Set Instagram's instant reply and FAQs so every DM is acknowledged in seconds, keep one sheet of your fit notes, stock, delivery and returns answers for an AI assistant to draft replies from, and give one person phone alerts plus a short inbox check every hour you are open. AI drafts the answer; a person confirms anything about stock or fit.
One caveat first: "within an hour" should mean opening hours. Out of hours, the honest move is an automatic reply that says when a person will pick it up. And the hour is measured to the real answer, not the acknowledgement, which is why the setup below has three layers rather than one clever bot.
What boutique DMs are really asking
Before automating anything, scroll back through two weeks of messages and sort them. In most clothing boutiques the same six kinds of question account for nearly everything:
| Type | Typical message | Who answers | Why |
|---|---|---|---|
| Stock and size | "Do you still have the green midi in a 12?" | AI drafts, person confirms | Stock changes by the hour |
| Fit | "I'm usually a 10 in most shops, what would I be in this?" | AI drafts from fit notes, person checks | Wrong advice means a return |
| Price and details | "How much is the jacket in your last post?" | AI | Facts that rarely change |
| Delivery, returns, holds | "Can you post it by Friday?" "Can you hold it until Saturday?" | AI | Policy you set once |
| Styling | "What would you wear with the linen trousers?" | AI drafts from current stock, person adds warmth | Sells a second item |
| Problems | "The seam split after one wear" | Person only | Tone and judgement matter |
Count how many fall into each row. That tally tells you where the hour is being lost and which layer below will save the most time.
Layer one: acknowledge every message in seconds with Meta's own tools
Meta Business Suite, the free dashboard for your Instagram and Facebook accounts, has inbox automations that cover the first seconds of a conversation. The ones that matter for a boutique:
- Instant reply: sent automatically to someone messaging you for the first time.
- Away message: sent outside the business hours you set.
- Frequently asked questions: tappable questions shown when someone opens a chat, each with a set answer. Put delivery, returns and opening hours here.
- Keyword replies: an automatic answer when a message contains a word you choose, such as "returns".
- Saved replies in the Instagram app: your own templates, inserted by hand with a couple of taps.
An instant reply buys time; it does not answer anything. Keep it short and specific:
Hi, thanks for messaging us. One of us will reply properly within
the hour while the shop is open (10-5.30, Tuesday to Saturday).
If it's about sizing, tell us your usual size and one shop whose
fit you like, and we'll come back with a straight answer.The last line matters. It gets the customer to send the details a fit answer needs, so the first human reply can be the real answer rather than a question.
Keyword replies need testing against real messages, because single words catch more than you meant. A boutique sets "return" to send its returns policy, and the first customer to trigger it wrote, "Will the green midi return in a 12?" She got a paragraph about 14-day refunds and a question left unanswered. Scroll back through a fortnight of DMs, search each keyword you plan to use, and look at every message it would have caught. Phrases such as "return an item" or "send it back" are safer than the bare word, and anything that catches a stock question should be dropped.
Layer two: a fit-and-stock sheet the AI drafts replies from
AI replies are only as good as what they can read. Build one sheet, shared between whoever answers DMs, with a row per style:
- Style name and the words customers use for it ("the green midi", "the one from Tuesday's post").
- Price, fabric and care.
- Fit notes: runs small, true to size or generous; stretch or none; what the person in your photos is wearing and their height.
- Garment measurements per size, at least bust, waist and length.
- Sizes in stock, updated at opening and after any busy spell.
- Two things from current stock that go with it.
One filled-in row, as an illustration of the level of detail that makes AI fit answers usable:
| Column | Entry |
|---|---|
| Style and customer words | Tiered cotton midi, green; "the green midi", "Tuesday's dress" |
| Price, fabric, care | $89; 100% cotton, no stretch; wash at 30 |
| Fit notes | Generous in the waist, fitted across the shoulders. Size down only if narrow-shouldered. Photographed on our 5ft 6 model, wearing a 10 |
| Measurements (bust / waist / length) | 10: 92 / 80 / 118 cm; 12: 97 / 85 / 119 cm; 14: 102 / 90 / 120 cm |
| Sizes in stock (updated 10:05) | 8, 12, 14, 16 |
| Goes with | Cropped denim jacket; tan leather sandals |
The fit note says where the dress is generous and where it is not. "Runs large" on its own would send narrow-waisted, broad-shouldered customers a size that does not close across the back.
Add a policy tab: delivery options and cut-off times, returns window, whether you hold items and for how long, and the one line you use about discounts.
Three ways to put AI on top of the sheet
- A general assistant. Put the sheet in a ChatGPT or Claude project (both offer shared projects on their business plans), and paste each DM in with the prompt below. Cheapest and most controllable, but it means copying and pasting.
- Meta's own agent. Meta Business Agent went global in the WhatsApp Business app in June 2026 and also answers Instagram and Messenger DMs. It replies from your business information and catalogue and decides when to pass a chat to a person. Since 1 August 2026 it has been charged by usage, at $2 per million tokens, which Meta puts at roughly 4-5 cents a message, with no free allowance. Availability varies by account, so check what your Business Suite offers.
- A dedicated DM tool. Automation platforms that connect through Meta's official interface can draft or send replies inside Instagram. Compare them in our tutorial on Instagram DM automation tools for small brands.
A rough sum helps choose between the first two. Illustrative figures: a boutique gets 35 DMs a day, or about 1,050 a month. If Meta's agent handles two-thirds of them and sends around two messages per conversation, that's roughly 1,400 agent messages; at Meta's 4-5 cents each, about $55-$70 a month, rising with every busy restock. A $20-a-month assistant plan with copy-and-paste costs less but needs a person for every reply. Neither is wrong. The agent makes sense once the paste-and-send time is the real bottleneck; test it on a quiet week first and read every conversation it handled.
The drafting prompt for option one:
You help a clothing boutique reply to Instagram DMs.
Use ONLY the fit-and-stock sheet and policy tab attached.
Tone: friendly, brief, like a shop assistant who knows the rail.
British English, no emojis unless the customer used them.
Rules:
- Never confirm an item is in stock. Say which sizes the sheet
shows and add "[STAFF: check rail]" for the person to confirm.
- For fit, compare the customer's usual size with our fit notes
and measurements. If unsure, suggest two sizes and say why.
- Never offer a discount or promise delivery dates not in the policy.
- If the message is a complaint, reply only "[STAFF ONLY]".
- Suggest one item from the sheet that goes with it, if relevant.
Customer message:
[paste DM, with the customer's name removed]An illustrative run. The DM: "hi is the green dress still available in 10?? I'm usually a 10 but a 12 in jeans". The draft that comes back:
Hi! The green midi is showing in 8, 12, 14 and 16 today, so no 10
at the moment [STAFF: check rail]. It's generous in the waist but
fitted across the shoulders, so as a usual 10 I'd try the 12 if
you like room to move, or the 8 if you're narrow-shouldered. Our
model is 5ft 6 and wearing a 10. Want me to hold a 12 for
you until tomorrow? It looks lovely with our cropped denim jacket.What the person answering fixes: the 8 is a poor suggestion for someone who wears a 12 in jeans, so they delete it; they check the rail, find a 10 was returned that morning, and change the first line. The reply goes out 40 minutes after the DM arrived, with the right size in it, which is the whole point of keeping a person on stock questions.
Strip out names and phone numbers before pasting, and use a plan where your chats are not used for model training. The tutorial on automating Instagram DMs without sounding like a bot has more on keeping the tone human.
The stock rule: the AI never says "yes, we have it"
This is the single rule that protects the whole setup. A stock sheet updated at 10am is wrong by lunchtime on a Saturday. If the AI tells a customer the size 12 is there and she drives in to find it sold, you have lost more goodwill than a slow reply ever would.
A realistic version of the mistake: a boutique let its tool send stock answers directly for a week. On the Saturday, three customers were told a linen blazer was available in a 14; the sheet had been updated at opening and the last 14 sold at 11am. Two of them had asked to collect that afternoon. The fix was to take stock confirmation back to a person and add the time of the last update to the sheet, so drafts could say "as of 10am".
So the AI's job on stock questions is to draft everything except the confirmation. The person answering glances at the rail (or the till system), deletes the staff note and sends. That takes thirty seconds and keeps the reply inside the hour. Pair it with a clear hold policy in the sheet, for example "We can hold one item until close of business tomorrow."
Layer three: the rota that makes "within an hour" true
Most boutiques miss the hour because nobody is sure whose job the inbox is. Give it an owner per shift and fixed check points:
| When | Who | Task | Time |
|---|---|---|---|
| Opening | First person in | Update sizes in the sheet; clear overnight DMs | 15-20 min |
| Quarter to each hour | DM owner for the shift | Open the inbox, send drafts, confirm stock | 5-15 min |
| After a new-stock post | Whoever posted | Stay on the inbox for the next hour | Varies |
| Closing | DM owner | Reply to all open chats or say when you will | 10 min |
Turn on phone notifications only for the DM owner, and schedule new-stock posts for times when someone can watch the inbox. A drop on Instagram is the moment DMs spike, and those are the customers most ready to buy.
Reply templates for the questions boutiques get most
Save these as saved replies, and give the AI the same wording as examples so its drafts match.
SIZE AND STOCK
We've got the [style] in [sizes] on the rail right now. It runs
[fit note], so if you're usually a [size] I'd go for the [size].
Want me to hold one until [day]?
FIT
Good question. The [style] has [stretch/no stretch] and measures
[measurement] across the [area] in a [size]. Our [person] is
[height] and wearing the [size]. From what you've said I'd pick
the [size], or the [size] if you like it looser.
DELIVERY
We post [days] with [service]; orders before [time] go the same
day. Delivery is [price], or free over [amount].
COMPLAINT (first reply, always from a person)
I'm sorry, that shouldn't happen. Could you send a photo of the
[problem]? I'll sort a repair, exchange or refund for you today.A two-person boutique handling 35 DMs a day
Take a hypothetical boutique with two staff and around 35 DMs a day, rising to 60 after a new-stock post. Sampling 20 conversations shows the first real reply took about five hours on average, mostly because DMs were answered in the evening.
After setting up the three layers:
- Every message gets the instant reply or an away message immediately.
- About half (delivery, returns, holds, prices) are answered from FAQs, keyword replies or AI drafts sent with a glance.
- About a third (stock and fit) are AI-drafted and confirmed at the hourly check.
- The rest (styling chats and problems) get a personal reply at the next check.
Time spent on DMs drops from roughly two hours each evening to around 75 minutes spread across the day, and a fresh sample shows the first real answer arriving in 20 to 50 minutes during opening hours. Your numbers will differ; the point is to measure them the same way before and after.
When a new-stock post brings 40 DMs in an hour
The hourly check falls apart the evening you post a restock of a style everyone asked about. Plan for the spike instead of reacting to it:
- Update the sheet before you post. Sizes, prices and fit notes for every new style go in first, so the AI's drafts are right from the first message.
- Answer the common question in the post itself. If the caption gives the price, sizes and "runs small, size up", a third of the DMs never get sent.
- Set a temporary keyword reply for the style name that gives price, sizes and the hold policy, and switch it off the next morning.
- Work the inbox in passes. First pass: send every draft that needs no stock check. Second pass: walk the rail once with a list of the sizes people asked about, then send those replies together. Walking the rail once is far quicker than walking it twenty times.
- Keep a waiting list. When a size sells out, the AI can draft a short "we'll message you if it comes back" reply, and the names go on a list you work through when stock arrives.
A spike handled this way often produces more sales than the post itself, because the people who DM about a size are close to buying. The companion tutorial on outfit and styling posts shows how to write captions that answer the questions before they arrive.
Staying inside Instagram's messaging rules
Tools that send messages through Meta's official interface work within a 24-hour window from the customer's last message. Meta's developer policy allows a longer human-agent reply window of seven days for cases a business could not resolve in time, and it is for people only, not bots. In practice: reply the same day, never mass-message people who have not written to you, and use tools that connect through Meta's approved route rather than your password. Our tutorial on what gets accounts banned covers the edge cases.
How to check you are hitting the hour
Once a week, take 20 conversations from opening hours and note two times: when the customer first wrote, and when a person (or an AI answer you approved) actually answered the question. The acknowledgement does not count. Also note any AI draft you had to correct, and why. Three corrections for the same style usually means its fit notes are wrong.
An illustrative extract of one week's log: 10:12 to 10:41 (29 minutes); 11:02 to 11:30 (28); 11:50 to 12:58 (68, a new-stock post went up at 11:45 with nobody watching); 14:20 to 14:37 (17); 16:55 to 17:40 (45, draft corrected: suggested a jacket that had sold out). Across all 20, the middle time was 31 minutes and three conversations went over the hour, all within an hour of a post. The fix for that week isn't faster typing; it's moving the post to a time the DM owner is free.
Keep the log in the same sheet as the fit notes so patterns are easy to spot. And if the DMs sit alongside WhatsApp and Facebook enquiries, bring them into one AI inbox before adding more automation.
What boutique owners ask about AI and DMs
Can I let the AI reply to DMs without checking them?
For delivery times, returns windows and opening hours, yes, once you have watched it answer correctly for a couple of weeks. For stock, fit and anything involving a complaint, keep a person in the loop. Those are the replies where a confident wrong answer costs a sale or a return, and they are exactly the ones where an AI working from a stale sheet is most likely to be wrong.
Does the same system work for Facebook and WhatsApp messages?
Mostly. Meta Business Suite brings Instagram and Facebook messages into one inbox, and Meta's business agent covers WhatsApp too. The fit-and-stock sheet and templates carry over unchanged. The main difference is WhatsApp's own rules on messaging customers outside a conversation, so check those before automating anything there. If you connect a tool through the WhatsApp Business Platform, service replies become chargeable from 1 October 2026 after the first 1,000 a month per number.
What should the reply say when someone asks for a discount?
Decide the policy before the AI sees the question. Most boutiques do best with one consistent line, such as mentioning the loyalty scheme or the next event, rather than improvised offers. Put that line in the sheet, tell the AI never to offer or invent discounts, and route any haggling to the owner.
Further reads
- How Day Spas Can Take Bookings From Instagram DMs With AI — A similar DM setup built around bookings rather than stock.
- How to Turn Instagram Comments Into Leads With Keyword Auto-DMs — Turn comments on new-stock posts into private replies.
- How Clothing Boutiques Use AI to Spot Trends Before Buying Stock — Use DM questions as buying signals for next season.
- When Should an AI Chatbot Hand Over to a Human? — Set clear rules for when the AI passes a chat to you.
- How to Build the FAQ Your AI Chatbot Needs Before Launch — Turn your sheet into a full FAQ any AI tool can use.
- Is It Safe to Put Customer Data Into ChatGPT? — What to strip out before pasting a DM into ChatGPT.
- How to Answer Etsy Customer Messages Faster With AI — Quick replies for the repeat questions, AI drafts for the rest, and a tone check for the hard ones, so Etsy messages get answered well inside 24 hours.
- AI Model Shots for Boutiques: Try-On Images Without a Photoshoot — How AI on-model images work, what tools like FASHN and Photoroom cost, which garments come out wrong, and how to check and label every image honestly.
- How to Write Reply Templates That Keep AI Replies On-Script — Rebuild your canned responses so AI fills them in without drifting: locked lines, marked slots, never-say lists and a clear rule for when to hand over.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Meta Business help pages on Business Suite inbox automations; Meta newsroom announcement of Meta Business Agent (June 2026) and its per-token pricing from 1 August 2026; Meta WhatsApp Business Platform pricing update (1 October 2026); Meta for Developers Messenger Platform and Instagram Messaging API policy.