If your organization is exploring agentic analytics, start with the foundation. Contact The Information Lab at info@theinformationlab.com to build the governed data, semantics, and analytics framework needed for trusted agentic solutions.

In analytics circles, there is no avoiding the increasing prevalence of AI agents in conversations. The premise is straightforward: a business user asks a question in natural language, receives an answer, and moves on without waiting for access to a dashboard or submitting an ad hoc request to the analytics team.

Platforms across the modern analytics stack have embraced this shift from different angles. Specifically, dbt, Databricks, and Tableau Next, while having different approaches, each invest in the same underlying requirements such as preserving context, semantics, and governance to ensure AI-driven analytics is reliable. Those foundations are becoming increasingly important as AI-driven analytics moves closer to everyday business use.

dbt approaches this from the modeling and business logic angle. It extends governed metrics and business context into AI workflows through its Semantic Layer and access to MCP capabilities. 

Databricks approaches it from the data and analytics platform angle. It combines Unity Catalog semantics with AI/BI and Genie. It connects the governed business context directly to conversational and AI-assisted experiences.

Tableau Next approaches it from the business user experience angle. It brings agentic analytics into the business user experience through a platform built on unified data layers and trusted semantics. 

What ties these interfaces together is not the desire for a conversational interface alone but the foundation it is built on. As the deployment of AI agents increases, the more need there is to increase the bar for governed analytics.

The agent is not the starting point

Before rolling out a conversational analytics experience, organisations need to make the business logic behind every answer clear and justifiable.

If a user asks, “How did revenue perform last quarter?” the answer relies on definitions that should not be decided when the query is made:

• What counts as revenue?

• Are refunds included?

• Which data drives the calculation?

• Is the revenue recognized, booked, billed, or collected?

• Who owns the definition?

An AI agent will surface the same discrepancies already buried in the dashboards, SQL queries, and spreadsheets. The experience may feel more modern and intuitive, however, the underlying problem remains the same.

The answer is only as trustworthy as the logic behind it.

Where dbt fits

dbt has carved out an important role in this puzzle. It specializes in turning raw data into documented, reusable, tested models. This is important because trusted analytics starts before the dashboard. It starts with clear data models, defined relationships, tested assumptions, and business logic that teams can review and maintain over time.

Without that structure, analytics logic spreads. A calculation starts in SQL, an analyst uses it in Tableau, adjusts it for one dashboard, and another team builds a slightly different version elsewhere. Over time, the organization ends up with multiple answers to the same business questions.

End users see this as a data problem but it really is an operating model problem.

Where Databricks fits

Databricks shows the same pattern from a different angle. Its expansion into AI/BI and conversational analytics through Genie depends on governance and semantic structure underneath it. Unity Catalog provides that governed layer for data and business context.

Semantics in Unity Catalog lets organizations define reusable metrics and business concepts once, instead of leaving every dashboard, query, or agent to interpret them independently. Metrics views, for example, let teams define measures such as revenue and reuse them across dimensions and analytical scenarios.

That context feeds experiences such as Genie One, where business users ask questions of organizational data in natural language. The more flexible the interface, the more the system depends on authoritative definitions, trusted assets, permissions, and the business context behind every answer.

AI ready analytics means making the underlying environment understandable, governed, and consistent enough for an agent to be used reliably. An agent layered on top of ungoverned data doesn’t fix that.

The interface is conversational, but the foundation still has to be intentional and governed.

Where Tableau Next fits

Tableau Next introduces a different approach to experiencing analytics. The dashboard is no longer the only destination. It seeks to meet the expectations users have for analytics to be conversational, contextual, and available within applications and workflows where decisions are being made.

Tableau describes Next as an agentic analytics platform where analytics, AI, and action are increasingly connected. This again brings us to business context, a foundational requirement  for this experience to work reliably. The AI needs more than just data, it needs the context contained in definitions, permissions, relationships, and metadata.

This need for business context is why the semantic layer is essential. In Tableau Next, Tableau Semantics allows ‌metrics, dimensions, relationships, terminology, permissions, and other analytical context to be reusable, governed, and understandable across the full analytics experience rather than locked in individual dashboards.

Tableau Next represents a move to a more natural interface, so foundational elements become more important. An agent may change how users interact with analytics , but it does not remove the need for trusted definitions and clear ownership, quite the opposite, it increases it.

The real decision

The role each layer plays in creating trusted, decision ready analytics matters more than which tools you pick.

Regardless of tools, organizations still need to answer the operating model questions:

• Who owns each metric?

• Where is the certified definition maintained?

• How are the changes reviewed?

• How is trust communicated to the user?

These are foundational questions, not afterthoughts.

A simple readiness test

Before introducing an AI analytics experience, analytics leaders should ask a simple question:

If a user asks the same business question through a dashboard, a self-service workflow, and an AI agent, should they get the same answer?

Leaders who answer yes need a governed semantic strategy. Leaders who answer no need to be explicit about why different answers are acceptable.

Having variation is not the issue, context changes the answer depending on who is asking. The real issue is unjustified, ungoverned variation.

Start with trust

AI ready analytics means answers are consistent, explainable, and trustworthy enough to support real decisions, whether an agent is used as the interface between users and data.

Building that starts before the agent: with the data model, metric ownership, and governance.

The future of analytics is trending to conversational interfaces, but without a trusted foundation underneath, AI only makes inconsistent logic easier to access.

If your organization is exploring agentic analytics, start with the foundation. Contact The Information Lab at info@theinformationlab.com to build the governed data, semantics, and analytics framework needed for trusted agentic solutions.