Skip to content

Slack AI Analyst Tutorial ​

End-to-end walkthrough: build a stock-market analyst pipeline with the Bruin CLI, deploy it to Bruin Cloud, and expose it as a Slack agent your team can query from any channel.

Audience: data professionals deploying an AI analyst to Bruin Cloud and Slack.

Prerequisites

  • Bruin CLI installed and authenticated.
  • Claude Code available for pipeline generation and bruin ai enhance.
  • A Bruin Cloud account with access to Team settings and Projects.
  • A Git repo containing your Bruin project.
  • Slack workspace with bot credentials and channel access.

1. Initialise the Bruin project ​

Run bruin init empty <pipeline-name>. If the current folder is already git-initialised, this creates <pipeline-name> unless you pass --in-place. If the current folder is not a Git repo, Bruin creates a bruin/ folder first and then creates the project and pipeline inside it.

See Project for context on Bruin projects.

2. Build the pipeline ​

Use Claude to extract stock data from Yahoo Finance and Wikipedia. Build assets that clean and join the data into something useful for an analyst: daily price tables, market-cap rankings, revenue and free-cash-flow rollups, and so on.

3. Enhance metadata ​

Run bruin ai enhance across the assets. This adds descriptions, column metadata, quality checks, and lineage. Review the output before committing.

4. Add the repo to Bruin Cloud ​

  • Open Bruin Cloud → Team settings → Projects and add the repo to your workspace. See Projects.
  • Enable the pipeline and trigger the first run. See Pipelines.
  • Confirm backfills and the daily schedule run as expected.

5. Create the AI agent ​

  • Open AI → Agents and create a new agent. See Configure Agents.
  • Select the project (the repo you just added).
  • Attach the connection set the agent should query against.
  • Add the Slack integration and pick the target channel. See Slack.
  • Name the agent and save.

6. Add agent instructions ​

Create an AGENTS.md file in the project root with the pretext, context, rules, and instructions for the analyst. A good AGENTS.md should:

  • Describe what the analyst is for, who uses it, and what kinds of questions to expect.
  • Tell the agent which assets to prefer for which question types.
  • Require bruin query for all data access, and use --dry-run while testing.
  • List any business rules or definitions (revenue growth, free cash flow margin, ticker matching) that the agent needs to apply.

7. Test in Bruin Cloud ​

  • Open the agent's chat in Bruin Cloud and ask a few questions. See Chat with Agents.
  • Confirm it can query the data and self-correct when its first query is wrong.

8. Test in Slack ​

  • Mention the agent in a Slack channel and ask a stock-market question.
  • Open the generated SQL to validate the logic.
  • Request a PDF report and confirm it lands in the channel.

Sample prompts ​

  • "Which companies had their free cash flow margin improve in the past 4 quarters but saw their stock price decrease more than 10% during the same period?"
  • "Summarize the top 10 tickers by revenue growth and generate a PDF report."

Next ​