A customer messages a small shop about a delivery, gets three product cards back, taps one, pays, and receives a receipt — all without leaving the thread and without installing a thing. We watched a merchant do this on a phone that had no other commerce app on it. The whole transaction lived inside a conversation that had started as 'hi, are you open?'.

It's tempting to call this a stripped-down store, but that undersells it. The chat already has the two things every commerce product spends years chasing: the customer is reachable, and they trust the number on the other end. A catalog, labels and in-thread payments are just plumbing on top of that. The storefront isn't coming. It's already here.

What the chat is genuinely good at

Look at what a business account actually gets. A catalog with photos and prices, labels to sort customers, quick replies, away messages, and a payment request that arrives as a normal message. For a shop selling a handful of things to people who already know it, that's not a compromise — it's less friction than a website. The buyer doesn't create an account, doesn't remember a password, doesn't wait for a page to load on a bad connection. They just answer a message.

The reach side is the part people underrate. A message arrives where people already are, on a device they keep in their pocket, with an open rate no email team would dare promise. And because the thread is tied to a phone number and a real conversation, the default posture is trust rather than suspicion. We've seen a single reply to a question convert faster than a landing page the same shop had paid to build.

That's the honest case for conversational commerce: high-intent, low-friction, personal. It works brilliantly right up until the service stops being simple.

The honest tradeoff

Here's where the wheels come off, step by step.

StepWhat happens in the chatWhat the catalog carriesWhere it breaks
First contactA customer replies to an ad or scans a QRNothing yet — it's a greetingNo browseable surface, only a thread
BrowsingThe shop sends a few product cardsImages, prices, short descriptionsLong lists and filters turn into endless scrolling
ConfigurationThe buyer asks about size, date, add-onsFixed variants at bestMulti-step choices need a real form
CheckoutAn invoice or payment link lands in the threadA cart totalDisputes, refunds and receipts live elsewhere
After the saleLabels and a follow-up templateA segment tagAnything past 24 hours needs an approved template
Account historyThe chat itself is the recordNothing durableNo login, no saved preferences, no orders view

Now the uncomfortable part. The reason chat feels effortless is also the reason it can't hold much: a thread is linear, and a real service isn't. The moment a booking needs a date picker, a comparison table, a saved address and an order history, you're stuffing a form into a conversation and hoping nobody gets lost. We've tried to push a configuration flow into messages and watched people answer question four while forgetting question one. That's not a UX bug you polish away; it's the shape of the medium.

Then there are the rules, and they're not suggestions. Outside the 24-hour window a customer opened, you can't just send a message — you need a pre-approved template, and templates get reviewed and rejected. Getting onto the Business API at all usually means going through a provider, verifying a number, and accepting per-conversation costs. For a small merchant this is a real gate, not a checkbox. And every extra automated message flirts with being the spam that makes people mute the whole channel. Reach here is borrowed, and the lender sets the terms.

Let the chat do the opening, not the whole job

So we'd stop asking the conversation to be the product. Let it be the front desk: the place trust starts, the place a question gets answered, the place a reminder lands. For anything with depth — configuring, comparing, managing an account, tracking an order — hand off to a lightweight app that opens in one tap and gives the buyer a real interface. The thread stays the relationship; the app does the work.

The practical test is simple. If your flow fits in a back-and-forth of five messages, keep it in chat. If it needs state, a form, or a history view, it doesn't belong there, however much you'd like one surface to rule them all. Build the deep part so it isn't welded to a single messenger — an entry point can change, a template policy can tighten, a region can restrict what's allowed. That's the boring reason a portable service layer, of the kind Cross Mini App targets, is worth having: not to replace the chat, but so the chat never has to pretend it's a store.

What you keep in-house is the judgment about which step goes where. Get that split right and the conversation feels personal instead of pushy. Get it wrong and you've built a call centre that never sleeps and nobody wants to talk to.

Sources