Data
Three teams, three definitions of active user, one meeting
Every definition of a metric found and listed side by side, with the source of each.
What the worker does
An AI worker can find every definition of a metric that exists in a company. It reads dashboards, documents and past discussions through the Supabase, Notion and Slack connections, lists each distinct definition side by side with its source, and drafts a single definition page stating the rule in words. Which definition becomes canonical is ratified by people, and the resulting document is versioned.
- Runs
- Once, then on demand
- Connections
- Supabase, Notion, Slack
- Output
- Definitions, side by side
Nobody decided to have three definitions
Marketing counted anyone who opened the app. Product counted anyone who completed a core action. Finance counted anyone on a paying account, because that was the number in the model. Each definition was reasonable in the room where it was made and none of the three rooms knew about the others.
The discovery usually happens in a meeting, with a slide up, when two numbers that should match do not. Then forty minutes go on archaeology instead of on the decision the meeting was called for.
Where definitions hide
In query logic
The real definition is the WHERE clause, and it usually differs subtly between two dashboards that carry the same title.
In documents
Written definitions in Notion or Google Drive, often accurate on the day they were written and never revised after the logic changed.
In Slack threads
The most common location for the actual working definition, agreed in a thread by three people and never written anywhere permanent.
In the deck
A number in a board deck with a footnote nobody has re-read, which is frequently the definition the company is actually managed by.
What the delivered comparison looks like
| Column | Contents |
|---|---|
| Definition | The rule, stated in words rather than as SQL |
| Source | Where it was found, with a link |
| Last touched | When that source was last edited |
| Population | Who it counts and who it excludes |
| Used by | Which dashboards, docs or decks depend on it |
After the decision
The ratified definition goes into a versioned document in Polaris Docs, with review comments, so a future change to it is visible rather than silent. Each dependent dashboard that needs updating becomes a task in the data workstream.
Six months later, when somebody proposes changing the definition again, the version history shows what it was, when it changed and who agreed, which is the argument that the meeting would otherwise have to reconstruct from memory.
Questions people ask
+How does the worker find definitions in Slack?
Through the Slack connection, searching the channels you point it at for discussion of the metric by name. Threads where a definition was agreed are listed with links, so the archaeology is checkable rather than something you take on trust.
+Can it rewrite our dashboards to match the ratified definition?
No. It lists the dependent dashboards it found, and each one becomes a task for a person to update. Changing query logic on live dashboards is exactly the kind of write action that stays with the data team.
+How many metrics should we do at once?
Start with the three or four that appear in board reporting, because those are where a mismatch is most expensive. Doing twenty at once produces a document nobody ratifies, and an unratified definition page is no better than the confusion it replaced.
+What if two definitions are both legitimate?
Then the answer is two named metrics rather than one contested one, and the definition page says which is used where. The worker will surface the conflict; naming both is the human decision that ends the argument.
Related
One analyst, thirty requests, and an AI worker in between
Five jobs that stand between a data team and the analysis they were hired to do.
The Monday metrics email that somebody writes on Sunday night
The numbers pulled and the paragraph written, delivered as a file rather than a link.
Can you pull the numbers, sent as a direct message at 6pm
Every ask restated as a question, with the clarifying questions asked before anyone writes SQL.
Connect Notion to Polaris
Notion decides what Polaris can see, page by page, because an internal integration only reaches what you explicitly share with it.
Connect Slack to Polaris
Six scopes, no access to direct messages, and a bot that only reads the channels somebody invited it into.
Hire an AI data analyst
It settles what active user means, writes the query that matches the definition, and turns the numbers you hand it into a paragraph an executive can read.
Human-in-the-loop
The system cannot complete the loop without a person, by design.