The Marketing Department Is About To Disappear Because AI Is Now Selling To AI

This voice experience is generated by AI. Learn more.This voice experience is generated by AI. Learn more.Richard Freisberg, CTO, All My Sons. Has built, scaled and exited multiple companies. Writes on tech trends shaping how businesses operate.

gettyWalk into any outdoor retailer’s website looking for a tent, and you’ll see the same formula: aluminum poles framed against a mountain ridge at golden hour, copy promising you’ll be “built for adventure” or ranking itself “#1” without saying compared to what. None of that tells you the tent’s square footage, its waterproof rating or whether it survives a real storm. It doesn’t need to. It’s not selling to your judgment. It’s selling to your nervous system.

That’s what marketing has always done well: engineer emotion, trust and desire in a brain that doesn’t always decide rationally. The whole discipline—brand strategy, creative direction, copywriting and campaign design—exists because the buyer can be moved by a feeling.

I don’t think that buyer stays human much longer. That changes marketing more than most executives are prepared for.

This isn’t speculative. It’s infrastructure that’s already live. Google unveiled its Universal Commerce Protocol at NRF 2026 with Shopify, Etsy, Wayfair and Target, covering the full purchase journey. Visa has moved its Trusted Agent Protocol beyond the concept stage, working with more than 100 partners, and Mastercard completed its first live agent-initiated transactions across Asia-Pacific. American Express announced purchase protection built for registered AI-agent purchases. The card networks aren’t simply experimenting anymore; they’re underwriting a world where software swipes the card.

A 2026 IBM study found 45% of consumers already use AI for part of the buying journey, and McKinsey puts the agentic commerce opportunity at $3 to $5 trillion by 2030. Morgan Stanley estimates AI-driven agentic shoppers could account for 10% to 20% of U.S. e-commerce spending by 2030, as much as $385 billion, with 23% of Americans already having bought something using AI in the past month. The question isn’t whether agents will shop on our behalf. It’s how fast the transition finishes.

Meta made this concrete on September 8, 2026, when it launched Muse, a personal AI agent available on iOS, Android and the web. Muse can book things, fill out forms and check out with a connected payment method. This is precisely the personal-agent-as-buyer scenario this piece is built around, now shipping to consumers.

The tell is what marketing is doing right now. A discipline called generative engine optimization (GEO) has emerged overnight, built around a single anxious question: How do we get chosen by an AI system instead of a human? WPP’s VML puts it to brands directly: “being seen is not enough — you must be chosen.” Miss that trust test with an AI agent, the agency warns, and a brand disappears from the conversation.

That instinct is correct. The execution is not. GEO still treats it as a persuasion problem, a different flavor of charm to win over a purchasing agent, the way a waterfall backdrop wins over a hiker. But an agent evaluating a tent isn’t feeling anything. It’s parsing structured data: weight, dimensions, waterproof rating, price and delivery window, matched against constraints a human gave it once.

As Shopify’s own commerce research puts it, AI agents rely on structured, real-time data to make purchase decisions, not editorial content. Reviews fit the same pattern: An agent cross-checks specs against ratings and sentiment before buying, and as more reviews get written by agents themselves, that becomes another dataset to parse, not a testimonial meant to move a person. A gorgeous campaign does nothing for a system that only reads a feed.

Even granting that agents need data “optimized” to be favorably selected, ask who does that. It’s not a human job. It’s a rules-based, measurable, constantly-shifting task: watching competitor pricing, adjusting attributes and testing which framings an agent responds to, faster than any team of people could. That’s precisely the work AI is best suited to automate, not the work that justifies a marketing department.

So the honest endpoint isn’t “marketing survives by becoming more technical.” One AI system optimizes how a product presents itself, and another evaluates and buys it. Persuasion still happens as a mechanism. It just no longer runs through a human on either side. AI is persuading AI. Humans are only in the room where the constraints get set, once, at the very top: a budget, a brand line that can’t be crossed or a short list of non-negotiables an agent operates within from then on.

That is not a smaller marketing department. It’s the absence of one.

A handful of people will stay close to the brand—not marketers, but two or three from product and legal—setting positioning guardrails, pricing floors and boundaries an AI system can’t violate when negotiating with another AI on a customer’s behalf. That’s governance, not marketing. It’s closer to a policy file than an org chart. The rest of what we call marketing, including campaigns, copywriting and creative direction built to move a human emotionally, has no audience left.

If you run a company, the mistake right now is optimizing your marketing team to reach AI agents better, as if that buys time. It doesn’t. It’s optimizing a department for a function about to be automated out from under it. The better question almost nobody in a marketing seat wants to ask is, “When the whole transaction—discovery, comparison, negotiation, purchase—happens machine-to-machine, what does a marketing department actually do, and for how much longer does it need to exist at all?”

I don’t think that’s a five-year-out question. Based on what’s already live in payment rails, commerce protocols and agent adoption curves, I think it’s closer to two or three years. Companies that start restructuring around that answer now, rather than defending the org chart they already have, won’t be caught flat-footed when the last human buyer logs off and lets an agent take over.

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?