Ask your data a question. Trust every answer that comes back.

Lightdash AI agents answer questions in plain English using your governed semantic layer.

Lightdash AI agents answer questions in plain English using your governed semantic layer.

The data teams that stopped being a ticket queue.

Ask questions where you already work.

The question usually lands in Slack or on a dashboard. The answer should land there too.

In the dashboard

Hit "Ask AI" on any chart or dashboard and start a conversation with that view already pinned as context.

In Slack and Teams

Embedded in your product

Get governed answers each time.

Runs on the semantic layer

Every answer is built from the metrics you defined once in code, so "revenue" means the same thing here as everywhere else.

A semantic query, not raw SQL

The agent returns a governed query you can open in the normal explore view and check, so you can see exactly how it got there.

Permissions follow the person

Turn on OAuth and each Slack question respects that user’s own data access. People only ever see what they’re allowed to.

Says when it can’t

When the data isn’t there, the agent tells you, instead of hallucinating and inventing wrong answers.

Get governed answers each time.

Runs on the semantic layer

Every answer is built from the metrics you defined once in code, so "revenue" means the same thing here as everywhere else.

A semantic query, not raw SQL

The agent returns a governed query you can open in the normal explore view and check, so you can see exactly how it got there.

Permissions follow the person

Turn on OAuth and each Slack question respects that user’s own data access. People only ever see what they’re allowed to.

Says when it can’t

When the data isn’t there, the agent tells you, instead of hallucinating and inventing wrong answers.

AI that gets better the more you use it.

Memory

Correct the agent once and it remembers, so the same mistake doesn't come back next week.

Verified answers

Mark a great answer as verified; the agent reuses it for similar questions and surfaces it as a suggested starting point.

Reviews

Lightdash flags likely-wrong answers and suggests one-click fixes to your context layer.

Knowledge documents

Upload glossaries, metric definitions, and business context the agent reads before it answers.

AI that gets better the more you use it.

Memory

Correct the agent once and it remembers, so the same mistake doesn't come back next week.

Verified answers

Mark a great answer as verified; the agent reuses it for similar questions and surfaces it as a suggested starting point.

Reviews

Lightdash flags likely-wrong answers and suggests one-click fixes to your context layer.

Knowledge documents

Upload glossaries, metric definitions, and business context the agent reads before it answers.

AI that gets better the more you use it.

Memory

Correct the agent once and it remembers, so the same mistake doesn't come back next week.

Verified answers

Mark a great answer as verified; the agent reuses it for similar questions and surfaces it as a suggested starting point.

Reviews

Lightdash flags likely-wrong answers and suggests one-click fixes to your context layer.

Knowledge documents

Upload glossaries, metric definitions, and business context the agent reads before it answers.

Built for the people who don't write SQL.

Ask in any language

Ask questions in any language: French, Japanese, Portuguese, and more. All answered, with the chart, in the same language.

Show me what I can ask

New users can ask the agent what it knows, and see suggested questions drawn from your verified answers.

Built for the people who don't write SQL.

Ask in any language

Ask questions in any language: French, Japanese, Portuguese, and more. All answered, with the chart, in the same language.

Show me what I can ask

New users can ask the agent what it knows, and see suggested questions drawn from your verified answers.

"We create specialized agents with different instructions and data access for different departments. Each agent has personalised context specific to its team’s domain, allowing it to do nuanced jobs."

Aleksandr Zolotukhin

VP of Data & Analytics @ JustWatch

86%

reduction in BI tool spend

~4000

queries per day

45%

increase in active BI users

"We create specialized agents with different instructions and data access for different departments. Each agent has personalised context specific to its team’s domain, allowing it to do nuanced jobs."

Aleksandr Zolotukhin

VP of Data & Analytics @ JustWatch

86%

reduction in BI tool spend

~4000

queries per day

45%

increase in active BI users

Your questions, answered

Your questions, answered

For any other questions, reach out to our team.

Can each team have its own agent?

Yes. You can create specialized agents for each team (Marketing, Finance, Growth, etc), each scoped to the right data and business context, and the AI Router sends every question to the best-fit one automatically. You control which tables, fields, and metrics each agent can access.

Which LLM does it use?

Lightdash is LLM-agnostic and works with multiple models from Anthropic and OpenAI. You're not locked into one provider, and you can use the model that performs best for your data.

Why not just point ChatGPT or Claude at our warehouse?

A raw LLM on your warehouse guesses at table joins and invents metrics, with no shared definition of "revenue." Lightdash agents answer from your governed semantic layer, return a query you can inspect, and learn from your corrections, so the answer is one you can actually put in a board deck.

Does it hallucinate numbers?

The agent builds answers from the metrics you defined in your semantic layer, not free-form SQL, and returns a query you can open and check. When the data isn't there, it says so rather than inventing a figure. Reviews also flags likely-wrong answers so you can fix the source.

What does it cost?

Lightdash is a flat fee with unlimited users and unlimited AI queries. You don't pay more as more of the company starts asking questions.

Do I need dbt to use Lightdash?

dbt is the most common setup, and Lightdash builds its semantic layer inside your dbt project. You can also define the semantic layer in Lightdash YAML against your warehouse without dbt. Either way, every dashboard runs on governed metrics defined once, in code.

I have an infrequently asked question.

Great! We're always around to help. You can chat to us through the live chat on this webpage, or talk to us in our Slack community.

See it on your own data

Book a walkthrough, or jump straight into the live demo.

The question isn’t whether your BI can be cleaned up. It’s why you’re the one doing it.

If your semantic layer already lives in code, you’ve done the hard part. Let Lightdash keep it honest, so your team spends its time on the analysis only they can do.

Join our newsletter to be en-lightened

© 2026 Telescope Technology Limited. All rights reserved.

Join our newsletter to be en-lightened

© 2026 Telescope Technology Limited. All rights reserved.

Join our newsletter to be en-lightened

© 2026 Telescope Technology Limited. All rights reserved.