A fashion store’s inbox has one question in it more than any other, and it is not about
delivery. It is “will this fit me”.

That question is asked in a hundred forms, it is asked by a shopper who is one tap from
buying, and it is answerable from data the store already has. Which makes it the most
valuable thing to automate in this business, and the one most stores answer by hand or not
at all.

The sizing question

Shoppers do not ask for measurements. They describe themselves and expect you to
translate.

  • “im 1.65m 55kg which size”
  • “usually wear M in uniqlo, same?”
  • “is this true to size or runs small”
  • “does it stretch”
  • “i have broad shoulders, will the S fit”

Answering these needs three things: the product’s own size chart, the fit note for that
garment, and the fabric composition. All three are usually already on the product page,
and all three are usually ignored by the chat tool, which answers from the marketing
description instead.

A workflow that reads the actual size chart for the actual product, applies your fit
guidance, and gives a recommendation with the reason is better than most human replies,
because it does not guess. “For 1.65m and 55kg the M fits with room. This style runs
slightly small through the shoulders, so if you are between sizes take the L. It is 95%%
cotton with 5%% elastane, so there is a little stretch.”

What it must not do is invent a chart. If the product has no measurements recorded, the
honest answer is to say so and offer a person. A confident wrong size recommendation
produces a return, which costs you the margin and the shipping both ways. See
AI hallucination.

Stock by variant

“Got the black in S” is a live inventory read, not a document lookup. This is the
clearest example of why a tool trained on your website text cannot do the job: the page
says the product exists, it does not say the black small is gone.

Two follow-ups to handle in the same workflow. If it is out of stock, offer a
back-in-stock alert rather than ending the conversation. And if a
near variant is available, offer it, because a shopper who wanted black small will often
take navy small.

The delivery-date question, which is a sale

“Will it arrive before Saturday” is not a logistics question. It is the last objection
before purchase, and it has a deadline attached: a wedding, a trip, a birthday.

Answering it properly means a shipping estimate for their actual address, not your
generic policy. That is a rate lookup. Get it right and a share of those shoppers buy
immediately. Answer with “3 to 5 working days” and they go and check a competitor who
promises a date.

Instagram is where the sale starts

For fashion, the direct message is the shop floor. A shopper sees a post, asks about
it, and either buys in that conversation or does not buy at all.

Three workflows matter here specifically.

Comment to direct message. Someone commenting “price” or “size S?” on a
post is a customer. An automatic reply that opens a direct message captures them in the
seconds they are still looking. AeroChat replies to comments on Instagram and Messenger posts as well as handling the direct messages, so the comment becomes a conversation. See conversational selling through Instagram DMs.

Story reply handling. Replies to stories are high intent and almost
always about one specific item. The workflow needs to know which item the story was
about.

Selling in the thread. A shopper who has asked two questions and
decided should not be sent to the website. A payment link in
the conversation finishes it. The documented average cart abandonment rate across published
studies was 70.22%% as of September 2025
(Baymard Institute), and the
checkout you avoid is the one you cannot lose them in.

Returns and exchanges

Fashion has the highest return rate in retail, and the return conversation is where
goodwill is kept or lost.

What automation should handle: the policy, whether this specific order is still inside
the window, starting the return, and issuing the label or instructions. That is a lookup
and an action, and it is identical every time.

What it should not handle: a return outside the window, a damaged item, a wrong item
sent, or anything where the customer is annoyed. Those need a person with discretion, and
the refund ceiling that person works within should be a limit the system enforces rather
than a guideline. In AeroChat the merchant sets the refund and discount limits and the
system applies them, passing anything larger to the team with the order details
attached.

Exchanges are the better outcome and the harder workflow, because they involve checking
stock in the replacement size. Worth building, because an exchange keeps the revenue.

Channels and platform

Instagram and WhatsApp carry the conversations, web chat catches the shopper mid
checkout. The store itself sits on Shopify or WordPress, and the useful distinction is
that those are platforms rather than channels: the store is where your product and order
data live, and the channels are where customers talk to you. The automation has to span
both, which is the whole point.

Drops and peak weeks

Fashion volume is not flat, and the weeks that break a store’s service are predictable:
a collection drop, a sale, and the fortnight before a major gifting occasion.

Two things happen in those weeks. Message volume multiplies, and the mix shifts towards
exactly the questions automation handles best: is this still available, what size, will it
arrive in time. So peak is when automation is worth most, and it is also when a badly
configured setup does the most damage.

Three preparations are worth making before a drop rather than during it. Make sure
variant-level stock is accurate, because a sold-out size answered as available produces a
cancellation and a refund. Set the delivery cut-off date for the occasion and have the
system state it plainly. And decide in advance what happens when stock runs out mid
conversation, because during a drop that will happen while someone is choosing.

One figure to watch through a peak: the share of conversations that ended without an
answer. That is your real capacity ceiling, and it is invisible in revenue reporting
because the orders you did take look fine.

Where to start

  1. Sizing answers from the real size chart, with a handover when there is no chart.
  2. Variant-level stock, with back-in-stock alerts and near-variant offers.
  3. Delivery date estimates for the shopper’s own address.
  4. Instagram comment to direct message, then selling in the thread.
  5. Returns inside the window automated, everything else to a person.

Read next

All industries

Start for free

AeroChat runs customer service and sales workflows you describe in plain English. There is a free plan, so there is nothing to decide today.

Start for free on the Shopify App Store

Not on Shopify? Create an AeroChat account here.