<aside> 📊

Help every team answer data questions without waiting on an analyst

Data Scout is an always-on data analysis agent that turns natural-language questions into governed, source-backed answers.

For teams with growing data needs, this agent helps product, operations, and go-to-market teams explore adoption, usage, pipeline, and performance trends while keeping analysis grounded in trusted sources.

At Notion, Data Scout serves data questions across 1,400 employees—turning a data bottleneck into self-service analytics and freeing analysts to focus on work that actually needs them.

</aside>

How it works

• Answers data questions in plain language, identifying the metric, timeframe, segment, and detail level before querying.
• Queries approved read-only sources — warehouse connections, semantic views, and dashboards — using governed paths instead of ad hoc pulls.
• Returns the answer with full context: key trend, timeframe, filters applied, and the source used.
• Escalates to the data team when the request needs a large extract, unsupported visualization, or deeper investigation.

Build your own

1. What to have ready before you start

• Your workflow: The sequence of steps the agent runs, from question intake to final answer. Know this before you build.

  For Data Scout:

  1. A teammate asks a data question.
  2. The agent identifies the metric, timeframe, segment, and level of detail needed.
  3. The agent checks approved definitions and queries the best trusted read-only source.
  4. The agent replies with the answer, supporting context, and the next useful cut when relevant.

• Context it needs: The information the agent draws on to answer questions accurately. You'll swap these into the starter prompt in step 2.

  For Data Scout:

  • [ ] Read-only warehouse or analytics source → [WAREHOUSE OR ANALYTICS SOURCE]
  • [ ] Semantic layer or metric definitions → [SEMANTIC LAYER OR METRIC DEFINITIONS]
  • [ ] Table documentation → [TABLE DOCUMENTATION]
  • [ ] Approved dashboards or reports → [APPROVED DASHBOARDS OR DOCS]
  • [ ] Data team escalation path → [DATA TEAM ESCALATION PATH]

• Tools to connect: The apps the agent needs access to for reading, querying, and responding.

  For Data Scout:

  • [ ] :notion: Notion (to read metric definitions, documentation, and approved reports)
  • [ ] Read-only analytics or warehouse connection (to query governed data sources)
  • [ ] Slack or Notion intake channel (to receive questions and reply in context)

2. Copy and paste this starter prompt

<aside> <img src="/icons/info-alternate_gray.svg" alt="/icons/info-alternate_gray.svg" width="40px" />

In your left-hand sidebar, go to the Agents section and click the + button.

</aside>

You are a data analysis agent for the [TEAM NAME] team. Use [WAREHOUSE OR ANALYTICS SOURCE], [SEMANTIC LAYER OR METRIC DEFINITIONS], [TABLE DOCUMENTATION], and [APPROVED DASHBOARDS OR DOCS] to answer natural-language data questions. Clarify only when the request is ambiguous. Return a direct answer, explain the timeframe and filters, and note the source you used. If the request needs a large extract, unsupported visualization, or deeper investigation, say so and recommend the right next step through [DATA TEAM ESCALATION PATH].