Key Takeaways
- Conversational commerce emerged around 2015 as chat-based buying; agentic commerce has fundamentally different premises.
- The biggest difference is subject: conversational keeps the human deciding; agentic puts the agent in charge of execution.
- For merchants, integration targets shift from messenger platforms to the MCP, UCP, and A2A protocol layer.
Chatbots Were Conversation — Agents Are Execution
"Conversational commerce" has been in the ecommerce lexicon for about ten years. Chat support on WhatsApp, Messenger, or LINE; chatbot order intake; voice-assistant shopping — all of these fit the category. So is the newly trending "agentic commerce" just the next version?
No. They look similar on the surface, but the premises, the design points, and the KPIs are all different. This article clarifies the differences and shows how to carry conversational commerce experience into agentic commerce. For protocol background, see protocol comparison; for consumer behavior, see agentic shopping.
Different Starting Points — 2015 vs 2024
Conversational commerce was coined by Uber's Chris Messina around 2015 and took off when Facebook opened the Messenger Platform. The premise at the time: getting mobile users to open apps is hard, so meet them where they already are — messengers. The technical toolkit was chatbots, basic NLU, and template responses.
Agentic commerce rose sharply from late 2024 onward. The starting point was the release of MCP and the arrival of general-purpose agents like ChatGPT and Claude. The technical toolkit is production-grade LLMs, agent frameworks, and the MCP/A2A/UCP/AP2 protocol stack. The quality of the tooling is fundamentally different, and that's what separates the experiences.
Different Subject — Human vs Agent
The most fundamental difference is "who decides and who buys?"
In conversational commerce, the subject is always the human. The user chats with the bot, sees candidates, checks prices, and ultimately says "I'll buy this." The bot proposes and informs; the human presses the purchase button.
In agentic commerce, the subject is the agent. The user says "red running shoes under $200" and the agent handles narrowing, comparing, final selection, and payment. The human may review or confirm but the per-transaction decision authority sits with the agent.
This changes what merchants should design around. In conversational commerce, the priorities were bot tone, how to present options, and where to fix drop-off points. In agentic commerce, the priorities are product data quality for being picked by agents, protocol support, and reputation.
Different Interfaces
Conversational commerce fought its battles on messenger platforms: Facebook Messenger, WhatsApp, LINE, KakaoTalk. Merchants built bots for each, using Publisher APIs and per-platform bot frameworks.
Agentic commerce fights on the protocol layer, not a specific UI. General-purpose agents — ChatGPT, Claude, Gemini — reach merchants through UCP and MCP. Which agent the user uses is up to them; merchants support one protocol and don't worry about which agent it is. In that sense, agentic commerce is more platform-independent than conversational commerce ever was.
Different Metrics
KPI design changes too. Conversational commerce centered on bot sessions, message counts, drop-off rates, and chat-attributed conversion rates. Tracking "how many messages did the user exchange with the bot" was standard.
In agentic commerce, those metrics lose meaning. How many API calls the agent makes behind the scenes doesn't relate to UX. What matters instead: agent-attributed revenue, per-agent recommendation rates, AP2 Mandate execution success rates. As of 2026, analytics for these are still early, and merchants often build pieces themselves.
Conversational Commerce Assets Aren't Wasted
That said, the assets you built in conversational commerce don't go to waste. Several areas transfer directly.
Structured product data has the same requirements for both. Catalogs polished for messenger display transfer directly into structured data for agentic commerce. Return and support automation built for chatbots slots almost unchanged into AI-agent automation. The mental model of intent understanding — the sense you develop for "what is the user actually trying to say" through NLU work — connects directly to prompt engineering.
Merchants who took conversational commerce seriously typically have a stronger foundation for agentic commerce than those who skipped it.
Replacement or Coexistence?
Short-term, they coexist. Messenger-based customer service and ordering won't disappear just because agentic commerce goes mainstream. Especially in Asia, where LINE and KakaoTalk are culturally embedded, agentic commerce layers on top of conversational commerce rather than replacing it.
Long-term, the boundary blurs. If LINE and WhatsApp embed agent capabilities, the "is it conversational or agentic?" distinction loses meaning. LINE is reportedly evaluating MCP support, and WhatsApp is progressively integrating Meta's AI assistant — the two categories look likely to merge clearly around 2027.
Conclusion — Treat Them as Different Games
Just because the names sound similar doesn't mean conversational commerce and agentic commerce are the same game. Premise, subject, interface, metrics — all different. Many assets transfer, but the strategies should be drawn up independently.
The practical move for merchants is to continue running existing conversational commerce while standing up a separate parallel track for agentic commerce. Start with protocol comparison and agentic commerce platforms compared, and treat agentic-commerce-specific design questions as their own independent thread.
