Chat intake assistant that triages ad hoc data requests
Problem
A commercial manager or coaching staff member wants a number and messages the data team directly, on whatever chat tool the organisation already uses, at whatever hour it occurs to them. The request is usually vague, something like can someone pull last month's numbers for me. An analyst has to reply, ask what months, what number, what fixture list is meant, then build a one off query. The same question was probably answered for someone else three weeks earlier, but there is no record of it, so it gets rebuilt from scratch. Requests sent after six in the evening sit unread until the next working day, and whoever eventually answers has no way to tell an urgent board request from five routine ones, because they all arrive in the same thread.
Product idea
A chat assistant that lives in the messaging tool the team already has open. Someone types a question in plain language. If it matches a metric the analytics team has already defined and documented, the assistant answers immediately, states the definition being used, and links to the source report rather than inventing a fresh number. If there is no match, it asks for the missing detail (date range, entity, purpose, deadline) and turns the exchange into a structured ticket in a queue an analyst owns, tagged with who asked and why, and checks first whether an open ticket already covers the same ask. It does not write new SQL and does not answer a question that is not already modelled; unmodelled questions go to a person.
Who it is for
Insight and BI analysts who own the request queue, analytics engineers who maintain the metric catalogue, and the commercial, marketing and coaching staff who currently message the data team directly for numbers. Sponsored by the head of data.
Possible first version
A chat app connected to the messaging platform already in use, with a manually maintained catalogue of the twenty or so metrics the analytics team already reports on, each with its definition, owner and a link to the existing report. Natural-language matching against that catalogue returns an instant answer with the link when there's a hit. Anything unmatched becomes a structured ticket with requester, deadline and purpose, visible in a simple queue view. Out of scope for version one: no live connection to the warehouse to compute a fresh number, no automatic SQL generation, and no integration with a separate ticketing tool.
- Build classification
- Workflow application
- Rough effort
- 5-week first release
- Roles involved
- Insight and BI analyst, Analytics engineer, Head of data, Head of IT
- Relevant to
- Professional club, League office, Federation / governing body, Collegiate athletics
- Systems in play
- Messaging apps, Data catalogues and quality monitoring, Business intelligence and visualisation tools, Spreadsheets
- Product framing
- Voice or chat interface
Questions we get asked
What has to exist before this can go live?
A documented catalogue of the metrics the analytics team is willing to stand behind, even if it starts as a spreadsheet of twenty rows with a definition, an owner and a link to the report. If that documentation does not exist yet, building it is the real first project, not the chat assistant. Skipping it just moves the same vague requests into a slightly different chat window.
Our team already just asks in the group chat and someone always answers eventually. Why change that?
That holds up until you try to work out how many hours a week go into answering the same question a second or third time, or explain to a board member why last month's attendance figure does not match the one given out in April. A group chat has no memory and no record of what was asked or by whom. This gives the same informal habit a queue and a paper trail, without asking anyone to file a proper ticket.
Does this replace the BI dashboard we already pay for?
No. It sits in front of it. The assistant never builds a new report or a new visualisation, it only points a requester at the existing one and, where nothing exists, tells the analyst that a gap has been found. The dashboard and its data stay exactly where they are; what changes is how a plain-language question either finds them or gets queued for someone who can build them.
What happens if someone asks a question at 11pm and no analyst is on duty?
The assistant still answers if the question matches something already in the catalogue, because that answer does not depend on a person being awake. If it does not match, it tells the requester it has logged the request and gives an honest expectation of when someone will look at it, rather than guessing at a number to seem responsive. It never fabricates an answer to fill the silence.
Is this your workflow?
Tell us one sports workflow that still runs on paper, spreadsheets, WhatsApp or an outdated system. We will map it and show you what a simpler product looks like.
Tell us about itMore in Data platform & engineering
- Access permission drift monitor for warehouse and reporting toolsA monitor that compares entity access rules across the identity provider, warehouse and reporting tool, and alerts a named owner when they drift out of sync.
- Attendance reconciliation report across ticketing and till systemsA reconciliation report that turns a ticketing export and a till export for the same fixtures into one defensible attendance and spend figure, with the variance between them shown and explained.
- Consent and deletion request tracker across the vendor stackA workflow tool that turns a consent revocation or deletion request into tracked, assigned tasks across every downstream system, replacing the spreadsheet checklist and email chain nobody can prove was completed.