A few months ago I decided to buy a new pair of running shoes. I opened ChatGPT and had a 20-message conversation where I told the AI about my gait, my weekly mileage, the trails near my house, and my price range. It asked follow-up questions. It made a recommendation with a detailed argument for the Asics shoe it had in mind. I said I wanted a cheaper option, and it pushed back, because it remembered a knee issue I had brought up in a previous conversation. It walked me through why a minimal-cushion shoe was a bad idea on the trails I described. We argued back and forth for 30 minutes. Once we (the AI and I) had finally reached an agreement on a different, Adidas shoe that satisfied both the AI's concern for my knee and my concern for my wallet, I visited the Adidas website directly and bought it.

That conversation is representative of the new, AI-driven buying journey. I made a decision without reading a review, seeing an ad, or (minus the final step) visiting a website. The entire process happened in a conversation with ChatGPT. It was closer to an argument with a knowledgeable friend than an online shopping experience.

Things look even weirder from a brand's perspective. **The brand had to win an argument it could not see or participate in.** From Adidas' perspective, I was a totally anonymous buyer who searched for their shoe directly, immediately bought it, then left. From the other brands' perspectives (Nike and Asics), they have no idea I was in the market at all. Every year, new research comes out that shows you need 10, 13, 17, 20 touchpoints with a customer before they buy. In this case, the number was zero.

As AI becomes more reliable, buyers will trust its suggestions more. They will ask fewer questions, use less judgment, and just buy what AI suggests. Eventually, human buyers will stop being involved in the buying decision at all. The AI will know what you need and when, and the humans will be the finance department signing the checks.

This is a guide for brands that want to be prepared for the world where AI agents are the buyers. The brands that thrive will be the ones who started preparing while their buyers were still the ones making decisions.

The four stages of AI-influenced buying

It helps to be specific about the steps along the way. I think about a ladder with four rungs.

1. Research assistant. You ask the AI to compare options, summarize reviews, outline tradeoffs. It does the legwork. You make the decision. This is where most buyers are most of the time today.

2. Advisor. You tell the AI your specific situation and constraints, and it comes back with a recommendation. It pushes back when you are about to make a bad call. The running shoes conversation is advisor behavior. The model has a point of view, and it is willing to disagree with you. The first time this happens to you in a category you care about, you start to trust the model differently. You also start asking it questions you would never have typed into a search bar.

3. Decision maker. You tell your AI assistant what you need, and it handles the rest: research, evaluation, selection. You review its recommendation and approve the transaction. The agent did the work. You are ratifying its judgment.

4. Autonomous purchaser. You do not tell the AI anything specific. It already knows what you need, based on everything it knows about you. It buys the thing. You just pay the bill at the end of the month.

Where we are today

Most readers will look at this ladder and place the autonomous purchaser rung somewhere safely in the future. However, buyers in select industries are already delegating their buying decisions to AI.

In developer tooling, engineers are using Claude Code and similar coding agents are routinely letting the model pick their tech stack: which database to use, which hosting provider to use. The agent reads the codebase, evaluates the options, makes a decision, and then tells the engineer to sign up for the service and paste in their credentials. That is rung three, with flickers of rung four, happening every day in a category that earned the trust earliest.

Stripe is already building the infrastructure to support agent-initiated transactions. Their APIs are designed to authorize and settle transactions where the buyer is a software agent acting on behalf of a human. They are doing that because their largest customers are telling them that agent-initiated transactions are going to be a meaningful share of volume, and that share is going to start growing in categories that are not developer tools.

The pattern that spreads fastest is the one where the product is complex, the decision is more about meeting requirements than expressing taste, and the cost of researching every option yourself is high. Most B2B purchasing fits that description. So does a large fraction of higher-ticket (>$100) consumer purchases.

The migration to the agentic-buyer future is already underway, moving from the categories that trust the agent earliest toward the ones that trust it last. Your category is somewhere on that gradient, and it is moving in one direction.

What this means for brand strategy

If AI is becoming the buyer, then the marketer's audience is becoming AI. Brand strategy does not disappear. Positioning, proof, narrative, evidence: all of it still matters. In many ways, persuading an AI is simpler (not easier, necessarily) than persuading a human.

Think of AI as the most intelligent, rational, careful, and well-informed buyer you could imagine. It has already read the entire internet. It has formed a view on you and your competitors. It has opinions about which dimensions matter in your category and which ones are window dressing. It is patient and methodical in a way human buyers are not. It does not respond to clever copy. It responds to evidence that a careful reader would find convincing.

Below are the four principles I have come to think matter most. They came out of a discussion with a top global brand consultancy, and they have been load-bearing in our work ever since, so I thought I'd share them here.

Principle 1: Play the endgame.

Buyer behavior is changing rapidly. Today, most buyers are still on phase 1--using AI to help them do research before making the decision themselves. However, each day, more people recognize how powerful AI is and are willing to delegate more of their decision-making to it. Our recommendation is that brand's should start orienting their strategy towards a phase 4. A brand that decides in 2027 it needs to be the recommended option for AI agents will discover the model already has views shaped by content from years earlier; the brands that win the recommendation in 2028 are the ones who started shaping the belief in 2026.

Principle 2: Treat this as an AI research problem.

AI models are inference systems whose reasoning is sensitive to a lot of things that are not obvious from the output. Many seemingly innocuous things can bias what a model says about a brand. Understanding why an LLM forms a particular belief is closer to AI interpretability research than marketing research. There is a tendency to hope for "one number" to optimize. Unfortunately, it's not that simple.

Principle 3: Focus on understanding AI's beliefs.

What an AI _says_ about your brand can change depending on who is asking, how they phrase the prompt, and the previous context of the conversation. What an AI _believes_ about your brand is stable across any conversation. We've verified this gap experimentally. When designing brand strategy, focus on what AI models believe.

Principle 4: Alignment beats reach.

The instinct most brands have, when they realize AI is shaping how buyers see them, is to publish more. The model has already read everything: your work, your competitors' work, the reviews, the case studies, the analyst notes. A single precise sentence in the right place, closing a perception gap the model has been carrying for a year, will move recommendations more than a hundred new posts of generic content. The agentic-buyer era rewards cohesion over volume.

What changes for the discipline

Brand strategy is changing. The audience is changing; the discipline is maturing around it.

When Stripe's rails for agent-initiated commerce are routine, and a meaningful share of transactions in your category are initiated by an agent on the buyer's behalf, the brands that thrive will be the ones that spent the next few years learning to do brand work with a new audience in mind. They will have a clear view of what the AI audience believes about them today, why, and what would change it. They will have done the work of sharpening their positioning, closing their proof gaps, and earning the recommendation in conversations they were not part of.

There is an opportunity to start the race today.