Data
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.
What the worker does
An AI worker handles the intake step for data requests. When a request arrives in Slack, the Polaris Inbox turns it into a prefilled task suggestion, and the worker restates the ask as an answerable question, names the tables it would need through the Supabase connection, and posts the clarifying questions that were skipped. The analyst decides whether the request is worth doing at all.
- Trigger
- Every incoming request
- Connections
- Slack, Supabase, Notion
- Output
- A scoped, answerable question
The request arrives without any of the information needed to answer it
Can you pull last month by plan. Which last month, calendar or trailing thirty days. Which plan field, the one on the subscription or the one on the account. Does churned count. Is this for the board deck on Thursday, in which case the definition has to match the last three decks, or is it curiosity.
The analyst asks two of those questions, gets an answer to one, and builds something that turns out to be the wrong shape. Then it gets rebuilt. The rebuild is the actual cost of bad intake.
What the worker adds before the request is queued
The decision it informs
If nobody can name a decision that changes based on the answer, that is worth knowing before an afternoon is spent. The worker asks, and records the answer or its absence.
The restated question
The ask rewritten so it has exactly one interpretation, with the date range, the grain and the filters stated explicitly.
The tables it would touch
Read from the schema through the Supabase connection, so the analyst knows whether this is a ten-minute query or a joins problem before committing.
The existing answer, if there is one
Half of all requests have been answered before. The worker checks the documented metrics and previous requests, and links the answer instead of queueing a repeat.
From Slack message to scoped task
- 1
The message arrives
Someone asks in a channel or a direct message. The Polaris Inbox catches it and turns it into a task suggestion with the bucket, lane, labels and owner already filled in.
- 2
One click accepts it
Suggestions never become tasks silently. Accepting one creates a fully specified task in the data workstream, which is the point where it becomes real work.
- 3
The worker scopes it
A cloud machine picks the task up, reads the schema, checks whether the question has been answered before, and posts the restated question with its open clarifications.
- 4
The requester answers, or does not
Requests whose clarifications go unanswered for a week are usually requests nobody needed, and now that is visible instead of being an analyst's private suspicion.
- 5
The analyst decides
Do it now, schedule it, or reply with the existing answer. The scoping note is attached, so the decision takes a minute rather than a meeting.
Questions people ask
+Does this make it harder for people to ask for data?
Slightly, and that is the point. A request that survives being restated as a specific question with a named decision is a request worth doing. Teams typically find a meaningful share of asks stop at the clarifying question stage.
+What if the request comes by email instead?
Give the worker the Gmail connection and the same intake applies. The connection catalog covers Slack, Gmail, WhatsApp and the rest of the fixed list, so intake can run wherever your requests actually arrive.
+Can the worker answer simple requests itself?
It can look up documented metrics and previous answers and reply with those. Producing a new number from the warehouse is a different job, briefed separately, and most teams keep query execution behind human approval.
+Where does the scoping note live?
As a comment on the task, which stays with the request permanently. That record is what lets you answer next quarter's identical request with a link rather than a rebuild.
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.
Thirty requests in the backlog and one analyst
A scoping note per request, so the backlog is triaged on value rather than on arrival order.
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.
Connect Slack to Polaris
Six scopes, no access to direct messages, and a bot that only reads the channels somebody invited it into.
Connect Supabase to Polaris
Supabase is the connection where a careless key choice does the most damage, and the only one Polaris cannot yet check for you.
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.
Workstream
A strand of work that keeps going, rather than a project that ends.