AI for Power BI semantic models

Ask Power BI like a colleague, get instant metric answers.

forbi.ai connects large language models directly to your Power BI semantic models. Ask questions in natural language, explore metric definitions, and debug your dashboards without writing a single DAX query.

Try the live demo

Currently in private beta. Built on Microsoft Fabric and Power BI.

Connected to demo workspace Preview of the chat experience
You
How many engaged sessions did we have last quarter?
AI

I’m querying the EngagedSessions measure in the GA4 Traffic model for last quarter.

Result From Power BI
128,420
Engaged Sessions Last fiscal quarter

The full interactive version lets you inspect DAX, filters, and breakdowns per query.

Why start with semantics, not screenshots

Most “chat with your BI” tools work on exported tables or screenshots. forbi.ai connects directly to your existing Power BI semantic models: measures, relationships, and security.

For data leaders

  • Fewer ad-hoc metric definition requests
  • More consistent understanding of KPIs

For analysts

  • Explain DAX and filters in plain language
  • Debug misaligned dashboards faster

Bring your Power BI semantics where work happens

forbi.ai turns your Power BI semantic models into a conversational layer for business users, while giving data teams the governance and observability they need.

Answers, not screenshots

  • Ask questions in natural language and get metric-level answers.
  • Always-on context from your semantic model, not ad-hoc exports.
  • Guide users back to the right dashboards when needed.

Documentation and observability

  • Explain DAX measures and filters in plain language.
  • Understand which KPIs are actually being asked about.
  • Use insights to prioritise new metrics and dashboards.

Built for governed data

  • Reuse existing Power BI security and workspace boundaries.
  • Keep a single source of truth for definitions and lineage.
  • Stay compliant by avoiding data duplication in random tools.

How forbi.ai plugs into your Power BI

A simple flow that starts from your semantic models and ends in faster, more confident decisions.

1

Connect your Fabric / Power BI workspace

Authorise a secure connection to your Fabric or Power BI tenant so we can read semantic model metadata (not raw fact data).

2

Select models, measures and dimensions

Choose which semantic models should be exposed, which measures are queryable, and which dimensions are allowed for filtering and breakdowns.

3

Embed in the tools your users already live in

Roll out a chat interface (e.g. Teams, Slack or an internal portal) that talks directly to your governed semantic models.

4

Iterate with real usage signals

See which questions are being asked, where definitions are unclear, and which metrics deserve better dashboards or documentation.

Frequently asked questions

A quick overview of how forbi.ai works with your existing Power BI setup.

How does forbi.ai connect to Power BI semantic models?

We connect via Microsoft Fabric / Power BI APIs to read model metadata: tables, measures, relationships and security roles. Query execution still happens in your Power BI environment – forbi.ai orchestrates questions and explanations on top.

What data does the AI actually see?

The primary source of truth is your semantic model. We focus on metadata, metric definitions and query results – respecting the same row-level security and permissions you already have in Power BI.

Can I start with just one or two models?

Yes. Many teams begin with a single “source of truth” model such as revenue or product analytics, then expand once they see how questions and usage patterns evolve.

How do I try it with my own data?

Start with the public demo to see the experience, then reach out using the “Talk about your data” button or email below. We'll walk through a short discovery and help you connect your first workspace.

Explore forbi.ai for your own semantic models

We're working closely with a small number of teams. Share a bit about your Power BI setup and we'll follow up with next steps.

Email details instead