> For the complete documentation index, see [llms.txt](https://help.daasity.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.daasity.com/core-concepts/ai-analyst.md).

# AI Analyst

Ask questions about your Daasity data in plain English - in the app, or from the AI tool you already use.

The AI Analyst lets anyone on your team ask questions about your business in plain English and get answers back from your governed Daasity metrics - the same definitions behind your dashboards and Explores. No SQL, no exports, no waiting in a report queue.

This page explains what the AI Analyst is, why brands use it, and how to choose between the two ways of using it. For setup, see [**Ask in Daasity (AI-Analyst in app)**](https://help.daasity.com/core-concepts/ai-analyst/ask-in-daasity-ai-analyst-in-app) and [**Connect Your AI (MCP)**](https://help.daasity.com/core-concepts/ai-analyst/connect-your-ai-mcp).

### What is the AI Analyst?

Your dashboards answer the questions you anticipated. Most days bring one you didn't - a number, a ranking, a trend. Answering it usually means digging into an Explore (knowing which one, which fields, which filters), asking the teammate who lives in your BI tool, or - for the occasional deep dive - querying the warehouse directly. Multiply that across a team, and questions pile up in a queue, or worse, go unasked.

The AI Analyst removes that friction. Ask a question the way you'd ask a colleague - *"top 10 products by revenue last 30 days"* - and it finds the right fields, runs the query against your Daasity data, and answers in plain English, with the underlying rows one click away.

Because every answer comes from your governed semantic layer, the number you get in chat is the same number your dashboards and Explores show. That's the difference between AI Analyst and pointing a general-purpose AI at a spreadsheet export: the answers are grounded in metrics your team has already agreed on.

### Why brands use it

**Everyone can self-serve an answer.** Merchandising, marketing, ops, and leadership can all pull a number, ranking, or trend the moment they need it - no SQL, no ticket, no analyst dependency for the routine questions that make up most data requests.

**One set of numbers.** Answers use the same metric definitions as the rest of Daasity, so the revenue figure in a chat thread matches the one in the Monday dashboard review. Teams stop debating whose spreadsheet is right and start debating what to do.

**Answers cover your whole business.** Because Daasity harmonizes DTC, marketplace, retail, and wholesale data into one model, "revenue by channel" means *all* of your channels - not just the ones a single tool can see.

**It shows its work.** Every answer includes the assumptions it made (like the exact date window used), a short interpretation of the result, and a **View data** option to expand the rows behind it. When an answer maps to a specific Explore in your model, it's marked **Daasity Verified**.

**It meets your team where they work.** Quick answers live in the Daasity app. Teams already working in an AI tool can [bring their Daasity data to them instead](https://help.daasity.com/core-concepts/ai-analyst/connect-your-ai-mcp).

**Built on your own model.** If you use Looker, answers come from your own explores and definitions, on any Looker setup. If you don't use a BI tool at all, that works too - your workspace is provisioned automatically on Daasity's standard semantic views.

{% hint style="warning" %}
Tableau and Power BI support is on the roadmap.
{% endhint %}

{% hint style="info" %}
**Availability:** AI Analyst is a Daasity add-on. If you don't see it in your workspace, contact your Daasity account team.
{% endhint %}

### Two ways to use it

|               | Ask in Daasity                                                                                      | Connect Your AI (MCP)                                                                                           |
| ------------- | --------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------- |
| Where you ask | Chat assistant inside your Daasity workspace                                                        | Your own AI tool - Claude Desktop, Claude on the web, Cursor, Codex CLI, or ChatGPT                             |
| Best for      | Quick answers while you're already in Daasity; saved threads you can return to and build on         | Weaving Daasity data into work you're already doing in an AI tool - drafting, analysis, prototypes, automations |
| Setup         | None - open it and ask                                                                              | Add Daasity as a connector in your AI tool (a few minutes)                                                      |
| Learn more    | [Ask in Daasity (AI-Analyst in-app)](/core-concepts/ai-analyst/ask-in-daasity-ai-analyst-in-app.md) | [Connect Your AI (MCP)](/core-concepts/ai-analyst/connect-your-ai-mcp.md)                                       |

{% hint style="info" %}
Both run on the same semantic layer, so the answers match no matter where you ask.
{% endhint %}

### What to expect

AI Analyst is built for answering questions about your data. A few expectations worth setting with your team:

* **Data pulls are the sweet spot.** Numbers, rankings, breakdowns, and trends come back reliably. For open-ended analysis, iterate with follow-up questions and sanity-check results before relying on them.
* **It answers questions; it doesn't build reports.** Use AI Analyst for the question of the moment, and dashboards and Explores for the views your team reviews on a schedule.
* **Read the assumption notes.** When your question is ambiguous ("last 30 days"), the answer tells you exactly what window and filters it used.

### Security and governance

AI Analyst only sees what your workspace is allowed to see. MCP connections are read-only, scoped to your workspace, time- and usage-limited, and revocable at any time. Queries run through your validated semantic layer and respect your workspace's governance rules - connecting an AI tool never opens a side door around them.

### Licensing

AI Analyst is an add-on license. If you're a workspace administrator and don't see it in your workspace, contact your Daasity account representative.

### Get started

{% stepper %}
{% step %}

### **Ask your first question in the app**

Open AI Analyst in your workspace and try one of the example questions that open by default.
{% endstep %}

{% step %}

### **Connect your own AI tool**

Add Daasity as a connector in Claude, Cursor, Codex CLI, or ChatGPT. See [Connect Your AI (MCP)](https://help.daasity.com/core-concepts/ai-analyst/connect-your-ai-mcp).
{% endstep %}
{% endstepper %}

### FAQ

**Why should I connect my AI through Daasity instead of straight to my warehouse?**\
A direct warehouse connection gives your AI every table and leaves it to guess the joins, filters, and metric definitions - which is how you get a number that doesn't match your reports. The Daasity connector points your AI at your curated data models instead: the same models behind your dashboards, with valid-order logic, timezone handling, and metric definitions already built in. It can only query those models. Raw connector tables like your Shopify or Amazon extracts aren't reachable through it, so there's no path to an unmodeled number.

**How do I know a number is right?**\
Every answer names the view it came from, and every metric arrives with its definition and the table it reads from, so your AI can show its work. If a question can't be answered from a curated view and your AI falls back to writing SQL, it's required to say so alongside the number and point you back to your dashboards to reconcile. In the in-app AI Analyst, an answer that maps to one of your Explores is marked **Daasity Verified**, and you can expand any answer to see the rows behind it.

**Where does the answer actually come from?**\
Your curated semantic views, first and always. Only when no view can answer does your AI fall back to SQL - and then only against Daasity's curated data models, never raw connector tables. Fallback answers are labeled, so you always know which kind you're looking at.

**Who can use it, and how do we turn it on?**\
Access is enabled per user. Send your account manager the names and emails of everyone who should have it. Anyone with AI Analyst automatically has the connector too. Each person connects individually - connections are listed and revoked independently - and everyone gets the same answers because they all resolve against the same views.

**Does this work if we don't use Looker?**\
Yes. Your workspace is provisioned automatically on Daasity's standard semantic views the first time you log in, and everything above applies. If you do use Looker, your own Explores are added on top.

**Does it work with the AI tool we already use?**\
Claude Desktop, Claude on the web, Cursor, Codex CLI, and ChatGPT all have step-by-step setup guides. Any other MCP-compatible tool can connect using the manual token method.

**Which data warehouses are supported?**\
Snowflake, BigQuery, and Redshift. The connector tells your AI which one you're on, so it writes the right SQL dialect.

**Do we keep our custom Explores, or only get the out-of-the-box ones?**\
You keep them. Custom Looks and Explores are regenerated as semantic views alongside the standard set, and your AI can list the full catalog available to your workspace so you can see exactly what's covered.

**We're on a Team or Enterprise Claude plan and don't see "Add custom connector."**

**Can it change or write anything?**\
No. The connector does two things: list the views available to you, and run a read query. Anything that would modify data or schema is rejected. Connections are scoped to your workspace, time-limited, and can be revoked at any time.

**Can different people have different levels of access?**\
Everyone on a connection sees the same modeled, permissioned dataset your workspace already governs. There aren't separate AI-specific permission tiers.

**What are other merchants doing with it?**\
Quick numbers, rankings, and trends without waiting on a report. Period-over-period questions with the reasoning spelled out. Scheduled morning briefs that run when your data lands. Exec-ready summaries built directly from governed numbers. And prototyping a view in chat, using your dashboards as the reference before anything becomes a scheduled report.

**Which model does it run, and can we choose?**\
In the in-app AI Analyst, Daasity manages the model. Through the connector, you use whatever model your own AI tool runs.

**Can it build a dashboard or a Look for me?**\
No - it reads, it doesn't build. Your AI can tell you exactly which fields and filters would produce the view you want, and you create it in Looker.

**Will it see the customizations we built on top of your transforms?**\
Anything modeled inside your Daasity account is visible to it. Files you maintain outside Daasity - separate YAML, external tables - aren't read.

**Can we give it our own definitions or context?**\
In conversation, yes. Tell it which table or definition to use and it carries that forward for the rest of the session. There isn't a file-based context layer today.

**Can several people or tools connect at once?**\
Yes - each connection is listed and revoked independently.