layer with artificial intelligence and why it might be the best place to start with GenAI. Let's explore the integration of semantic layers with existing data infrastructure and AI.
## The Power of Combining Semantic Layers with AI
The semantic layer is the ideal foundation for natural language queries, providing a **structured bridge** between data sources and business requirements with AI capabilities. Its declarative definition of metrics, comprehensive data models, and metadata management enable the large language model to understand and interpret the business context accurately. This promises higher-quality results from natural queries, human interaction, and data insight.

Generative AI makes natural language queries accessible for discovering complex business insights through simple questions. Cube is pioneering natural language querying, especially with the [acquisition](https://cube.dev/blog/investing-in-cubes-ai-driven-future) and integration of DelphiLab, among other things. It makes **data accessible** to more non-technical users and enhances data exploration, benefiting from faster insights without the technical bottlenecks. 
## How Semantic Layers Enable Natural Language Querying
How does it work? If we start with the top-down approach, we have an instruction - a prompt that gets augmented. It uses existing database schemas and data, queries them with SQL to get the answer, and finally transpiles it back to text. This would look something like this:

![](https://ucarecdn.com/683e0474-0330-43e5-b914-1aa8c0e548eb/)
An example from instruction to answer | More on [Natural language AI queries drive customer analytics](https://cube.dev/case-studies/customizable-embedded-dashboards-and-natural-language-ai-queries)

Many different interfaces are involved, and using natural language isn't as simple as it looks. So, the question is, how do they work?
### Architecture and Components
A simplified architecture that helps Cube transform **text-to-SQL via Cubes API** based on retrieval augmented generation (RAG). Cube will make sure this process is **deterministic**. It uses metadata from the semantic layer, and the data catalog looks like this with the data warehouse holding our data and Cube our semantics about the data, allowing Cube to interact with text with the help of a generative large language model such as the models from Open AI:

![](https://ucarecdn.com/3ddb7478-f75d-4ef7-9c86-0220581c120d/)

There are more components involved. Below are the critical components for generative and natural interaction with a semantic layer to get a better understanding:

- **Semantic layer** (like Cube) - providing declarative metric definitions, data modeling, and metadata
- **Large language models** (like OpenAI) and natural language response generators are used for natural language processing, generation, and interpreting results into human-readable text
- **RAG** (Retrieval Augmented Generation) - to extend internal business knowledge and context
- A **[Semantic Catalog](https://cube.dev/docs/product/workspace/semantic-catalog)** lets users search, understand, and reuse trusted data products from Cube and external sources such as cloud data warehouses and business intelligence tools.
- **API and SQL Transpiler**: The Cube API compiles user prompts deterministically to SQL. With the intermediate API, Cube adds constraints for the request format and reading from semantic layer artifacts to enable simpler output generation for the AI model.
- **Query execution and planner engine** - running the SQL and returning results
- **Cache layer** - for query optimization and performance
- **Access control & governance** - managing security and permissions
- **Visualization/dashboard tools** - for displaying results graphically

A lot is going on until we have our intuitive human interface, and many topics are constantly improving and changing. Luckily, Cube has integrated all of this. If you'd like to try it, try it and [get started](https://cube.dev/docs/product/getting-started). There is an OSS and a cloud version; some of these features are available only.

> 📝 Evolution of the Semantic Layer
> 
> To get an even better overview, here is a short recap of the evolution we've gone through until we arrived where we are today (long story [here](https://airbyte.com/blog/the-rise-of-the-semantic-layer-metrics-on-the-fly#the-history-of-a-semantic-layer)):
> 1. SAP BusinessObjects Universe and BI semantic layer (1991)
> 2. SSAS and MDX, with their logical modeling layer with MDX, define business metrics and dimensions in a structured way (1997)
> 3. Kimball discussed the concept of a semantic layer in [#158 Making Sense of the Semantic Layer](https://www.kimballgroup.com/2013/08/design-tip-158-making-sense-of-the-semantic-layer/) , around 2013
> 4. Maturing BI tools with an integrated semantic layer, such as Tableau, TARGIT, PowerBI, Apache Superset, etc., have their own metrics layer definition around 2016
> 5. Looker and LookML popularized as the first semantic layer around 2019
> 6. Declarative Semantics and metrics with tools such as Cube, MetriQL, MetricFlow, Minerva, and dbt arose with the explosion of data tools around 2022
> 7. The evolution of data querying and natural language queries with the integration of LLMs and catalogs around 2023/24

## What are AI and Semantic Layer Use Cases?
After we understand how it works and what is needed under the hood, let's explore typical use cases with semantic layer and AI. The best-suited use cases for a semantic layer involve centralized metrics definition, unified data modeling, query optimization with a cache layer, and security and access control governance.
### Typical Use Cases
The chosen use cases below illustrate the power and suitability of a semantic layer, among many others:

> 💡 AI-Powered Semantic Layer Use Cases
> Key scenarios where semantic layers enhance AI and natural language capabilities
>
> **🗣️ Natural Language Data Exploration**
> - Transform business questions into accurate SQL queries through LLMs
> - Context-aware query generation using metadata and business definitions
> - Extended context through RAGs with internal systems like CRM, ERP
> - Real-time data exploration through conversational interfaces
> - Automated insight discovery and anomaly detection
> - Enhanced data accessibility for non-technical users
> 
> **🎯 Rapidly Creating Business Dashboards through GenAI**
> - Quick response to new data sources or management report requests
> - Requirements engineering & business needs validation
> - Dimension/fact definition with the appropriate [granularity](https://en.wikipedia.org/wiki/Granularity)
> - Dashboard-oriented metric building
> - Accelerated transition from business request to functional dashboard
> 
> **🎯 Unified Data Governance**
> - Ideal for both fresh starts and consolidating fragmented data landscapes
> - Seamlessly connects siloed departmental data sources into a cohesive structure
> - Easy [integration](https://cube.dev/docs/product/configuration/data-sources) of **heterogeneous data sources** (internal & external)
> - Natural language and intelligence metric management
> - Automated discovery of related metrics and dimensions with API abstraction
> - Unified access with in-build features: [Query pushdown](https://cube.dev/blog/query-push-down-in-cubes-semantic-layer), caching optimization
> - Organization-wide [DRY](https://en.wikipedia.org/wiki/Don%27t_repeat_yourself)
>
> **🤖 AI Chatbot Integration**
> - Embedded analytics with AI chatbot solutions
> - Customizable dashboard implementations
> - Natural language query capabilities
> - Real-world case studies:
>   - [Embedded Analytics and AI Chatbot in one Semantic Layer](https://cube.dev/case-studies/embedded-analytics-and-ai-chatbot-on-one-semantic-layer)
>   - [Customizable dashboard implementations and natural language AI queries to drive customer analytics](https://cube.dev/case-studies/customizable-embedded-dashboards-and-natural-language-ai-queries)
> 
> Find more on [Customer Stories](https://cube.dev/case-studies).

### Copilot
Unfortunately, every organization is largely complex. Therefore, AI has difficulty producing the correct measures and data models. With generative AI we can further automate this complex business process and make it more accessible for non-technical users, especially those who know the business inside and out. These domain experts have worked in the industry for a long time.

A semantic layer has an edge here with the combination of **declarative metrics definitions** and the **know-how of business expert**, we have chance to enhance the model with the critical business context from internal systems such as ERP, CRM, or any others; and iterate quickly with more concise and correct output.

With Cube, we can generate the first iterations of the data model and measures, joins, etc. with **[Cube Copilot](https://cube.dev/blog/introducing-cube-copilot-your-new-partner-in-building-semantic-layers-with)** for example. It can suggest metrics in real-time based on the current context, empowering data practitioners working with business requirements and SQL for greater productivity.

![](https://ucarecdn.com/e2a73a45-e030-47f2-b008-205100e5fabe/)
Cube Copilot in action, where you define your needed measure as a comment

## AI-Powered Semantic Layer and Its Future
Looking ahead to what might be possible, we see that AI is here to stay. So, the question is, how can we use this power in our everyday lives? The semantic layer is one of the first areas I would start experimenting with; as we have seen, the **intersection** between data sources and the business already offers excellent opportunities with Copilot, AI assistant, and many more features.

But what else would be possible? We could ask our questions directly to our data apps to generate the dashboard we'd like to see, and AI would take over and create everything autonomously and entirely correctly. This will still take time. However, as assistants to humans, we supervise the outcome (as we did with the stewardship of master data management), and we can approve or decline generations as we do with code generations and other applications of AI today.

However, the most significant future that GenAI with a semantic layer will enable is much **faster iterations**, making it more approachable for business experts and feeding into more accurate outcomes, fed with more context than any other tool in the data stack and adding sub-second response time to my data warehouse with an OLAP cube featuring a solid business context and AI assistant.

Above all, enhancements in **natural language querying** will improve daily, and data lineage and governance will improve along the way with quality checks and automatic tests.

Besides the challenges of implementing a central tool for metrics or using it for a small landscape, the evolution shows that the need for semantics and a universal layer to define business logic has been here for a long time and that a new semantic layer with new generative capabilities will evolve even further.

## Conclusion: The Path Forward
In closing this three-part series, we've taken a comprehensive journey through the evolution and potential of semantic layers. We began by comparing the [semantic layer to a Model-View-Controller pattern](https://cube.dev/blog/exploring-the-semantic-layer-through-the-lens-of-mvc), explored the [key benefits and components of a universal semantic layer](https://cube.dev/blog/universal-semantic-layer-capabilities-integrations-and-enterprise-benefits) with centralized metrics definition, and finally demonstrated how semantic layers are becoming critical tools for implementing natural language integrations due to their unique position bridging business logic and technical implementation.

Integrating AI with semantic layers represents a new way organizations can interact with our data. By providing a unified data modeling layer, optimized query performance, and robust governance, semantic layers are positioning themselves as essential components in the data stack; creating alignment between business and technical teams while ensuring data consistency and governance.

Looking ahead, the combination of semantic layers with AI capabilities promises to democratize data access while maintaining control and accuracy. I'm sure this evolution will continue to enhance business intelligence capabilities, making data more accessible to non-technical users while preserving the rigorous standards required for enterprise-scale operations.

---

Thank you for following all three parts of this series. We'd love to hear your thoughts, experiences, or questions about implementing semantic layers in your organization. What topics would you like to see covered next? Please share your feedback and reach out to me or Cube directly.

Those interested in learning more about Cube's semantic layer capabilities can visit [Cube.dev](https://cube.dev) or check out the documentation to [get started](https://cube.dev/docs/product/getting-started).

Simon Späti

Semantic Layer and AI: The Future of Data Querying with Natural Language

Embedded analytics has long been a critical feature for delivering insights within applications, empowering users to make data-driven decisions without leaving their workflow. Now, Generative AI is reshaping expectations, enabling users to interact with data in more intuitive and conversational ways. The lines between analytics and AI are blurring, creating a new frontier for product leaders: Should you build or buy when adding these transformative capabilities into your solution?

At Cube, we believe the answer is clear—__*build your composable embedded analytics and Generative AI solution to your unique requirements*__. By choosing Cube Cloud, you can create tailored [embedded analytics](https://cube.dev/use-cases/embedded-analytics) and [Generative AI](https://cube.dev/use-cases/llm-and-ai-semantic-layer) experiences that align perfectly with your audience’s needs. You can even monetize extended capabilities beyond your data product, such as Data-as-a-Service with Cube’s extensive suite of APIs and integrations. While buying off-the-shelf solutions may seem quicker, it often results in compromises that limit your product’s impact. Here’s why building with Cube Cloud is the smarter choice for your embedded analytics and Generative AI solutions.

## The Limitations of Buying Embedded Analytics and Generative AI Solutions
Modern BI platforms and pre-built Generative AI solutions offer a tempting promise of rapid deployment of analytics and AI capabilities without significant engineering effort. However, their one-size-fits-all approach often introduces challenges that can derail your product strategy:

- __Too Much or Too Little Functionality:__ BI platforms and AI tools are designed to cater to a wide range of users. This means they often include features your users don’t need or lack the specific capabilities your application demands, leaving gaps you can’t fill without additional customization.
- __Fragmented Experiences:__ Buying embedded analytics and Generative AI capabilities from different vendors can create new data silos from disjointed workflows. Users may have to navigate separate interfaces for analytics and AI-driven interactions, creating friction and diminishing your product’s value.
- __White-Label Challenges:__ Even with white labeling, pre-built solutions rarely feel native to your product. The experience can appear bolted on, undermining your brand identity and leaving users questioning its authenticity.
- __Lack of Control and Scalability:__ Pre-packaged solutions often lock you into a vendor’s ecosystem, limiting your ability to be agile and adapt as your product grows or your users’ needs evolve.

## The Case for Building Embedded Analytics and Generative AI with Cube Cloud
Building your own embedded analytics and Generative AI capabilities allows you to create a seamless, scalable, and brand-aligned experience for your users. With Cube Cloud as your foundation, you gain the flexibility to innovate without compromise.

- __Future-Ready Data Stack:__ Cube’s universal semantic layer provides a unified, governed, and optimized data foundation for both analytics and AI. This ensures your solution can scale as your data grows and your AI needs evolve, without being tied to a vendor’s roadmap.
- __Unified Functionality:__ With Cube, embedded analytics and Generative AI capabilities can be integrated into a single, cohesive user experience. Users don’t have to switch between tools to access insights or interact with data—they get everything they need, right in your application.
- __Tailored User Experiences:__ Cube Cloud enables you to design analytics and AI workflows that are fully aligned with your product’s purpose and your audience’s needs. Whether it’s crafting interactive dashboards, enabling conversational data exploration, or providing predictive insights, you’re in control.
- __Brand-First Integration:__ Building with Cube keeps your product at the center of the experience. Embedded analytics and AI workflows feel like a natural extension of your application, enhancing your brand rather than diluting it.

## How Cube Cloud Powers the Next Generation of Embedded Capabilities
Cube Cloud is more than a tool for building embedded analytics; it’s the foundation for delivering transformative data experiences. As a [universal semantic layer](https://cube.dev/use-cases/semantic-layer), Cube bridges the gap between your cloud data platform and the capabilities you want to offer, while enforcing governance and security for every data consumer.

![](https://ucarecdn.com/d0142439-c019-49fa-812d-421f007fd884/)

- __For Embedded Analytics:__ Cube unifies and optimizes your data, ensuring real-time insights are accurate, consistent, fast, and ready for consumption in your purpose-built application.
- __For Generative AI:__ Cube provides AI models with clean, trusted, and contextual data, enabling conversational data exploration, intelligent recommendations, and personalized predictions—all seamlessly integrated into your application.
- __For Data-as-a-Service:__ Cube allows you to monetize your data in new ways by providing on-demand data access for external users through Instant APIs. Now you can support both ad-hoc analytics and tool choice from BI platforms to spreadsheets, without your data engineering team’s involvement.  
This unified foundation means you don’t have to choose between analytics and AI. Cube Cloud supports both, empowering you to build a cohesive, cutting-edge experience for your users.

## Why Build, Not Buy? 
Embedded analytics and Generative AI are no longer separate considerations—they’re part of a unified user experience that drives engagement and delivers value. Buying pre-built solutions might promise speed, but it forces you to compromise on functionality, flexibility, and user experience.

Building with Cube Cloud gives you the power to innovate on your terms. You can create tailored solutions that integrate seamlessly into your application, scale with your growth, and set your product apart in a competitive market.

Don’t limit your vision to what pre-built solutions offer. Build with Cube Cloud and deliver the future of embedded analytics and Generative AI today. Ready to build smarter with Cube Cloud? [Contact sales](https://cube.dev/contact?utm_source=cube&utm_medium=blog?utm_content=embedded-analytics-gen-ai-blog) to learn more about Cube’s embedded analytics and Generative AI capabilities.

> Interested in learning more? Register for our webinar, __[Stop the AI Hallucinations: Giving Context to Gen AI with a Semantic Layer](https://cube.registration.goldcast.io/webinar/41f5706b-6c97-4a16-a191-3a0f2e248a42?utm_source=cube&utm_medium=blog&utm_content=embedded-gen-ai-blog)__.

Embedded Analytics

Why Building Embedded Analytics and Generative AI Solutions with Cube Cloud Is a Winning Strategy

Data engineers are constantly seeking tools that not only simplify their workflows but also enhance their ability to deliver robust data models to meet the demands of the business. Today, we're excited to announce the release of Cube Copilot, a groundbreaking assistant designed to empower data engineers to build and manage the universal semantic layer more efficiently than ever before.

## Empowering Data Engineers for Greater Productivity
Implementing a universal semantic layer is a foundational step in structuring and interpreting complex data. It bridges the gap between raw data and meaningful insights, enabling organizations to make informed decisions. However, building a semantic layer isn't always straightforward. It requires knowledge of various objects to construct and an understanding of the precise formats to declare them in.

With Cube Cloud, data engineers have access to a powerful open semantic layer platform that offers a plethora of features to model, aggregate, and serve data efficiently. But let's face it—keeping track of all the possible Cube objects and mastering the exact syntax of Cube's data modeling language can be challenging, especially for those who aren't working with it daily.

## The Challenge: Mastering Cube Modeling Language Without Daily Practice
Cube Modeling Language is a declarative, versatile language tailored for defining data models in Cube. It allows you to describe dimensions, measures, joins, segments, and more. While its flexibility is a strength, it also means there's a lot to learn and remember. For data engineers new to Cube or those who only dip into Cube Cloud occasionally, recalling every detail of the syntax or every available feature isn't always feasible.

Imagine you're trying to define a new measure or dimension but can't quite remember the exact syntax. Or perhaps you're aware that Cube supports a certain functionality, but you're unsure how to implement it in your data model. The natural solution would be to consult the documentation, but this interrupts your workflow and consumes valuable time.

## Help Is at Hand: Meet Cube Copilot
This is where Cube Copilot comes into play. We're introducing an intelligent assistant that's integrated directly into your Cube development environment. Cube Copilot is designed to help you edit and expand your data models faster.

Cube Copilot streamlines the coding process, allowing you to focus on designing effective data models. By offering contextually relevant suggestions, Cube Copilot helps minimize syntax errors and mismatches in your data models. Instead of searching through documentation, now you can spend more time building. 

For new users of Cube, Copilot accelerates the onboarding process. By providing guidance and suggestions, it helps you learn Cube organically as you work. Even experienced Cube users can benefit from Copilot by discovering new features or more efficient ways to implement functionality within their models.

### How Cube Copilot Enhances Your Workflow
As you work on your data model, Cube Copilot provides real-time suggestions based on the current context. Whether you're defining a new measure or adjusting a join, it offers relevant code snippets and syntax guidance to keep you moving forward smoothly.

Cube Copilot is aware of your database schema, including column types and existing tables. This means it can suggest appropriate dimensions and measures that align with your data's structure, reducing the likelihood of errors and enhancing model accuracy.

It doesn't just consider what's in the database; it also takes into account the existing parts of your data model. By understanding how you've structured your models so far, Cube Copilot can provide suggestions that are consistent with your current approach.

We've equipped Cube Copilot with numerous examples of how to build various Cube objects. This rich repository allows it to offer informed suggestions that reflect best practices and advanced techniques.

### Direct Prompting with Comments
One of the standout features of Cube Copilot is its ability to respond to natural language prompts. Simply write a comment in your code describing what you'd like to create, and Cube Copilot will generate the corresponding code for you. For example:

```
// Create a measure for total sales that sums the amount column
```

Upon entering this comment, Cube Copilot will provide the code necessary to define the `total_sales` measure, saving you time and reducing the need to look up syntax.

## Get Started with Cube Copilot Today
We're thrilled to bring Cube Copilot to our global community of data engineers. It's more than just an assistant; it's a partner that enhances your abilities and streamlines your workflow. Whether you're a seasoned Cube user or just starting your journey to unified data, Cube Copilot is here to support you every step of the way.

## Conclusion
Building a robust semantic layer is now more accessible than ever with the introduction of Cube Copilot. By simplifying the process of writing code and providing intelligent, context-aware assistance, we're confident that Cube Copilot will become an indispensable tool in your data engineering toolkit. Try Cube Copilot and experience a new level of productivity and efficiency in creating semantic layers. [Contact sales](https://cube.dev/contact?utm_source=cube&utm_medium=blog&utm_content=copilot%20blog) to learn more.


All Things Cube

Introducing Cube Copilot: Your New Partner in Building Semantic Layers with Ease

We’re pleased to be featured in the July 10, 2024 issue of insideAI NEWS, formerly insideBIGDATA. The revamped publication offers comprehensive coverage of AI  technology and implementation strategies across multiple industries, including healthcare, security, financial services, automation, and agriculture. Below is a summary of our article, **“The Rise of Embedded Analytics and Embedded AI: Proportional Business Value at Last?” [Click here for the full text](https://insideainews.com/2024/07/10/the-rise-of-embedded-analytics-and-embedded-ai-proportional-business-value-at-last).**

Although business intelligence (BI) solutions have gone a long way to make more data available to non-technical users, challenges remain. Business users often give up when trying to find the right data because of the proliferation of data sources and dashboards. This is where curated data experiences — embedded analytics and embedded artificial intelligence (AI) can help — not just for customers and external use cases but in internal, user-facing applications, too.

## Access data where you need it with Embedded Analytics
With embedded analytics, users can access data when and where they need it most: in the context of their workflows. Instead of viewing a dashboard in the analytics platform and switching to the business application to act on it, businesses can curate the analytics experience within an application or a custom solution so there’s less jumping among applications and more immediate relevance. Taking it one step further, organizations can drive the experience using embedded AI so that users can bring up the right data within the business application using a simple voice command or chatbot.

But before organizations can deliver trusted embedded analytics and embedded AI, they need a universal semantic layer—an independent yet interoperable translation layer—between the data repository and the data-consuming endpoints. The semantic layer provides a consistent and trusted view of unified data by organizing, simplifying, and accelerating its consumption. Once a universal semantic layer is in place, curated data experiences are relatively easy to deliver, internally and externally. Although many organizations start with customer-facing applications, it is worth mentioning that substantial value can be derived quickly from focusing on internal processes first.

## AI and embedded analytics transform how data is used in an organization

There are many examples of internal processes that can benefit from embedded analytics and AI, including better employee experiences, decision-making in context, and better workflows and processes. When paired, a universal semantic layer and AI unlock profound capabilities in informing business users in real-time and context. For example, a universal semantic layer makes it possible to embed AI-assisted analytics into a tool like Salesforce, allowing for analysis on deals, prospects, and other key metrics to be done without context-switching — and via a near real-time process that can be as easy as querying an AI chatbot.

AI and embedded analytics powered by a semantic layer transform the way data is used in an organization, replacing the conventional, frequently fragmented method with one that is more integrated, perceptive, and useful.


Lissa Daniels

Data is the new oil, but it must be refined

As data continues to grow in importance and complexity, Cube Cloud's latest innovations, built on the [universal semantic layer](https://cube.dev/use-cases/semantic-layer), ensure that enterprises can connect the entire organization with the right data. These advancements promise to simplify the way businesses discover, explore, and access modeled data, making it more accessible to everyone. 

With the introduction of [Semantic Catalog](https://cube.dev/product/semantic-catalog), Cube Cloud extends its utility even further to data analysts and business users within your organization. Let’s take a closer look at how Cube’s universal semantic layer unlocks the value of trusted data.

## Semantic Catalog: Search a Unified View of Connected Data Assets
Semantic Catalog provides a comprehensive, unified view of connected data assets that is essentially self-documenting and always up-to-date, based on the data models you define. Users can effortlessly search through data modeled in Cube Cloud, explore downstream connected BI content, and examine upstream tables. 

This capability allows users to gain insights into the lineage, entities, relationships, dimensions, and metrics of their data. Additionally, users can visually navigate upstream connections to source tables and downstream connections to charts and dashboards across all BI platforms using a Data Graph. 

By providing a unified view, Semantic Catalog reduces duplicate efforts in data modeling and analysis, fostering better collaboration between data engineers and data analysts. This unified approach not only increases productivity but also takes the guesswork out of who to ask or where to go to find trusted data.

![](https://ucarecdn.com/051e290c-e12f-465e-af2a-2ee851462c13/-/quality/smart/-/format/auto/)

## Cube’s Vision for Data-Driven Decision Making

For the first time, Cube's universal semantic layer is not just a tool for data engineers but an invaluable resource for everyone else in your organization. Administrators can define roles with specific permissions for Semantic Catalog and [Playground](https://cube.dev/docs/product/workspace/playground) access and apply those roles to users, creating a tailored user experience for non-data engineering users. 

Whether you are a data analyst searching for data to start your analysis or a business user seeking instant insights, the new capabilities in Cube Cloud ensure that everyone can participate in the data-driven decision-making process.
<Quote authorName="Artyom Keydonuv" authorAvatar="https://media.graphcms.com/mvIb32XHRoibCutwm7WA"  authorRole="CEO & Co-Founder, Cube"> As the demand for data grows, delivering value consistently, efficiently, and at scale becomes more challenging. Cube Cloud’s AI-driven capabilities enable more people throughout the organization to make more informed decisions and derive value from the single source of truth the universal semantic layer provides. When it comes to AI, the universal semantic layer is the critical piece of the puzzle to ensure accuracy without hallucinations. </Quote>

Cube is making it possible for a broader audience within organizations to benefit from data-driven insights. Data analysts can now interpret data more consistently, while business users gain a better understanding of the data, facilitating better, faster business decisions. Combining [AI with the universal semantic layer](https://cube.dev/blog/semantic-layer-the-backbone-of-ai-powered-data-experiences) to provide context, you can ensure accuracy and reduce hallucinations. 

## Transforming Data Interactions

While data continues to grow in importance and complexity, Cube Cloud's innovative features ensure that enterprises can unlock the full potential of their data assets and their people. By making modeled data easy to discover, explore, and access, Cube Cloud empowers organizations to deliver value to everyone from it. 

[Contact sales](https://cube.dev/contact "Contact sales") to explore the new Cube Cloud features today and transform the way your organization interacts with data.

Unlock the Value of Trusted Data with Semantic Catalog

## Discoverability and accessibility at the point of data consumption
### Why the Semantic Layer is the perfect place for a user-focused catalog

I’ve always thought the semantic layer is the perfect place for a catalog. Semantic layers are in production so their metadata remains fresh automatically as engineering work drives change. [Semantic layers](https://cube.dev/use-cases/semantic-layer) are at the point of consumption, so are the most relevant abstractions for data users to know about. Other data catalogs seem more geared towards data engineers who are wading through the swamp instead of business users who are looking to see what data they can get hold of to answer their questions.

Released today by Cube, [Semantic Catalog](https://cube.dev/product/semantic-catalog-and-ai-assistant) is centered around the semantic layer. The semantic layer contains the analytics vocabulary that your business users know about; the semantic layer is where data engineering meets business. This inherently makes this catalog more useful for business users. Yes, like other catalogs we have lineage from data source to consumer, but Cube’s Semantic Catalog contains the abstractions understood by business users: entities, metrics and dimensions. They can start at these abstractions, the analytics vocabulary they know about, and go backwards to understand their definition and forwards to understand where they are used in BI tools, data products, and elsewhere. This expands the use of the semantic layer to semantic intelligence.

This is not a data catalog, focused on a data engineer’s need to understand ELT, orchestration and data provenance. Semantic Catalog is focused on data analyst and business user needs, met by their data and analytics engineers.

![](https://ucarecdn.com/665e357f-dcde-4842-bbcd-df9aac427347/-/quality/smart/-/format/auto/)

## Allowing users to discover both data and metadata using AI

On top of the Semantic Catalog, we are also launching our new AI Assistant. This builds upon our launch of the AI API last month and further enhances with the metadata stored in Semantic Catalog. 

Business users can come here and ask questions in natural language. They don’t need to know any special syntax or terms. They don’t need to learn how to use a complex user interface. They can ask questions about what objects are in Semantic Catalog but also ask data questions and receive answers and results as line or bar charts here. They can ask these questions in the analytics vocabulary they already know. They are line-of-business experts; they don’t need to be part-time data analysts too.

This is not a BI tool. It is not our aim to build all the features required to deliver a modern BI experience - this is well served already by our partners. This is a data intelligence experience - a way to discover your data and use it in the same place with no technical knowledge required.

![](https://ucarecdn.com/c7838b9f-f9df-4234-83e5-e972dea4502f/-/quality/smart/-/format/auto/)

## Concluding Thoughts

With Semantic Catalog and AI Assistant, we have enabled users to have semantic intelligence - knowledge of how data flows through their stack into their semantic layer, what data is trusted and available in their Cube semantic layer, what downstream assets it relates to and also the ability to query it with natural language and AI.

Semantic Catalog and AI Assistant make Cube Cloud more useful to business users than ever before as well as our core engineering users.


Semantic Intelligence for All

The need for accurate, comprehensive, and easily accessible data is paramount. Ensuring that data is not only available but also meaningful and trustworthy is a challenge that many organizations face. 

Today, we are excited to announce the preview release of the [Semantic Catalog](https://cube.dev/product/semantic-catalog), a new capability designed to simplify data discovery and exploration. Let's dive into what Semantic Catalog offers and how it allows anyone to search and reuse connected data assets.

## What is Semantic Catalog?
Semantic Catalog is a groundbreaking innovation that provides a comprehensive, unified view of connected data assets. It is self-documenting and continuously up-to-date by default, based on the data models you define. This means that as you adjust and refine your data models within [Cube Cloud](https://cube.dev/product/why-cube-cloud), the Semantic Catalog automatically updates to reflect these changes, ensuring that your metadata is always up-to-date. This will be very different to the experience that Data Engineering teams have had previously with data catalogs, which require a great deal of manual entry and maintenance and invariably become stale. This is why we believe the semantic layer is the perfect place to have a catalog. Metadata from the semantic layer is in production and is kept current as a byproduct of engineering work.

## Key Capabilities of Semantic Catalog
### 1. Unified Search of Connected Data Assets
Previously, data engineering teams would look through the asset lists of multiple tools in their data stack to find specific objects. With Semantic Catalog, users gain the ability to search through a unified view of connected data assets seamlessly. 

Whether you are looking for modeled data in Cube Cloud, downstream BI content, or upstream tables, you can now find it all within a single, cohesive interface. This reduces the time spent jumping between different data sources and platforms, offering a more streamlined and efficient data discovery process for both engineers and consumers.

![](https://ucarecdn.com/b3a05158-db44-4593-9368-b1bbc3fc14fb/-/quality/smart/-/format/auto/)

### 2. Insights into Data Lineage and Relationships
Understanding the relationships and lineage of your data has often been a complex and time-consuming task. Semantic Catalog simplifies this by providing clear insights into the lineage, entities, relationships, dimensions, and metrics of your data. This is crucial for data governance, as it helps you ensure the quality and integrity of your data throughout its lifecycle. 

Additionally, this understanding empowers data analysts to make more informed decisions based on a comprehensive view of the data's history and interconnections.

__product screenshot: data graph from source tables to cubes and views__

### 3. Exploration of Downstream Content
Before Semantic Catalog, it was hard to know what the impact of engineering changes would be on downstream data assets like reports and dashboards. Engineers would have to check the code being run in these downstream tools to see what upstream assets they used.

Semantic Catalog facilitates the exploration of downstream charts and dashboards across all BI platforms. This means you can easily visualize how data flows from source tables through various transformations and into the reports and dashboards that stakeholders depend on. By visually navigating these connections using a [Data Graph](https://cube.dev/docs/product/workspace/data-model#data-graph), you can quickly identify the impact of engineering work to upstream data on downstream analytics, helping you to proactively manage and mitigate potential issues. For data analysts, starting with a Semantic Catalog search will help them find trusted data faster and also understand whether BI content already exists, rather than starting from scratch.

![](https://ucarecdn.com/5d4489af-9416-4c3f-9c17-fdb44663b177/-/quality/smart/-/format/auto/)

## Reducing Duplicate Efforts and Enhancing Collaboration
One of the standout benefits of the Semantic Catalog is its ability to reduce duplicate efforts in data modeling and analysis. In many organizations, data engineers and analysts often find themselves working in silos, leading to redundant work and inconsistencies in data interpretations. 

Semantic Catalog fosters better collaboration by providing a single source of truth that both data engineers and analysts can rely on. This unified approach not only increases productivity but also ensures that everyone in your organization is working with the same, trusted data.
Increasing the discoverability of data assets also reduces the likelihood of duplicate versions being created organically by different engineering teams or by analysts and engineers separately.

## Conclusion
The introduction of the Semantic Catalog marks a significant milestone in our ongoing commitment to empowering data-driven organizations. By providing a broad view of connected data assets, centered on the semantic layer, this product addresses some of the most pressing challenges in data discovery and management. From reducing duplicate efforts and enhancing collaboration to offering clear insights into data lineage and relationships, Semantic Catalog will greatly enhance the way your organization interacts with data.

We invite you to preview the Semantic Catalog and experience how it can improve data discovery. Stay tuned for more updates and enhancements as we continue to innovate and expand the preview capabilities. [Contact sales](https://cube.dev/contact?utm_source=blog&utm_content=Semantic%20Catalog%20blog) to learn more about Semantic Catalog and how it can connect your organization with trusted data.

Introducing Semantic Catalog (Preview)

At Cube, we’re on a mission to power the next-generation of data experiences by bringing consistency, context, and trust to all of your data. Today, we’re excited to unveil new advancements in Cube Cloud that accelerate the development of Generative AI (GenAI) and embedded analytics solutions. 

The latest updates include our new, native AI API, now in preview, and the general availability of Chart Prototyping. These innovations are built on the flexible foundation of Cube’s universal semantic layer, increasing the value you receive from your Cube Cloud deployment.

## Get Conversational with the AI API
The universal semantic layer is essential for building GenAI data applications. It provides context and governance to LLMs, reducing errors and hallucinations. Before the AI API, connecting GenAI to your data required either tedious manual work, such as vectorization and custom prompting, to provide the LLM with the right context and metric definitions, or going straight to a database without the governance and consistency of a semantic layer.

Customers like [spyne](https://cube.dev/case-studies/embedded-analytics-and-ai-chatbot-on-one-semantic-layer) and [Quantatec](https://cube.dev/case-studies/customizable-embedded-dashboards-and-natural-language-ai-queries) already benefit from the context provided by a universal semantic layer to LLMs. We were inspired by these customer use cases to simplify and accelerate development with the AI API. Now, Cube provides a standard interface for interacting with Anthropic Claude models as a turnkey solution for text-to-SQL queries from LLM-generated semantic layer requests or BYO-LLM. 

![](https://ucarecdn.com/e130a926-80c2-4679-9bf6-e3f05186e4f3/-/preview/1000x406/)

The new, native AI API reduces the complexity of typical application-LLM integrations. With a well-documented data model, shown below, and the AI API you can more easily create AI agents, chatbots, copilots, and Slack apps, using natural language interactions. Even better, your users will love it because it lowers the barrier to entry for working with data. The AI API will be generally available this fall.

![](https://ucarecdn.com/6dc9578a-9e83-4d6d-93f1-997a68cd5ed1/-/preview/932x295/)

## Streamline Development with Chart Prototyping
In addition to the AI API, we’re introducing Chart Prototyping for embedded analytics. Embedded developers and data engineers can now create and test charts within Cube Cloud’s Playground 2.0. This speeds up the development process, so that you can ensure insights are delivered in the most impactful way. 

Whether you’re integrating charts into internal tools or external customer applications, Chart Prototyping enables you to generate front-end TypeScript code, preview different chart types, and select your integration method via native REST or GraphQL APIs.

![](https://ucarecdn.com/1ca405f4-171b-4e94-a45f-2a7e9907999c/-/preview/1000x592/)

## Build Next-Generation Data Experiences
Cube Cloud’s new features are part of our commitment to helping you unlock the full potential of your data. With these capabilities, you can build secure, accurate, and cost-effective data experiences, enhancing the value of data for every user both inside and outside of your organization. 

[Contact sales to learn more](https://cube.dev/contact?utm_source=blog&utm_medium=site&UTM_Publisher=Cube&leads[…]tion&campaignid=701UH000009UsImYAK&utm_content=Contact%20Request) about how you can activate and leverage the AI API today. 

Introducing the AI API and Chart Prototyping in Cube Cloud 

Cube’s AI API is here, and it is available on all Cube Cloud paid tiers, including Starter. In our newest native API, we built the ideal architecture to extend the value of your Cube Cloud investment and accelerate the development of natural language capabilities with a turnkey solution integrated with Anthropic Claude models or [BYO-LLM](https://cube.dev/docs/product/apis-integrations/ai-api#other-llm-providers). 

The AI API includes Retrieval Augmented Generation (RAG) and prompt engineering, for an LLM to interface with the Cube’s universal semantic layer. This simplifies the development of AI applications and features in your data products.  

We leverage our deep knowledge of Cube’s universal semantic layer, its APIs, and metadata. You benefit from this API without needing to build your own RAG, prompt engineering, or vector store of your own. This saves a huge amount of time, effort, and cost compared to developing in-house. Now, you can focus on developing the best AI agents and chatbot experiences for your customers.

## Why Is Text-to-Semantic Layer Better?

The semantic layer is the foundation for trustworthy AI access to data using LLMs. With the public preview release of our new AI API you can now use Cube’s universal semantic layer with AI. The new AI API interfaces with our existing REST API to request the data the user has asked for; it does not generate raw SQL.

The semantic layer provides the ideal middleware to stand between LLMs and data. When asking an LLM to generate raw SQL on a database schema, it’s far too easy for it to hallucinate or get small things wrong as we saw in our webinar, [Delivering on the Promise of AI](https://cube.dev/events/delivering-on-the-promise-of-ai-increasing-accuracy-with-a-semantic-layer). 

The semantic layer provides context and also constrains what the LLM can answer with. LLMs have been trained on language itself–nouns, adjectives,  etc. Semantic layers provide LLMs with objects that are equivalent to these language elements: entities for nouns, dimensions/attributes for adjectives,  and measures for quantitative description. The LLM has the choice of the objects in the semantic layer and their attributes defined there, to answer questions with - not the whole knowledge it has been trained on.

This is just the first part, and is true of providing knowledge graph access to an LLM too. Semantic layers are special knowledge graphs, designed for querying data. Semantic layers, like Cube, provide APIs which have much more constrained input than raw SQL. If an LLM is forced to generate such an API request, it is much less likely to make a technically valid hallucination. Add to this, the fact that Cube is still compiling the request into raw SQL, doing the joins, aggregation, and filtering deterministically with our best-in-class compiler. The LLM is again constrained from work that it couldn’t reliably get correct.

![](https://ucarecdn.com/35b2c82f-5ba9-43c9-a13e-f470d8bc076d/-/quality/smart/-/format/auto/)

## Preparing Cube Data Models for Generative AI
You will use existing views within Cube Cloud to keep it DRY (Don’t Repeat Yourself); however, there are likely some modifications needed to work with the new AI API, as compared to other use cases.

Imagine that a data engineer from another company is reviewing your data model. The baseline test for preparing your data model for Generative AI is: **Is it clear enough that a data engineer from another company would understand it?** If the answer is yes because you’ve used clear, plain-language, readable terms and don’t have a lot of acronyms or company lingo, then it is likely that it can work with this API.

Don’t worry if this isn’t the case though, a few simple modifications can make all the difference. You don’t need to rework your whole semantic layer, just the parts that you want to expose to the AI API. First, ensure that you are working with Views in Cube Cloud, then use the following best practices to prepare it for use with the AI API:

- **Use clear, standard names.** Make names understandable, i.e. call a customer a Customer not some strange acronym like CustAcctNm that only people in your company use.
- **Add verbose descriptions.** Explain the member's purpose clearly. If you don’t want to put too much in the description field, in case it confuses human users, you can use a meta-tag to add more detail for the Cube AI API to pick up on. It’s helpful to mention in either if people use synonyms or acronyms to refer to that entity or member, if specific values for the field have special meanings etc. 
- **Make member names distinct.** If you have multiples of the same type of member, such as timestamp fields, try to make it very clear as to the use case for the field. For example, order_timestamp for when the order was made, and shipped_timestamp for when an order was shipped. Again, where there is ambiguity, you can describe how to treat it in the descriptions or meta-tag as if you were talking to a colleague.

![](https://ucarecdn.com/6dc9578a-9e83-4d6d-93f1-997a68cd5ed1/-/preview/932x295/)
## Testing the AI API with Postman
Once you have your views built and the AI API enabled (see the topic for the AI API in docs for more info on that), you’ll naturally want to try it out. An easy way to test is to use an API testing tool like Postman that lets you send API requests and inspect responses.

To get started, click the “Import” button in the top left or press Command+O. Then, paste the following example request in and hit enter:

```
curl \
 -X POST  \
 -H "Content-Type: application/json" \
 -H "Authorization: Bearer EXAMPLE-API-TOKEN" \
 --data '{ "messages": [{ "role": "user", "content": "what cities do we have the most sales in?" }]}' \
 https://YOUR_CUBE_API/cubejs-api/v1/ai/query/completions
 ```

This brings up a request. Replace YOUR_CUBE_API in the URL with your own Cube Cloud API, and EXAMPLE-API-TOKEN in the authorization header with a valid JWT for your Cube deployment.

In the request body, replace the question in “content” with something relevant to your own data set and the view(s) that you created.

After you hit Run, you will get a response that looks like the following: 
![](https://ucarecdn.com/50f63e71-5c35-41c9-ad1b-0bb93952cae3/-/preview/1000x561/)

This is a Cube query on your data, generated by AI! To run it, just copy the Cube query into a request to Cube’s REST API’s /load endpoint. Alternatively, you can add` "runQuery": true` to the AI API request body to run it inline.

<video width="100%" controls><source src="https://ucarecdn.com/16eee2fd-02ea-4311-90e7-ae3f4142c7c3/" type="video/mp4"/></video>

For example, in the video above, we ask the AI API for the cities where we have the most sales. It gives us back a Cube query using the Line Items view: total amount by city, top 10 descending. When we run the query, we can see that Los Angeles is where we have the most sales, followed by Seattle and Mountain View.

And now you’ve run your first Cube query using AI! Feel free to explore, add more messages to the conversation, and ask other questions to get a feel for what it can do.

## Get Started with the AI API
The release of Cube Cloud’s new AI API simplifies the development of Generative AI experiences for embedded developers and data engineers, while natural language interactions lower the barrier for every user of your product or solution–both inside or outside of your organization. 

We will continue to invest in the AI API and improve it over time, allowing all teams who deploy using our AI API to benefit from these enhancements as we release them. When it comes to how to use the Cube Cloud’s universal semantic layer with AI, we’ll do the heavy lifting, so you can focus on building sophisticated data experiences. To get started with the AI API in preview, [contact sales to learn more!](https://cube.dev/contact?utm_source=blog&utm_medium=site&UTM_Publisher=Cube&leads[…]tion&campaignid=701UH000009UsImYAK&utm_content=Contact%20Request)

A Practical Guide to Getting Started with Cube's AI API

I’ve been writing about the [semantic layer](https://davidsj.substack.com/p/semantic-layers-a-buyers-guide) for some time now, and regularly at that! I’ve said before that AI + SL is a match made in data, and I really believe that semantic layer adoption is necessary for any organisation that wants to use AI with their data - whether this is to make AI experiences for their customers or to democratise data access internally.

This is the basis on which Michael and I founded [Delphi Labs](https://www.delphihq.com/). We could see many startups being founded to offer this democratisation by using LLMs to write SQL queries, but we believed that this method was very limited. After experimenting with a few we realised that this method really suffers from hallucination as soon as a data model is more than a handful of tables. Delphi was the first company founded to focus on using third party semantic layers, like Cube, as the governed interface for LLMs to use data.

Through experimentation with all of the main semantic layer offerings, Michael and I found Cube. We found the [Cube Semantic Layer](https://cube.dev/use-cases/semantic-layer) to be the very best to use with LLMs and for our AI application in Delphi. We also met some great Cube folks and formed a strong partnership with them, leading to our announcement today.

Michael and I are joining Cube full time. I will be joining as VP of AI and Michael will be working alongside me to help bring AI offerings into Cube. We will be leveraging our knowledge of both AI and data engineering to enhance Cube’s product even further.

We see huge potential in joining Cube and real affinity with our mission. If you ask any senior data stakeholder, CXO/VP/HoX, whether they want to simply ask a question and get an answer back in seconds or minutes rather than hours, days or weeks… they say yes please! Ask the same stakeholders what a semantic layer is and why they should invest in one, and they require a lot of education and persuasion of the benefits. When it’s simply the infrastructure needed for their AI data experience to work, to be safe and to be reliable - then it’s like explaining why a train needs rails.

This move will help a much larger audience access what we’ve built and learnt, more easily. They can get the benefits of the best semantic layer out there and AI in one place. Get in touch to learn more about LLM and AI built on top of Cube’s semantic layer.

Michael and I are both experts in deploying and using semantic layers in practice, having both used them since leading data teams in our practitioner careers. We are excited to help contribute to Cube’s semantic layer progression that isn’t related to AI too! Given what I’ve written about before in my blog, it’s clear to all that I’m a semantic layer fan!

There will be more specifics to come about integrations, what we’re building at Cube, timelines, etc. as we move forward. 


The Delphi Labs team joins Cube.

Investing in Cube’s AI-driven future. 

Every business today is data-centric and the role of semantic layers for optimizing efficiency is undeniable. Serving as the bridge between raw data sources and end-point data sources, semantic layers offer a consistent, comprehensive perspective of data across the board. This helps businesses make decisions rooted and organized in accuracy and reliability. By centralizing the definition of semantics and metrics, data teams can eliminate inconsistencies and ensure a unified understanding across all data users, be in internal, external or AI bots. Let us look into how semantic layers streamline data management and analytics, simplifying the concepts of a semantic layer into digestible and relatable terms. 

## How a Semantic Layer Works in the Data Stack

[Semantic layers](https://cube.dev/use-cases/semantic-layer) are not a new concept, but their importance has grown with the increasing complexity of modern data stacks. Today, businesses use a variety of visualization tools, customer-facing analytics, [embedded analytics](https://cube.dev/use-cases/embedded-analytics), and AI agents. Each time a new data visualization or data app is implemented, metrics and/or logic need to be defined. If not managed properly, this can lead to inconsistencies and discrepancies and, ultimately wasted resources. Semantic layers address this issue by centralizing these definitions, ensuring consistency across different data consumption endpoints.

The semantic layer has been around for a while, embedded in Business Intelligence (BI) tools. However, the recent shift towards standalone semantic layers is driven by the need for a universal, easy-to-use, and highly discoverable data interface. The democratization of data has led to fragmentation and silos, impeding ease of use and leading to failed analytics projects. A standalone semantic layer addresses these issues, ensuring a single source of consistent data that is easier for data teams to work with. 

The implementation of semantic layers can significantly streamline data management processes. Sitting on top of [data warehouses](https://cube.dev/blog/getting-started-with-a-snowflake-semantic-layer), semantic layers provide data semantics (context) to various data applications. They work seamlessly with transformation tools, allowing businesses to define metrics, prepare data models, and expose them to different BI and analytics tools. This not only enhances efficiency but also ensures that all data consumption places are working with the same, accurate data. So, no matter if the data is being used by a person looking at a dashboard, or a LLM that is giving a person answers to questions, the data is consistent. All of this makes it easier for data teams to quickly deliver data for the various data consumers they work with internally and externally.

## Examples of how companies use semantic layers

Businesses like [Drift](https://www.drift.com/) and [Breakthrough](https://www.breakthroughfuel.com/) have successfully leveraged semantic layers to enhance their data management and analytics. Drift, a customer communication platform, uses semantic layers to ensure consistency in its metrics across different systems. This has not only improved their reporting capabilities but also enabled them to add new data in days instead of months. Breakthrough, a transportation management platform, uses semantic layers to provide consistent data across their SaaS and consulting platforms, significantly improving productivity.

Adopting a semantic layer in an organization is a step-by-step process. It begins with identifying inefficiencies and building a thoughtful semantic model. This model is then incrementally applied to different products, aligning metrics and data definitions. The process also involves identifying and fixing bugs in the old system, ensuring that the data is well-understood and well-documented. This is crucial for the success of analytics and AI projects.

### How AI agents, LLMs and the semantic layers work together

The semantic layer also plays a pivotal role in the realm of [generative AI](https://cube.dev/use-cases/llm-and-ai-semantic-layer). As AI agents are increasingly used for various applications, they need to access and analyze hard data. Due to their non-deterministic nature, AI agents are not good with hard numbers, they need to rely on the context training that has been done on the Large Language Model (LLM) they are fed from. Use of LLMs to generate code (like SQL) can be executed to get the hard numbers for the AI agent. This is where the semantic layer comes in, providing AI agents with the necessary data access and semantics via the LLM. 

However, the use of generative AI also presents new challenges. One of these is the need for a new interface for interacting with data. Another is the need for explainability, ensuring that users can trust the data and understand how the AI is working. The semantic layer can help address these challenges, providing a well-documented and well-understood data interface for AI agents. Semantic layers provide both valuable context and necessary constraint to empower retrieval augmented generation models. While much has been written about providing context via embeddings, we are now learning that constraining LLMs via a semantic layer - choosing from a defined universe of metrics - can be very useful for improving accuracy.  A recent test was performed with an AI Chatbot on data – with and without a semantic layer feeding the LLM. The results were dang clear – AI Chatbot with a semantic layer, outcomes were 100% accurate — AI Chatbot without a semantic layer, outcomes were 83% accurate.  

In summing up, semantic layers are an invaluable asset for businesses keen on optimizing their [data management](https://cube.dev/product/caching-and-data-performance) and analytics. They centralize metric definitions and semantics, ensuring consistency across various data points, which in turn, bolster efficiency and precision. The future will bring more data sources and more data applications, creating even more chaos in the data stack. The significance of semantic layers in navigating the intricacies of the contemporary data stack is poised to increase exponentially. The semantic layer plays a crucial role in increasing confidence in data for non-technical users, fostering the incorporation of analytics and AI projects. It also paves the way for generative AI to access and scrutinize data. Looking ahead, the impact of the semantic layer on data analytics and AI is expected to be even more profound.


Use Cases

Optimizing Data Management and Analytics Efficiency with Semantic Layers

Ever since the wide availability of ChatGPT in early 2023, Large Language Models (LLMs) have become a subject of intense interest and investment in nearly every industry. As we all try to wrap our heads around exactly how the power of generative AI will transform the world, here at Cube it was obvious to us that [semantic layers](https://cube.dev/use-cases/semantic-layer) would have a part to play. This would take shape as the integration of the Cube semantic layer into applications that retrieve information from data stores, perform analysis and return requested results to a user who is interacting with them using natural language, interpreted by an LLM. The exact return method of these results could be multimodal - natural language generation, tabular result sets, or even chained into visualization libraries to return dashboards, but the big idea is “ask a question, get an answer”. We weren’t sure exactly what it would look like, but we knew it was coming.

To prepare for this tectonic shift, we focused on learning about how builders were thinking about this problem and we released features that we felt would empower them such as our semantic document loader for LangChain and our expanded meta endpoints to make it easier to provide the useful context within Cube to the LLM, improving consistency and reducing hallucinations. We also engaged with many builders focusing on next generation analytics experiences driven by natural language query interfaces enabled by LLMs such as our partners at [Patterson Consulting](https://pattersonconsultingtn.com/) and [Delphi Labs](https://www.delphihq.com/) who have been instrumental in delivering AI-enabled analytics use cases for our customers. 

While we all felt that it made sense that an LLM based application would provide more accurate results when answering questions that required examining data within a database with the aid of well designed semantic layer, when compared with more simple zero-shot text-to-sql methods, we will be the first to tell you that we were long on customer anecdotes and light on empirical evidence. 

Recently a group of researchers from data.world (Juan F. Sequeda, Dean Allemang and Bryon Jacob) have published a [ground-breaking paper](https://arxiv.org/pdf/2311.07509.pdf) studying the accuracy of LLMs answering questions with a text-to-sql pattern and comparing those results with the accuracy of introducing a knowledge graph into the process. For those who may not be familiar, knowledge graphs are somewhat similar to semantic layers in that they are also an additional layer of business context that can inform usage of data. 

In addition to the results, the paper also puts forward the detailed methodology for evaluating the accuracy of LLMs in answering business questions asked of a database. They provide a schema, sample data, natural language questions and expected answers in a format that is fairly easy to replicate their experiment. Our partners at Delphi, who build an AI-enabled analytics application on top of semantic layers, quickly realized that this methodology could be adapted to perform a benchmark of the accuracy of their own solution. **[They have recently published some interesting findings of their own](https://delphihq.substack.com/p/delphi-at-100-video), that they were able to achieve an outstanding 100% accuracy with Delphi connected to a subset of the data.world test modeled in Cube as the semantic layer.** This is an incredible result and not only a validation of the ability of semantic layers to provide necessary context to improve LLM accuracy, but also speaks to their thoughtful prompt engineering and design choices. 

## Why are semantic layers the missing piece?

The two driving themes for why we need to use semantic layers with LLMs are context and constraint. We see more and more teams moving towards providing context through prompt engineering and doing things like providing the LLM with a knowledge graph to interface with. This is definitely a step in the right direction, but you also need to constrain their output.

Semantic layers provide both a knowledge graph and a constrained interface for an LLM. As a special type of knowledge graph, a semantic layer provides the LLM with a model of a world, composed of entities and their measures and dimensions. LLMs have been trained on language, and all language is composed of entities (nouns) and their dimensions and attributes (adjectives). Language also contains mathematical terms such as ‘total’, ‘running total’, ‘average’… as a result, LLMs are well-suited to knowledge graphs. 

The reason LLMs are also quite good at SQL is because of how many documents and articles (think Stack Overflow) contain SQL. However, SQL is a minuscule fraction of what LLMs have been trained on. An interface closer to natural language is a better fit for LLMs, and a constrained one, to reduce chance of error and hallucination, is even better. Requests to Cube’s REST API are both close to natural language (a requested list of metrics, dimensions and filters with few, if any, additional terms) and highly constrained (there is only really one way to form a request), if the LLM generates the request incorrectly this will almost always result in an error rather than a hallucination.

However, this has been theory so far, with some anecdotal evidence from using Delphi. What we needed was a benchmark, and now that the data.world one has been created - we have one.

## Testing Methodology
We undertook an ambitious benchmark test, following the footsteps of data.world and dbt Labs, who [tested their semantic layer on a subset of 11 questions](https://github.com/dbt-labs/semantic-layer-llm-benchmarking/tree/main) from the original benchmark. Our goal was to replicate dbt Labs’s methodology on our platform, utilizing Delphi in conjunction with the Cube semantic layer. **The outcome was remarkable - while dbt Labs achieved an impressive 83% across 8 questions (excluding 3 with multi-hop joins), Delphi + Cube reached a perfect score of 100% on all questions.**

The benchmark began by setting up a Snowflake data warehouse. We created the necessary tables from the data.world benchmark using the [DDL available in our GitHub repository](https://github.com/Delphi-Data/delphi-semantic-layer-llm-benchmark/blob/main/ACME_small.ddl).

Once the database structure was in place, we proceeded to populate it with the data files sourced from the insurance dataset [available on data.world’s GitHub](https://github.com/datadotworld/cwd-benchmark-data/tree/main/ACME_Insurance/data), originally developed by Object Management Group. This model, which involves a 3NF structure with multi-hop joins, was crucial in replicating the complexity of real-world data, setting the stage for an authentic benchmarking experience.

Next, we started building out the Cube semantic layer. This involved signing up for Cube’s free tier and creating a new project using the credentials from our Snowflake account. Using Cube Cloud’s built-in IDE, we developed the Cube schema available in our GitHub repository, including all tables and dimensions from the benchmark (even “extra” ones our questions did not use) and some basic measures.

Finally, we created a new free account on Delphi, connected Cube via its REST API integration, and then were off to the races. Working off the benchmark results sheet created by dbt Labs, we ran each question in a separate Delphi conversation and tracked the results. We replicated this 5 separate times to ensure that our results were consistent and not impacted by the non-determinism that’s often possible with LLMs.

Our results are publicly available, including the Cube queries Delphi generated. Delphi was able to answer 100% of the questions correctly, even including some where it was able to calculate the needed measures on the fly.

In line with our commitment to transparency and community collaboration, we have made our Cube schema and benchmark instructions publicly available. These resources can be accessed on our GitHub repository at [Delphi Semantic Layer LLM Benchmark](https://github.com/Delphi-Data/delphi-semantic-layer-llm-benchmark), inviting others to replicate our benchmark and share their findings.

## Conclusion

There is now empirical evidence that using a semantic layer with an LLM, for analytics, is superior to other methods. We have chained a number of proprietary methods in addition to this to deliver the best-in-class performance we see with Delphi. We invite everyone to take a look at the testing methodology above and take both [Cube](https://cubecloud.dev/auth) and [Delphi](https://auth.delphihq.com/en/signup) for a spin. We’d love to show you how easy it is to start asking questions of your data and feel confident that the answers your team will receive are grounded in the reality of your data.


Semantic layers provide LLMs context and constraint.

Semantic Layers are the missing piece for AI-Enabled Analytics

For many years already, we live in the data-driven world where accessing the data and deriving insights from it efficiently is paramount. This year, we experience an explosion of interest to artificial intelligence (AI) and large language models (LLMs) in particular, fueled by the latest developments in the technology and vast perspectives unfolding.

[LangChain](https://www.langchain.com), an extensive toolkit for working with LLMs, has become one of the most important building blocks for developers of AI experiences. At Cube, we know that our semantic layer is also an important building block for AI applications because Cube not only centralizes metrics calculations but also serves an [antidote to AI hallucinations](https://cube.dev/blog/semantic-layer-the-backbone-of-ai-powered-data-experiences).

## Semantic document loader

Today, we're happy to present Cube's integration with LangChain. It comes as the [document loader](https://python.langchain.com/docs/integrations/document_loaders/cube_semantic) that is intended to be used to populate a vector database with embeddings derived from the data model of the semantic layer. Later, this vector database can be queried to find best-matching entities of the semantic layer. This is useful to match free-form input, e.g., queries in a natural language, with the views and their members in the data model.

![diagram](https://ucarecdn.com/32e98c8b-a920-4620-a8d2-05d57618db8e/)

We're also providing an chat-based demo application (see [source code](https://github.com/cube-js/cube/tree/master/examples/langchain) on GitHub) with example [OpenAI prompts](https://github.com/cube-js/cube/blob/master/examples/langchain/utils.py#L48,L87) for constructing queries to Cube's [SQL API](https://cube.dev/docs/product/apis-integrations/sql-api). If you wish to create an [AI-powered conversational interface](https://cube.dev/blog/conversational-interface-for-semantic-layer) for the semantic layer, functioning similar to [Delphi](https://www.delphihq.com), these prompts can be a good starting point.

## Chat-based demo app, dissected

Here's what you can build with the all-new document loader for LangChain, a vector database, an LLM by [OpenAI](https://openai.com), and [Streamlit](https://streamlit.io) for the user interface:

<video width="100%" controls><source src="https://ucarecdn.com/5500be7a-ab78-4de8-abc9-02df575ad3e0/video.mp4" type="video/mp4"/></video>

See how the tables, columns, aggregations, and filters in the SQL query generated by an LLM match the human input. Check the [README file on GitHub](https://github.com/cube-js/cube/blob/master/examples/langchain/README.md) for pointers on running this demo application on your machine or skim through the following highlights.

**Ingesting metadata from Cube and populating the vector database.** The `ingest_cube_meta` function in the `ingest.py` file loads the data model from Cube using the all-new `CubeSemanticLoader`. Note that only [views](https://cube.dev/docs/product/data-modeling/reference/view) are loaded as they are considered the ["facade" of the data model](https://cube.dev/blog/complementing-data-graph-with-views). Loaded documents are then embedded and saved in the FAISS vector store, which is subsequently pickled for later use.

```python
def ingest_cube_meta():
    ...
    loader = CubeSemanticLoader(api_url, api_token)
    documents = loader.load()
    ...
    with open("vectorstore.pkl", "wb") as f:
        pickle.dump(vectorstore, f)
```

**LLM setup.** In the `main.py` file, dependencies are imported and the environment variables are loaded. The OpenAI model (`llm`) is initialized with the provided API key.

```python
import ...
load_dotenv()
llm = OpenAI(
    temperature=0, 
    openai_api_key=os.environ.get("OPENAI_API_KEY"),
    verbose=True
)
```

**User input and vector store initialization.** In the same file, Streamlit primitives are utilized to get user input:

```python
question = st.text_input(
    "Your question: ",
    placeholder="Ask me anything ...",
    key="input"
)
if st.button("Submit", type="primary"):
    check_input(question)
    vectorstore = init_vectorstore()
```

**Querying the vector store.** The vector store is queried for documents similar to the user's question. The best match's table name is extracted and taken as the best guess to try to create a prompt using the columns from vectorstore:

```python
docs = vectorstore.similarity_search(question)
    # take the first document as the best guess
    table_name = docs[0].metadata["table_name"]

    # Columns
    columns_question = "All available columns"
    column_docs = vectorstore.similarity_search(
        columns_question,
        filter=dict(table_name=table_name),
        k=15
    )
```

**Building the prompt and calling OpenAI.*** The OpenAI large language model is called with the constructed prompt, and the response is parsed to extract the SQL query and any associated filters:

```python
# Construct the prompt
prompt = CUBE_SQL_API_PROMPT.format(
    input_question=question,
    table_info=table_name,
    columns_info=columns,
    top_k=1000,
    no_answer_text=_NO_ANSWER_TEXT,
)
llm_answer = llm(prompt + PROMPT_POSTFIX)
```

![](https://ucarecdn.com/1252fc98-75db-4887-be82-227b60d477a9/)

## Wrapping up

Cube's integration with LangChain provides a seamless interface for querying data using natural language. By abstracting the complexities of SQL and leveraging the power of LLMs, it provides the builders of AI experiences with a user-friendly and error-prone approach to data access.

<Quote authorName="Josh Patterson" authorAvatar="https://ucarecdn.com/a3e94c6f-5d02-4b57-b523-bbfaf04f4890/8s5_8mFF_400x400.jpg" authorRole="CEO at Patterson Consulting">

It is essential for enterprises to leverage internal knowledge correctly with the power of LLM’s reasoning. Cube’s semantic layer integration with LangChain is a great example of how most product companies will eventually write smart integrations for LLMs to better power these reasoning engines.

</Quote>

Our partners at Patterson Consulting have been hard at work building with Cube, LangChain and LLMs. They have written a wonderful ebook as a primer to [Building Enterprise Applications with LLMs](https://pattersonconsultingtn.com/content/intro_to_llms_ebook.html).

**You can try the LangChain integration with a Cube Core or [Cube Cloud deployment](https://cubecloud.dev/auth/signup) today.** Join our Slack community at [slack.cube.dev](https://slack.cube.dev) to share your thoughts and opinions about this update.

Mikhail Nitsenko

Powerful tooling for building AI experiences with LLMs and the semantic layer

Introducing the LangChain integration

## The AI-powered data experience

We’ve long seen that the data experiences landscape has been very diverse. From dozens of general-purpose BIs to hundreds of vertical purpose-built analytical solutions, data consumers have many, many options to visualize and consume data.

The evolution of front-end technologies made it possible to embed quality analytics experiences directly into many software products, further accelerating the proliferation of data products and experiences. 

And now, with the arrival of large language models, we are living through another step change in technology that will enable many new features and even in the advent of an entirely new class of products across multiple use cases and domains—_including_ data. 

LLMs are taking the data consumption layer to the next level with **AI-powered data experiences** ranging from chatbots answering questions about your business data to AI agents making actions based on the signals and anomalies in data. 

![](https://ucarecdn.com/5c8bd295-4317-4820-beea-532f8c9ba391/)

## Semantic layer gives context to LLMs

LLMs are indeed a step change, but inevitably, as with every technology, it comes with its own limitations. **[LLMs hallucinate](https://en.wikipedia.org/wiki/Hallucination_(artificial_intelligence)); the [garbage in, garbage out](https://en.wikipedia.org/wiki/Garbage_in,_garbage_out) problem has never been more of a problem.** Let’s think about it like this: when it’s hard for humans to comprehend inconsistent and disorganized data, LLM will simply compound on that confusion to produce wrong answers.  

We can’t feed LLM with database schema and expect it to generate the correct SQL. To operate correctly and execute trustworthy actions, __it needs to have enough context and semantics__ about the data it consumes; it must understand the metrics, dimensions, entities, and relational aspects of the data by which it's powered. Basically—LLM needs a semantic layer.

The semantic layer organizes data into meaningful business definitions and then allows for querying these definitions—rather than querying the database directly. 

The ‘querying’ application is equally important as that of ‘definitions’ because it enforces LLM to query data through the semantic layer, ensuring the correctness of the queries and returned data. With that, the semantic layer becomes a solution for the LLM hallucination problem.

![](https://ucarecdn.com/c79d8e5a-ad67-43b3-ac4f-b3d323c0fb65/)

Moreover, the combination of LLMs and semantic layers can enable a new generation of AI-powered data experiences. At Cube, we’ve already witnessed many organizations build custom in-house LLM-powered applications, and startups, like [Delphi](https://cube.dev/blog/conversational-interface-for-semantic-layer), build out-of-the-box solutions on top of Cube’s semantic layer ([demo here](https://cube.dev/events/unleash-the-power-of-ai)). 

On the edge of this developmental forefront, we see Cube being an integral part of the modern AI tech stack as it sits on top of data warehouses, providing context to AI agents and acting as an interface to query data.

![](https://ucarecdn.com/ea4d11af-b3f9-43b9-b02e-b8fa8056c4b1/)

Cube’s data model provides structure and definitions that are used as a context for LLM to understand data and generate correct queries. LLM doesn’t need to navigate complex joins and metrics calculations because Cube abstracts those and provides a simple interface that operates on the business-level terminology, instead of SQL table and column names. This simplification helps LLM to be less error-prone and avoid hallucinations. 

For example, an AI-based application would first read Cube’s meta API endpoint, downloading all the definitions of the semantic layer and storing them as embeddings in a vector database. Later, when a user sends a query, these embeddings would be used in the prompt to LLM to provide additional context. LLM would then respond with a generated query to Cube, and the application would execute it. This process can be chained and repeated multiple times to answer complicated questions or create summary reports. 

### Performance

In regards to response times—when working on complicated queries and tasks, the AI system may need to query the semantic layer multiple times, applying different filters. 

So, to ensure reasonable performance, these queries must be cached and not always pushed down to the underlying data warehouses. Cube provides a relational cache engine to build pre-aggregations on top of raw data and implements aggregate awareness to route queries to these aggregates when possible.

### Security

And, finally, security and access control should never be an afterthought when building AI-based applications. As we mentioned above, generating raw SQL and executing it in a data warehouse may lead to wrong results. 

However, AI poses an additional risk here: since it cannot be controlled and may generate arbitrary SQL, direct access between AI and raw data stores can also be a significant security vulnerability. Instead, generating SQL through the semantic layer can ensure granular access control policies are in place. 

### And more...

We have a lot of exciting integrations with the AI ecosystem in store and can’t wait to share them with you. Meanwhile, if you are working on an AI-powered application with Cube – [let’s talk](https://cube.dev/contact)! 


—and the antidote to AI Hallucinations.

Semantic Layer: The Backbone of AI-powered Data Experiences

Today, we’re happy to present the newest addition to the plethora of [data tools](https://cube.dev/integrations) that work with Cube, the semantic layer for building data apps. [Delphi](https://www.delphihq.com) provides a way to interact with Cube in natural language and delivers insights straight to business users in Slack. 

![Diagram: Cube and Delphi](https://ucarecdn.com/ed0d51c4-2ff7-40cc-891f-32badcc57a51/)

Powered by the newest large language model (LLM) technology from OpenAI, Delphi brings the accessibility of self-service analytics to the next level and enables a compelling use case for semantic layers. See Cube and Delphi in action:

<iframe width="790" height="447" src="https://www.youtube.com/embed/FotEaaf20gY" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" allowfullscreen allow="fullscreen;"></iframe>

Learn more about Delphi and [semantic superiority](https://davidsj.substack.com/p/semantic-superiority-part-4) from David Jayatillake, co-founder and CEO at Delphi Labs:

## Natural language interface for data

Our mission for Delphi is to solve the whole workflow, from question to answer, with data. Today, this is broken—self-serve BI tools haven’t delivered on this promise to date, and data teams are bombarded with ad hoc requests at a scale greater than ever before as businesses move to operate with data.

When we ask data leaders in the space what % of their organization is capable of self-serving with data, the answer is usually between 10 to 25%, with only 10% being very competent. It’s been three years since “dashboards are dead,” but unfortunately they’re still being used as the primary way to serve data. In my experience, no stakeholder ever wanted a dashboard. They wanted an answer, an explanation, a prediction, data to use elsewhere operationally, a graph or a slide to make a specific point.
This is the problem space we aim to solve with Delphi.

There are many other entrants who are trying to offer a natural language interface to data in this LLM gold rush. However, almost all of them are pointing the LLM at a database schema and getting it to generate SQL. We believe this is the wrong approach: every time a query is answered using this method, the LLM generates a minimal semantic layer on the fly in order to derive meaning from the data. As humans created the data structures and their meaning in the first place, why introduce a probabilistic element into the meaning of your data?

We believe that working with semantic layers curated by data folks is the best approach. It synthesizes the information already baked into the semantic layer, truly enabling safe and accurate answers. It also allows us to say that we can’t answer the question—a very normal outcome when asking a data team a question; LLMs connected directly to a data store will simply guess, as they don’t have a concept of a world and what does, and doesn’t, exist in it, which a semantic layer offers.

This is why we’re so excited to work with semantic layers like Cube—we’re already blown away by the interest we’ve had from the Cube community and hope to continue to work with you all!

As I’ve written in my [substack](https://davidsj.substack.com/archive), I believe the future of semantic layers is standalone. It’s what the people want.

## What’s next?

Here's what an early adopter of Delphi and Cube thinks:

> Getting started with Delphi on Cube was a breeze and we really quickly had a conversational AI interface to our data warehouse. The accuracy was great and some of the Delphi features really helped explain our data and ensure we knew how we were querying it. I'd highly recommend it for it's wow factor and enabling moving towards a data driven culture.

Interested? Implement your [semantic layer](https://cube.dev/use-cases/semantic-layer) with Cube by [getting started with Cube Cloud](https://cubecloud.dev/auth/signup) today or by [contacting us](https://cube.dev/contact). Also, [sign up for Delphi](https://docs.google.com/forms/d/e/1FAIpQLSc7nJc1Zeu5kEHcKrghbTPy3MSxJLOBsge0kj4Iiz5aqX6zJg/viewform) and see it in action at [slack.cube.dev](https://slack.cube.dev) in the `#random-delphi-demo` channel:

![Delphi in Slack](https://ucarecdn.com/ddf30e1e-c179-4fda-8aa8-2f75a05f808e/)
