The Network health brief lives on the HQ dashboard, the overview page for your multi-location network. It takes no input from you beyond the network itself — click the "Brief me" button and it reads across every location in your org using numbers your dashboard already computes elsewhere: recent revenue, royalty compliance and overdue statements, cancellation/no-show rates, and each location's latest compliance audit. It returns a short written brief naming which locations need a call this week, in priority order, and why. Nothing happens automatically as a result of running it — it's a read-only summary for you to act on; it doesn't send a message, change a setting, or touch any location's data.
Generating a brief
On the HQ dashboard, the "Network health brief" card starts empty with a "Brief me" button. Clicking it fans out across your organization's existing rollups and asks the AI to write the brief; while that's running the button shows a "Scanning the network…" state. Once it returns, the card shows the paragraph, a badge with your total location count, and — when at least one location is flagged — a "need attention" badge with the count. A small colored dot next to the card title tells you whether the text you're reading came from a live AI call (green) or a fallback (amber); see Things to know. The last brief you generated stays on the card if you navigate away and come back, and a "Refresh" button reruns the whole thing and replaces it with a fresh one.
How the priority list is built
The ranking isn't left up to the AI to invent. Before the AI is ever called, the app scores every location in your network against its own recent numbers: whether revenue is running notably below that location's own trailing average, whether royalty compliance is weak or statements are overdue, whether cancellations and no-shows are running high, and whether the location's latest compliance audit came back as a warning or a fail. Locations are sorted worst-first, and only the ones actually flagged for something are handed to the AI to narrate — capped at the eight neediest, so a large network doesn't bury the ones that actually need a call this week. The AI's job is limited to writing the paragraph from that pre-built, ranked list; it can't flag a location the numbers don't support or cite a figure that wasn't handed to it. If nothing in your network is flagged, the brief says so plainly instead of manufacturing something to report.
Things to know
- Plan and permission: this requires the HQ plan, and only an org owner/admin — or a teammate specifically granted the equivalent permission — can successfully generate a brief. A viewer-level teammate can see the rest of the HQ dashboard, including this card and its "Brief me" button, but generating a brief is gated behind the scenes to admin-level access, so clicking it won't produce a brief for them.
- Stub mode: if the workspace doesn't have ANTHROPIC_API_KEY configured, the brief falls back to a simple templated sentence built from the same real flagged data — it still names the actual worst-flagged location and its top reason (or says the network is clean, if nothing's flagged) — but it isn't the AI-written narrative you'd get live. That's what the amber dot means.
- The compliance signal depends on a scan having been run: a location's compliance-audit input only shows up if someone has actually run a compliance scan for it from the HQ Compliance page. It isn't recomputed automatically on a schedule, so a location nobody has scanned recently simply won't surface a compliance-related reason, even if everything else about it looks fine.
- It's a recent-window read, not all-time history: the revenue and cancellation numbers behind the brief reflect a recent period, not a location's full history — a location that had a rough stretch last quarter but has since recovered won't be flagged for that.
- Only the network's neediest locations get named: the AI only writes about the flagged locations, capped at the eight most urgent — if more than eight are flagged, the brief covers the worst of them, not the full list.
- Read-only: generating a brief never sends anything, changes anything, or takes action against a location — it only summarizes numbers that already live elsewhere on your dashboard.
Frequently asked questions
Q: Does the brief take any action on the flagged locations, or just tell me about them?
A: Just tells you. Generating a brief is entirely read-only — it doesn't send a message, adjust pricing, or change anything about a location. You decide what to do about anything it flags.
Q: Can the AI flag a location, or cite a number, that isn't actually accurate?
A: No. The app decides which locations get flagged and why, using your real numbers, before it ever asks the AI to write anything. The AI's only job is turning that pre-built list into a short paragraph — it can't invent a flagged location or a statistic that wasn't handed to it.
Q: Who can generate a Network Health brief?
A: It requires the HQ plan, and only an org owner/admin — or a teammate specifically granted the equivalent permission — can generate one. A viewer-level teammate can see the rest of the HQ dashboard, including this card and its "Brief me" button, but generating a brief is gated behind the scenes to admin-level access — clicking it won't produce a result for them.
Q: Why does the brief say a location is fine when I know it had a rough audit result recently?
A: The compliance signal only reflects the most recent scan actually run for that location from the HQ Compliance page — it isn't recomputed on its own, so if no one has re-run a scan since then, the brief may be working from an older or missing result. The revenue and cancellation signals are also limited to a recent window, not the location's full history.
Q: What does the colored dot next to "Network health brief" mean?
A: Green means the paragraph you're reading was actually written by the AI from your network's real numbers. Amber means the workspace's ANTHROPIC_API_KEY isn't configured, so you're seeing a simpler templated fallback sentence — it's still built from your real flagged data, just not the full AI-written narrative.
Q: Does refreshing the brief cost me anything or count against a limit?
A: The brief pulls fresh numbers every time you click Refresh, so it's meant to be re-run whenever you want a current read. It doesn't run automatically in the background — it only regenerates when you or another admin explicitly click the button.