From conversational to agentic: where the ad goes when the agent decides

Conversational advertising was defined around a person in the chat. The ad becomes a two-way exchange inside an assistant instead of a static message, and a human reads it, asks back, and decides. That assumption is now under pressure. As assistants turn into agents that book, compare, and act on your behalf, the question is fair: is conversational advertising becoming agentic advertising?

Short answer: conversational does not die, it moves from the main stage to the doorway. Conversation becomes the interface. The agent becomes the engine. And advertising always follows the engine, not the surface.

The part of “conversational” that was never about chatting

The label suggested the value was in the back-and-forth. It was not. The value was two things: intent expressed in natural language, and the collapse of discovery, decision, and action into one flow (the shift we set out in what conversational advertising actually is). An agent keeps both. You still say what you want in plain language, and the steps still collapse into one motion. What the agent removes is the human from the middle. It reads, compares, and picks so you do not have to.

So the disruptive part is narrow but real. The “two-way dialogue a person actually reads” is what thins out. The intent-driven, in-the-flow, no-banner core survives intact, and arguably gets stronger.

Two layers, not two eras

It helps to stop treating conversational and agentic as a before and after. They are two layers of the same system.

The front-end stays conversational. People still speak or type. OpenAI’s product chief describes a personal agent you reach “on your mobile, desktop or web. When you’re in the car, you can talk to it.” Source Natural language is the input, and that does not go away.

The decision layer turns agentic. Which hotel gets booked, which tool gets opened, which brand makes the shortlist is now settled by software. That is where the money goes. You can already see it in OpenAI’s super app steering users toward partner services such as Canva and Booking.com. Source The commercial layer did not disappear. It moved from a slot you can see to a choice you cannot audit.

The job of advertising shifts with it: from “persuade the person mid-conversation” to “be the option the agent selects when it acts.” That is a different discipline. It looks less like creative and messaging and more like structured product data, clean machine-readable content, and partner integrations. It is the router fight we described in ASO after search, where the ranking that decides everything is one you cannot see.

Where conversation survives as an ad surface

It does not vanish. It concentrates into a few high-value moments.

• Confirmation and negotiation points. “I found three hotels, this one fits your budget, shall I book it?” That is still a conversational ad, only compressed and high-stakes. The single recommendation at the moment of action is worth more than a feed of them.
• Considered and B2B purchases, where the buyer wants to stay in dialogue rather than delegate the whole decision.
• Voice, where there is no banner at all and the recommendation arrives spoken.

In each case the conversation is the front of an agentic system, not a separate channel.

What this changes for marketers

Three shifts worth planning for now.

The scope of the term widens. “Conversational advertising” still names the door. The work behind it is no longer only “how do I write a good dialogue ad,” it is “how do I become the option an agent picks.” Treat the agentic decision layer as the next chapter of the same discipline, not a separate one.

Measurement breaks and has to be rebuilt. Impressions and click-through assume a human looking. When an agent decides, the metrics that matter are selection rate, inclusion in the shortlist, and attribution of the completed action. This is the same move from impression to interaction we argued in AI agents don’t see your ads, now applied to the decision itself.

Transparency becomes the hard problem. A conversation was something a person could see and judge. An agentic recommendation is not. Who lands on the shortlist, and why, is less auditable than any labeled banner ever was. That gap is where disclosure rules and regulation will land, and it is the part most worth getting ahead of.

My read: conversational does not become agentic. Conversational becomes the human face of an agentic system. For marketers that is the less comfortable world, because the influence retreats into a decision nobody watches. For anyone working on how to be present inside that decision, it is the more interesting one. The brief is no longer just the ad in the chat. It is whether you exist at all when the agent decides.

TL;DR

• Conversational advertising assumed a human in the chat. As assistants become agents that act, that assumption weakens.
• The core of conversational (intent in natural language, a collapsed funnel) survives. The disrupted part is the dialogue a person actually reads.
• Think in two layers: the front-end stays conversational, the decision layer turns agentic, and advertising follows the decision layer. Source
• The job shifts from persuading a person to being the option an agent selects, which rewards structured data, parseable content, and partner integrations.
• Measurement moves to selection rate and completed-action attribution, and transparency of agent choices becomes the central open question.