Instagram DM Concierge

Instagram DM Concierge brings your salon's Instagram DMs into Engage and lets AI draft or auto-send replies grounded in your real business data — services, prices, portfolio, hours, and live availability.

Where to find it – Web: Settings → Integrations → Instagram (connect + configure). Engage → Inbox → Instagram DMs (day-to-day threads).

Mobile: Messages tab → Instagram DMs.

---

TL;DR

• AI classifies every inbound Instagram DM into one of seven intents and replies using your own venue data — never invented prices or fabricated availability.
• Per intent, you choose whether AI runs silently (auto-send), drafts for your review (draft), or stays out of it (off).
• Complaints always escalate to you — they never auto-send.
• Every auto-sent message has a 5-minute retract window.
• Kill switch pauses auto-send globally in one tap.
• Costs 2 Sparks per draft or auto-send; bundled into PRO.
• Mobile parity — draft, send, retract, handover from iOS or Android.

Before you connect

You need:

1. An active PRO subscription.
2. An Instagram Business account (not a personal account).
3. That IG account linked to a Facebook Page you manage — Meta's policy requires the link for any automation to access messages.
4. Permission on that page to manage messages.

If your Instagram account is currently "personal" or "creator", open Instagram → Settings → Account → Switch to Business, then link it to your Facebook Page in Meta Business Suite.

Connecting Instagram

1. Go to Settings → Integrations → Instagram.
2. Tap Connect with Instagram.
3. Sign in with the Facebook account that manages your salon's Page.
4. Grant the permissions Meta asks for — Instagram messages, Pages messaging, and Business management.
5. You'll be redirected back to Settings with a green "Instagram connected as @your_handle" toast.

If you see "Your Instagram account needs to be linked to a Facebook Page"

Meta rejects connection attempts when the IG Business account isn't linked to a Facebook Page. Open Meta Business Suite, go to Settings → Accounts → Instagram accounts, link your IG account to a Page you own, then return to OpenChair and tap Try again.

If the connection expires

Meta long-lived tokens last about 60 days. We refresh them automatically every day, but if your token lapses (you disconnected the app, changed your password, or revoked Meta permissions) the settings page will show an amber "Connection expires in N days" warning or a red "Connection expired" banner. Tap Reconnect to restore it.

Configuring what AI handles

Once connected, the settings page unlocks the full configuration form.

AI mode per intent

Six configurable intents (plus Complaint, which always escalates):

Intent	What it covers
Pricing	"how much is balayage?", cost questions
Hours	"are you open Sunday?", closed days
Portfolio	"can I see your colour corrections?"
Availability	"do you have Saturday?", open slots
Booking link	"can I book balayage with Sam?" — replies with a deep-linked booking URL
General	Greetings, thanks, small talk — always drafts, never auto-sends

For each intent you pick one of three modes:

• Off — AI doesn't touch messages with this intent. They land in your inbox as unread.
• Draft — AI writes a reply and shows it above the composer. You send, edit, regenerate, or handover.
• Auto-send — AI replies directly, subject to safety checks below.

Our advice: start with every intent on Draft. After a week of reading AI drafts, switch the intents you trust to Auto-send one at a time. Pricing and hours are usually safe first switches; bookingLink is strictest because the wrong link books the wrong thing.

Brand voice

Pick one of three presets — Warm & casual (default), Professional, Luxury — or write a custom voice in your own words. The custom textarea caps at 500 characters; be specific ("Warm and friendly, never formal, British English, drops the odd emoji").

Use Preview on sample threads before enabling Auto-send for the first time. It runs the AI on three canned customer messages (pricing, availability, portfolio) using your current voice and shows the output side-by-side. Each sample charges 2 Sparks (same as a real reply).

Avoid-list

Up to 5 phrases that AI is never allowed to use. Common entries: "sweetie", "darling", "hun", "love" — anything you don't want going out on your behalf. AI replies are checked against the list before sending; any leak triggers a retry with a tightened prompt, and a second leak forces the reply into Draft mode instead of auto-sending.

Safety caps

Hard limits on how much AI can auto-send per hour and per day, per Instagram account. Defaults:

• 20 auto-sends per hour — once exceeded, AI falls back to Draft until the hour rolls over.
• 100 auto-sends per day — once exceeded, AI falls back to Draft for the rest of the day.

Raise these if a viral post is generating legitimate traffic; lower them if you prefer tighter control. You'll get a push notification whenever a cap is hit.

AI disclosure (optional)

A short suffix (up to 120 characters) appended to the first AI-generated outbound message per thread. Some regions recommend disclosing automated messaging; this is a simple way to do it — e.g. "— replied via AI. A human will follow up if needed."

Day-to-day use

When a DM arrives

1. Your customer sends a DM to your Instagram Business account.
2. Within about 30 seconds, the thread appears in Engage → Inbox → Instagram DMs with an IG pill badge.
3. AI classifies the intent and — depending on your settings — drafts a reply, auto-sends, or escalates.

The AI draft card

When AI drafts a reply it appears in a card above the composer with four actions:

• Send — sends verbatim. We log it as draft_accepted so we can tell how often your AI gets it right first time.
• Edit — inline editor. Your edits are sent as-is; we log them as draft_edited so we can learn where AI tends to drift.
• Regenerate — new draft, new 2-Spark charge. First use shows a confirmation; ticking "don't ask again" suppresses it for 30 days.
• Handover — suppresses AI drafting + auto-send on this thread for 24 hours and flags the thread for you. Useful when a customer goes off-script and you want to take over personally.

Sometimes you'll see a small "AI chose to draft — review carefully" chip above the draft. That means AI had auto-send configured for this intent but fell back to draft because confidence was below threshold or validation failed — a signal to read more carefully before sending.

Auto-sent messages

When AI auto-sends a reply you'll get:

• A push notification on your phone: "AI replied to @customer_handle: '[first 60 chars]…'. Tap to review."
• An "AI sent" chip on the outbound message in the thread, next to a live Retract (4:32) countdown button.

You have 5 minutes to retract. Tapping Retract sends a follow-up disclaimer — "Apologies — please disregard my previous message. A human will follow up shortly." — and flips the thread into your needs-attention zone so you can follow up manually.

Retract is suppressed automatically when:

• The customer has already replied to your auto-send (the conversation has moved on).
• The 5-minute window has expired.

The "AI sent" chip stays on the message permanently for audit purposes.

Global kill switch

If anything goes wrong — a viral Reel floods you with DMs, AI misbehaves, or you just need a break — the Kill switch on the settings page pauses auto-send globally in one tap. Every intent's auto-send mode is coerced to Draft; your drafts keep working.

Turning the kill switch off again does not auto-restore your prior Auto-send settings. You re-enable each intent deliberately — this is on purpose so you're making a fresh decision after whatever incident caused you to pause.

Staff access

Team members can action threads assigned to them — send, edit, retract, handover — using the same inbox surface. Settings (connection, brand voice, kill switch, caps) are owner-only on web.

If a team member retracts an auto-sent message, the audit log records them separately from the original sender so you can see who caught what.

Costs

• Every AI draft or auto-send = 2 Sparks. This maps to the concierge_outbound tier — the same rate as an AI-composed SMS confirmation.
• Complaints and classifier-only skips cost nothing. If AI decides to escalate, no Spark is charged.
• Regenerate = 2 Sparks each time. First-use confirmation is surfaced; subsequent regenerates within 30 days go through without asking.
• Preview on sample threads = 2 Sparks per sample. Three samples = 6 Sparks.
• Included Sparks are bundled in PRO (500 per seat per month, capped at 5,000). Most salons won't exceed their included allocation for this feature alone. Overage is 1 cent per Spark, billed from your wallet.

Troubleshooting

"Your 24-hour messaging window has closed"

Meta policy: salons can only reply to customers within 24 hours of their last inbound message. If a customer messaged you 30 hours ago and you try to send, Meta rejects it. Ask the customer to send another DM before replying.

"You're out of Sparks"

Your included allocation is exhausted and your wallet is empty (or auto top-up is off). Top up from Settings → Billing on the web app, then the draft card refreshes. Auto-send quietly falls back to Draft whenever Sparks are unavailable — no messages are silently dropped.

A customer complained that the AI reply was wrong

Retract within 5 minutes if you catch it. For older issues, the thread's audit trail is in instagram_auto_send_log (our support can help you find it). Use what the AI got wrong to tighten your brand voice or add to your avoid-list.

"Instagram connection expired"

Reconnect via Settings → Integrations → Instagram → Reconnect. Existing threads remain read-only until you complete the reconnect flow.

Messages aren't arriving

Check the connection status on the settings page. If the webhook subscription was lost (rare), disconnect and reconnect — this re-subscribes the page webhook in one flow.

What we don't do

• We only respond to DMs, not Instagram Story mentions or comments on posts.
• Full booking completion inside the DM thread is not supported — we send a deep link to your booking page, which is where the booking completes.
• No broadcast / outbound DMs — Meta's policy and reputation risk prohibit it.

What's next

Facebook Messenger and WhatsApp Business parity will follow the same concierge pattern. If you use either of those channels heavily, let us know — we prioritise based on owner demand.