Connect your own AI assistant to your club data

Start from the architecture you own

This builds directly on the [club-owned data architecture](/explainers/how-a-club-owns-its-hockey-data): the club owns the gold data and models, and everything else is a swappable consumer. An AI assistant is just another consumer — a very powerful one — so it plugs in the same way, with read-only access and no special privileges.

The connector is the whole trick

Do not paste rows of stats into a chat window. Instead, expose your gold tables through a **connector** — a read-only SQL tool or an [MCP server](/explainers/what-is-mcp-and-why-it-matters-for-analytics). Now the assistant can request exactly the data it needs, and the connector is the only way in — and it can only read. Your data never leaves the infrastructure you control.

How a question becomes an answer

When you ask "who are our best penalty-kill forwards this month?", the assistant plans a query, calls the connector, reads the gold tables, sanity-checks the result, and returns a grounded answer with the numbers and a short summary. Because it is reading real data — [expected goals](/glossary/expected-goals), [possession](/glossary/possession), shot rates — the answer is traceable back to a query, not invented.

Keep a guardrail

Let the assistant plan and draft, but wrap it: log the generated SQL, check the output against a known total, and keep a human in the loop for anything that drives a real decision. Speed on the boring parts, judgement where it matters.

What you actually get

- **Answers in plain language** — ask like you'd ask an analyst, no SQL or spreadsheets. - **Instant scouting reports** built from your own player profiles. - **Opponent prep in minutes** — tendencies and matchups on demand. - **Post-game summaries** every night, automatically.

And because the gold layer is the contract, you can switch from one assistant to another whenever a better one appears — without changing anything underneath.