Reports
Intelligence Chat (Conversational Analytics)
Last updated 19 May 2026
Intelligence Chat (Conversational Analytics)
Ask questions about your business data in plain English. Intelligence Chat is a grounded Q&A surface inside the Intelligence dashboard that uses your real venue data for the selected period to answer questions about revenue, occupancy, staff performance, service mix, funnel, and goals.
Where to find it — Web: Intelligence dashboard → Concierge Insight Card → switch to Conversation mode. Mobile: Intelligence tab → same card.
TL;DR
- Conversational interface inside the Intelligence dashboard.
- Two modes on the card: Briefing (AI narrative summary of the period) and Conversation (interactive Q&A).
- Up to 5 turns per session; session resets when you change date range.
- 1 Spark per turn. PRO only.
What it answers
Questions grounded in your venue's real data:
- "Why is my revenue lower than last month?"
- "Who's my top-performing stylist this period?"
- "What service is making most money for me?"
- "How am I tracking against my monthly goal?"
- "My Tuesdays are quiet — when do I get most bookings?"
- "What's my booking conversion rate?"
The answers are bounded to the date range you've selected on the dashboard. Switch the range to switch context; that also resets your 5-turn limit.
What it knows
On the first turn of every session, the chat is fed a context block of your real data:
| Data point | What's in it |
|---|---|
| Current period KPIs | Revenue, bookings, average ticket |
| Previous period KPIs | Same metrics for the prior equivalent window |
| Top staff | Top 3 by revenue, with bookings and avg ticket |
| Top services | Top 3 by revenue, with % of total |
| Funnel summary | Visitors, page views, booking rate |
| Goal progress | If you've set a revenue goal — pace, days remaining, on-track/amber/red |
| Industry benchmarks | Relevant peer data |
| Currency awareness | AUD / NZD / GBP |
Subsequent turns in the same session reference the same context — the chat doesn't refetch live data on every message.
What it doesn't do
- No charts — text replies only. The dashboard around the chat shows the charts; the chat explains what's in them.
- No agentic tool calls — Intelligence Chat doesn't look up specific clients, send messages, or take actions. For that, use the operator Concierge.
- No date ranges outside the dashboard selection — you can't ask about "last year" if your dashboard is set to "last 7 days". Switch the range first.
- Limited to your own venue's data — no cross-venue comparison even for multi-venue accounts.
How it differs from Coach
| Coach | Intelligence Chat |
|---|---|
| Gives advice and recommendations | Answers questions about data |
| "What should I do about empty Tuesdays?" | "How quiet are my Tuesdays vs the rest of the week?" |
| Uses agentic tools (lookup_client, find_clients, get_schedule, get_waitlist) | Uses a single grounded data context, no tools |
| Can propose actions you approve and send | Read-only Q&A |
| Lives in the Concierge tab | Lives inside the Intelligence dashboard |
| Costs ~3 Sparks per action-producing turn | 1 Spark per turn |
Both are PRO. They complement each other: ask Intelligence Chat to understand what's happening; ask Coach what to do about it.
Suggested follow-up chips
Each reply ends with two follow-up question chips parsed from the AI's tail block. They appear under the reply as one-tap continuations so you can dig deeper without typing.
Example flow:
"Why is my Tuesday revenue down?"
"Your Tuesday revenue is A$1,840 this period, down 22% from last period's A$2,360. The dip is concentrated in the morning slot — 9-12 bookings dropped from 12 to 5. Sam took 4 fewer Tuesday-morning bookings this period; other staff are roughly flat."
Suggested follow-ups: "Are Sam's other days down too?" · "What services are Sam's morning bookings usually?"
Cost
Each turn costs 1 Spark (HEAVY_ORCHESTRATION). Sparks are 1 cent each.
If your wallet is empty, the chat returns INSUFFICIENT_BALANCE and you can't continue until you top up. The 5-turn session limit is independent — empty wallet stops the chat regardless of turn count.
Turn limit
5 turns per session. The limit resets when you change the date range (which starts a new session). The hard cap exists so an open-ended exploration doesn't escalate into 50 turns of Spark charges without you noticing.
If you hit the limit and want to continue:
- Switch the date range to something else (or back to the same range) — this resets the session.
- Or wait for the cap to clear (caps are per-session, not per-day).
Each turn supports up to 10 prior messages in history; messages over 4000 characters are rejected.
Mobile parity
Mobile shows the same Concierge Insight Card on the Intelligence tab with the same two modes. Identical 5-turn cap, identical 1-Spark cost, identical follow-up chip behaviour.
Telemetry and safety
- Out-of-scope responses (the AI saying "I don't have that data") are tracked for eval purposes; you'll see fewer of these as the system improves.
- Input is sanitised for prompt-injection attempts; flagged inputs alert ops.
- Output is canary-token-checked to detect prompt leakage; PII is scrubbed before any telemetry leaves the venue.
- The chat never reveals it's AI; if asked directly, it deflects to its actual job.
Common mistakes
| Problem | What to check |
|---|---|
| Chat says it doesn't have data on something I know I have | The chat only sees the date range currently selected. Expand the range and ask again. |
| Hit the 5-turn limit too quickly | Switch the date range to reset the session. Use Briefing mode for static summaries that don't consume turns. |
| Chat said something wrong about my revenue | Refresh the page; the context block is fetched on session start. If the data still looks wrong, check Intelligence dashboard directly — the chat reads the same numbers. |
INSUFFICIENT_BALANCE error |
Wallet is empty. Top up from Settings → Billing. |
| Wanted Coach-style advice | Switch to the Concierge tab — Coach is the advice surface; Intelligence Chat is Q&A. |
FAQ
What is Intelligence Chat?
A conversational interface inside the Intelligence dashboard. Ask questions in plain English ('Why is my Tuesday revenue down?') and get a grounded answer based on your real venue data for the selected period. Up to 5 turns per session; 1 Spark per turn; PRO.
How is this different from Coach?
Coach gives business advice and recommendations ('try a flash deal on Mondays'). Intelligence Chat answers questions about what's true right now in your data ('what's my Monday revenue trend?'). Coach uses agentic tools and takes actions; Intelligence Chat answers grounded Q&A only, no tool calls, no actions.
Does it return charts inline?
No. The chat returns Markdown text only. The Intelligence dashboard around the chat shows the charts; the chat is for asking questions about what's in them. Two follow-up question chips appear at the end of every reply so you can dig deeper without typing.
Why is the chat locked after a few questions?
Each session is capped at 5 turns. The cap resets when you change the date range (which starts a new session). Designed to keep cost predictable; one open-ended question can stretch into many follow-ups otherwise.
Can I export the chat?
Not in the current version. The conversation is held in the page state; refresh and you start fresh. If you need to preserve insights, copy them out into your notes.
Why doesn't the chat answer questions about specific clients?
Intelligence Chat is bounded to your dashboard data — aggregate metrics, not individual client history. To ask about a specific client, use the operator Concierge, which has lookup_client and get_booking_history tools for that.