OpenChair Assistant

OpenChair Assistant is OpenChair's AI operations assistant. It chats with you, reads your venue data live (clients, bookings, schedule, waitlist), and proposes specific actions you can approve and send — from filling tomorrow's empty slots to running a win-back campaign to texting a single client about their next visit.

Where to find it — Web: sidebar → Assistant. Mobile: bottom tab → Assistant. Requires PRO on both surfaces.

TL;DR

• Chat with Assistant in plain language. It calls live venue data (clients, bookings, schedule, waitlist) to ground its answers.
• 14 action types across gap-fill, reconnect, retention, scheduling, feedback, and automation setup. Owners see all; staff see a focused subset for their own clients.
• Actions either Draft (you review and Send), require Approval (Owner/Manager confirms), or Navigate (one-click to settings). Big sends get a hold-to-confirm gesture; queued sends survive a closed tab.
• Costs ~3 Sparks per action-producing turn (chat + selector + draft). 1 Spark per refine, capped at 3 refines per draft.

What Assistant can do

Three things happen in any given turn:

1. It chats with you. Ask in plain English. Assistant replies using your venue's real data. Two follow-up suggestions appear at the bottom of every reply so you can dig deeper without typing.
2. It calls live data. When your question needs venue context, Assistant calls one of five read-only tools: lookup_client, find_clients, get_booking_history, get_schedule, get_waitlist. A trace strip between the message and the reply shows each call as it happens ("Looking up Sarah…" then "Looked up Sarah (0.8s)") so you can see why a multi-tool turn took as long as it did.
3. It proposes actions. If the conversation looks like something Assistant can act on — fill a gap, win back a lapsed client, draft an inbox reply, share your booking link with a regular — a card appears alongside the reply. You review, optionally refine, and Send (or approve, depending on the action).

Costs

What	Cost
Each chat turn (Assistant thinking + reply)	1 Spark
Action selector (deciding whether to suggest an action)	1 Spark
Draft generation (when an action is suggested)	1 Spark
Refine a draft (up to 3 times per draft)	1 Spark each
Quick nudges (review request, no-show follow-up, gap waitlist offer)	1 Spark each

A typical action-producing turn (chat + selector + draft) costs about 3 Sparks (~3¢). Multi-refine turns add up — refine sparingly.

Sparks are 1 cent each. PRO venues get included Spark allocations per billable bookable team member (see Understanding Spark). When your wallet drops below 100 or 10 Sparks, a soft warning appears; when it hits zero, action drafts fall back to deterministic templates so the conversation keeps working but loses its voice match.

Action catalogue

14 action types, grouped by family. Assistant picks the right one based on what you've discussed.

Gap fill

Action	Risk tier	What it does	Roles
Flash deal	Draft	Creates a real flash-deal URL for an open slot and shows you the shareable link.	Owner / Manager / Staff
Service spotlight	Approval	Drafts a promotional SMS for a chosen service, sent to a cohort you pick.	Owner / Manager

Reconnect

Action	Risk tier	What it does	Roles
Win-back campaign	Approval	Drafts a batch SMS to clients who haven't been in for 60+ days (window picker: 30 / 60 / 90 / 180). Honours opt-out, recently-messaged dedup, and a 90-day per-wave cap.	Owner / Manager
Client outreach	Draft (Approval for non-owners)	One-tap personal note to a single named client. Requires Assistant to have just looked the client up. No cancel window: direct send.	Owner / Manager
Cohort outreach	Approval	Like win-back, but for non-lapsed cohorts you pick (VIPs, top spenders, regulars, new this month).	Owner / Manager

Retention

Action	Risk tier	What it does	Roles
Birthday outreach	Draft	SMS to clients with a birthday today or this week.	Owner / Manager / Staff
Share storefront link	Draft	One-tap message to a client containing your booking link.	Owner / Manager / Staff
Filter clients	Navigate	Opens the Clients page so you can apply filters yourself.	Owner / Manager

Scheduling

Action	Risk tier	What it does	Roles
Booking reschedule	Approval	Finds open slots and proposes a reschedule for a specific booking.	Owner / Manager / Staff

Feedback

Action	Risk tier	What it does	Roles
Review request	Draft	Batch SMS to today's completed bookings asking for a review.	Owner / Manager
No-show follow-up	Approval	Per-target SMS to today's no-show clients with a rebook nudge.	Owner / Manager

Inbox

Action	Risk tier	What it does	Roles
Inbox reply	Draft	Drafts a reply to a single inbox conversation; on Send, the message goes out via the standard inbox path.	Owner / Manager / Staff

Automation setup

Action	Risk tier	What it does	Roles
Enable reconnect automation	Navigate	Surfaces only when reconnect is off and lapsed clients exist. Opens /settings/reconnect.	Owner / Manager
Set up rebooking prompt	Navigate	Opens /automations to configure the post-visit rebook nudge.	Owner / Manager

How risk tiers work

• Navigate: no draft, no send. The card is a one-click route to a settings page. Free.
• Draft: Assistant writes a draft. You review, optionally refine, and Send. The action executes on send.
• Approval required: pending row in the approval queue. An Owner or Manager reviews and approves before anything sends. Owners always bypass approval for actions they originated. Solo venues (single staff, solopreneur onboarding intent) bypass approval entirely.

Approval-required drafts expire after 24 hours. Plain drafts expire after 7 days.

Surface: suggest vs stage

Independently of risk tier, Assistant picks how prominently to surface the card:

• Stage: full card, audience preview, Send button. Used when the conversation has clearly committed.
• Suggest: lightweight chip, "Here's what you could do…" Click to expand into the full card. Used when Assistant thinks the action might be useful but you haven't asked for it directly.

Audience preview and cost

Any action that sends to multiple clients shows an audience-preview card before you Send:

• Count of eligible recipients, refreshed live.
• Sendable sub-count (after opt-out, blocked, recently-messaged dedup, no-contact filters).
• Window picker (win-back only): toggle between 30 / 60 / 90 / 180-day inactivity windows; the count re-queries instantly.
• Sample names: up to 5 first names of recipients with "and N more…".
• Cost lives inside the Send button: Send to 47 clients · A$3.40. NZ shows zero (email-only).
• Wallet warning: if your wallet balance is short for this send, the Send button surfaces a "Wallet balance is short, top up before sending" note and disables until topped up.

Hold-to-confirm Send

For audiences over 50 recipients, the Send button becomes a press-and-hold gesture. Hold for about 600ms until the fill bar completes, then release to send. Tapping briefly shows a "Hold to send" hint instead. Reduced-motion users see a static 50% fill bar.

The hold gesture stops accidental blasts. Cmd/Ctrl-Enter doesn't bypass it: it focuses the hold button instead.

Scheduled-sends queue

Win-back campaigns and other server-tracked batch sends use a 30-second cancel window:

1. You approve and Send.
2. The campaign queues with scheduled_for = now + 30 seconds.
3. The card shows a countdown with a Cancel button.
4. If you tap Cancel before the timer fires, the campaign is cancelled. No SMS sent, no Sparks consumed beyond the draft.
5. After 30 seconds (capped at 5 minutes), the cron worker picks up the row and fires the send.

The dual-path design means:

• If your tab stays open, the in-page timer fires first. Snappy UX.
• If you close the tab or background the mobile app, the cron worker fires instead. Campaign still goes out.
• A client who books between queue and fire is dropped automatically from the audience. No double-send risk.

Currently this queue only handles win-back. Other actions execute synchronously on Send (no cancel window).

Suggested starters

Below the chat input, Assistant offers a row of starter chips so you don't start from a blank page.

• FREE venues see static starters on a locked upgrade screen, chips are presentational only; chat is paywalled.
• PRO venues see a mix:
  • Data-aware chips appear first, gated on 10+ historical bookings. Examples:
    • Occupancy below 60%: "How do I fill my empty appointment slots?"
    • Revenue trending down: "What can I do to boost revenue this month?"
    • One staff member generating 50%+ of revenue: "How do I reduce my dependency on one team member?"
    • No-show rate above 10%: "How do I reduce no-shows at my salon?"
  • Cluster-aware static chips: owner sees marketing / client-relations / team-ops / growth; staff sees consultation / skill-development / client-experience / productivity. Rotates daily.
  • Personalised follow-ups: each Assistant reply ends with two suggested follow-ups; reopening a session re-surfaces them.

Mobile parity

Mobile (iOS, Android, iPad) is at full feature parity with web for:

• Chat, tool calls, trace strip
• All 14 action types
• Audience preview, cost display, sample names, window picker
• Hold-to-confirm, scheduled-sends queue
• Suggest vs stage surface tier

Mobile-only surfaces add Smart Nudges in the booking inspector for relevant suggestions.

Mobile FREE shows a passive lock screen with no upgrade verb or pricing (App Store / Google Play compliance, see plan tiers for upgrade).

Role access

Capability	Owner	Manager	Staff
Open Assistant	Yes	Yes	Yes
Send chat messages	Yes	Yes	Yes
Bypass approval queue	Yes (own actions)	No	No
Win-back / cohort / spotlight / no-show actions	Yes	Yes (approval-gated)	No
Flash deal / inbox reply / birthday / storefront link	Yes	Yes	Yes
Reschedule / client outreach to own clients	Yes	Yes	Yes (own clients)
Enable automations	Yes	Yes	No

Staff are scoped to clients they've personally serviced via the find_clients, get_booking_history, get_schedule, and get_waitlist tools. They can't surface a stranger's data, intentionally.

Common mistakes

Problem	What to check
Chat returns "Upgrade to PRO"	Your venue is on FREE. Assistant requires PRO.
No actions are suggested no matter what I ask	You may have fewer than 10 historical bookings, the data gates on most actions need a baseline. Or the conversation isn't on a topic the registry recognises (it matches against keywords like "gap", "win-back", "review", "birthday").
The Send button shows "Wallet balance is short"	Your wallet doesn't have enough for the calculated send cost. Top up from Settings → Billing or pause the campaign.
Send fired but no SMS went out	Check Engage Activity Feed, the send may have been dropped by quiet hours, opted-out targets, or wallet pause. Recently-messaged dedup also silently filters duplicates.
Refine button is greyed out	You've hit the 3-refine cap on this draft. Discard and re-trigger the action if you need more iterations.
Approval queue shows expired actions	Approval-required actions expire after 24h; plain drafts after 7 days. Once expired they can't be sent, re-trigger from a new chat turn.
Trace strip shows "permission_denied"	A staff member tried to call find_clients (owner/manager only) or asked about a client they don't service. Assistant falls back to a deterministic message.

FAQ

What does OpenChair Assistant actually do?

It's a chat surface that can call your venue data live (clients, bookings, schedule, waitlist) and propose actions you can approve and send. Fourteen action types across gap-fill, reconnect, retention, scheduling, feedback, and automation setup. Owners get full access; staff get a tighter subset focused on their own clients.

What does each Assistant turn cost?

1 Spark for the chat turn. A turn that proposes a draft adds 1 Spark for the action selector and 1 for the draft generation, so a typical action-producing turn is about 3 Sparks. Each refine of a draft costs 1 more Spark, capped at 3 refines per draft. Sparks are 1 cent each; PRO venues get included allocations per billable bookable team member.

Why is the chat saying 'Upgrade to PRO'?

Assistant chat is PRO-only. FREE venues see a passive lock screen on mobile and an upgrade screen on web with starter chips. Switching to PRO unlocks chat, all action types, and data-aware suggestions.

What's the difference between Send, Queue, and Hold-to-confirm?

Single-recipient and small-audience sends fire instantly when you tap Send. Audiences over 50 recipients get a hold-to-confirm gesture (press the button for ~600ms) so accidental taps don't blast clients. Win-back campaigns and other server-tracked sends queue for a brief cancel window before firing, even if you close the tab the cron worker picks them up.

Will it ever send something I didn't approve?

No. Every action either drafts for your review (you tap Send) or requires explicit approval from an Owner or Manager. Navigate actions just open a settings page; they don't send anything. Owners bypass the approval queue for actions they originate; solo venues bypass approval entirely.

Why is a specific action expected but not appearing?

Most actions have data gates. Win-back needs lapsed clients; review requests need today's completed bookings; no-show follow-up needs today's no-shows. If the gate isn't met, Assistant won't suggest the action even if you ask for it. Check the Intelligence dashboard for the relevant counts.

Can I see what was sent?

Yes. Every batch send writes to the Engage Activity Feed. Single-client outreach appears in that client's conversation history. Approval-queue history is server-side; the operator who approved is recorded.

Related Articles

• Assistant Actions & Backfill
• Gap Fill Automation
• Coach (AI Business Advisor)
• Understanding Spark
• Engage Overview
• Intelligence Dashboard