Conversational Commerce: When a Support Chat Becomes a Checkout

Somewhere in your store's chat log this week, a shopper asked whether a jacket runs small, got an answer, and bought it. No campaign, no discount code, no landing page. A question, an answer, an order.

That's conversational commerce. Not the buzzword version, the ordinary, unglamorous version that already happens in most stores and mostly goes unmeasured because chat sits in a support dashboard rather than a revenue report.

The interesting question isn't whether conversations sell. It's why so few stores treat them as if they do.

What conversational commerce actually means

The term was coined by Chris Messina in 2015 to describe the collision of messaging apps and shopping. A decade on, the definition has settled into something more practical.

Conversational commerce is any commercial interaction that happens through dialogue rather than navigation. Instead of the shopper hunting through a size chart, a shipping page and a returns policy, they ask, and the store answers.

Three things distinguish it from ordinary customer service:

• It happens before the purchase as often as after it.
• It's two-way, the store can ask clarifying questions, not just respond.
• It can complete an action, not just supply information: check stock, apply an exchange, place or amend an order.

That third point is where most implementations fall down, and it's the difference between a chat that deflects and a chat that sells.

How it evolved: from SMS blasts to AI agents

The channel history matters, because each phase left behind a habit that still shapes how stores use chat.

Phase 1. SMS and email broadcast (2000s). One-way, promotional, measured by open rate. Conversation wasn't really possible, so nobody designed for it.

Phase 2. Live chat widgets (2010s). Two-way at last, but gated by staffing. Available 9 to 5, queued, and treated as a support cost. Most stores measured tickets closed, not revenue influenced.

Phase 3. Messaging apps and social (late 2010s). WhatsApp, Instagram DMs and Messenger moved shopping conversations to where people already were. Huge reach, but fragmented across inboxes and hard to staff.

Phase 4. Rule-based chatbots. Cheap coverage, poor experience. Decision trees broke on anything unexpected, and shoppers learned to type "agent" immediately.

Phase 5. AI agents (now). Systems that read your actual catalogue, order data and policies, understand a question in the shopper's own words, and answer or act. Always on, multilingual, and critically, capable of handling the long tail of questions a decision tree never anticipated.

The channels that matter

Not every channel deserves your effort. The right mix depends on where your buyers already are and how much you can realistically staff.

Most Shopify stores get the majority of the value from on-site chat plus one messaging channel. Adding a fifth inbox you can't answer is worse than having four.

What makes a conversation convert rather than just deflect

Deflection means the shopper stopped asking. Conversion means they bought. Those are not the same outcome, and optimising for the first can quietly damage the second.

Conversations that convert tend to share four traits.

1. Speed measured in seconds, not hours. Pre-purchase intent decays fast. An answer that arrives after the shopper has closed the tab is a support log entry, not a sale.

2. Access to real store data. "Our standard delivery is 2 to 5 days" is a policy. "To your postcode, this arrives Thursday" is an answer. The second requires the conversation to see order, inventory and shipping data, not just a help-centre article.

3. Willingness to recommend. A genuinely useful assistant says "that one runs small, the next size up is in stock" or "this works with the model you bought last year." Guided selling is the part most stores never switch on.

4. A visible, fast route to a human. Trust in the whole channel depends on the exit. If a shopper believes they can reach a person in one step, they're far more willing to start the conversation in the first place.

Illustrative example (not a real customer): Picture a Shopify homeware store where roughly a third of chat messages arrive between 8 p.m. and 7 a.m. Under a staffed 9 to 5 model, all of those become next-day email replies. Under an always-on model, they become answers while the shopper is still on the page. The size of the difference depends entirely on your traffic pattern and margins, which is exactly why it's worth measuring on your own store rather than assuming an industry benchmark applies.

Where an AI agent fits

Conversational commerce only works at scale if someone can answer. For most Shopify stores, that "someone" now needs to cover nights, weekends and several languages, which is a hiring problem before it's a technology one.

AskZoye is an AI customer support agent built for Shopify that answers shoppers in around three seconds, 24/7, in 50+ languages, using your live store data for order tracking, returns, shipping and product questions. It also recommends products and helps recover sales that would otherwise be lost overnight, and it goes live in under 60 seconds with no code.

The framing that matters: treat it as a sales channel with a support function, not a support tool with a sales side effect.

The honest limits

Conversation is not the right interface for everything, and pretending otherwise is how stores end up with irritated customers.

An AI agent should not handle an emotionally charged complaint, a request for an exception to your own policy, or a rare edge case where the correct answer requires judgement about a specific customer relationship. Those need a person, and the handoff should be fast and obvious rather than buried behind three attempts at deflection.

There's also a discovery limit. Chat is excellent at resolving a specific doubt. It's a poor replacement for good browsing, clear photography and honest product copy. If shoppers need a conversation to understand what you sell, the problem is upstream on the product page.

The bottom line

Conversational commerce isn't a new channel to bolt on. It's a recognition that some purchases need a question answered before they'll happen, and that the store which answers first usually gets the order.

Start narrow. Pick the channel your buyers already use, make sure answers arrive in seconds rather than hours, connect the conversation to real order and inventory data, and measure it against revenue instead of tickets closed.

Then be honest about the boundary. Automate the repeatable, hand the difficult conversations to a person quickly, and keep the exit visible.