Two-thirds of Gen Zers and more than half of Millennials are now using LLMs to research products before they buy. That's not a future trend. That's happening right now.

But here's what most CMOs are missing: when someone uses ChatGPT or Claude to research your category, you have no idea if your brand showed up in that conversation. You don't know if the information was accurate. You don't know if you were even considered.

And increasingly, it's not just humans using AI to research. It's AI agents acting on behalf of humans, making decisions without the person ever visiting your site, reading your content, or engaging with your brand directly.

We're entering a world where three distinct types of AI-mediated relationships are emerging simultaneously: brand agents (your AI engaging with humans), consumer agents (their AI acting on their behalf), and full AI intermediation (AI-to-AI autonomous interaction).

The strategic question every CMO needs to answer isn't whether to adopt agentic AI. It's which layer you're competing in, and whether your brand even shows up when the decision gets made without you in the room.

The behavior shift is already underway. ChatGPT searches OpenTable, picks restaurants based on preferences, and completes the booking. Claude can navigate interfaces, fill forms, and execute purchases autonomously.

We're seeing three distinct relationship models emerge:

Brand agents are what most companies think of first. Capital One's Auto Navigator checks inventory, schedules test drives, estimates trade-ins, and explains financing. It's controlled, on-brand, and designed to guide customers through a journey you still own.

Consumer agents flip the script. They work for the buyer, not you. They compare your offering against competitors using criteria you don't set. They optimize for factors you might not even track. The person isn't doing the research. Their agent is doing it for them.

Full AI intermediation is where it gets interesting. No human in the loop. Your brand agent negotiating directly with their consumer agent. Autonomous systems executing purchases based on rules established days or weeks earlier.

Here's the problem: most brand data isn't ready for this. When Pernod Ricard audited how major LLMs represented their brands, they found gaps and errors everywhere. Ballantine's Scotch, a mass-market whiskey, was being described as a prestige product. If the model has your positioning wrong, every recommendation it makes will be wrong too.

This changes how deals are won and lost.

When consumer agents handle research, your analytics tell you one story while reality tells another. You see the traffic. You don't see the human who never showed up because their agent already ruled you out.

When AI intermediation becomes standard, brand preference becomes a data structure problem. Your narrative, your differentiation, the emotional connection you've built... it only matters if the agent can interpret it and weigh it appropriately. If your value proposition doesn't translate into structured decision criteria, you're invisible.

The governance challenge is harder than it looks. When your systems interact with external agents, you're depending on someone else's guardrails. You don't control their training data. You don't see their decision logic. You're trusting that their agent has the judgment and constraints to interact appropriately with yours.

And if something goes wrong, if data gets shared that shouldn't be, if a transaction executes incorrectly, the accountability is murky. That's not a technical problem. That's a business risk that most legal and compliance teams haven't even started thinking about.

Connect your email and calendar, and Attio instantly builds your CRM. Every contact, every company, every conversation, all organized in one place.

Then Ask Attio anything:

No more digging and no more data entry. Just answers.

Start your free trial →

The instinct is to build a chatbot, label it an agent, and move on. It handles support tickets and answers product questions. That's helpful, but it's not agentic. Real agents make decisions, execute actions, and operate with judgment inside boundaries you define.

Most teams skip the hard work: deciding what the agent is allowed to do, what data it can access, how it should handle edge cases, and when it needs to escalate to a human. Without that foundation, you're just automating conversations, not delegating decisions.

The second mistake is treating this like an SEO refresh. You can't content-market your way into an LLM's recommendation. The models aren't crawling and ranking pages. They're synthesizing knowledge and applying it contextually. If your brand information is scattered, inconsistent, or buried in unstructured content, the model won't know how to represent you accurately.

The biggest gap is thinking this is a marketing initiative when it's actually a cross-functional data and systems challenge. If your product catalog is messy, if your positioning varies by channel, if your customer experience has gaps, the agent will expose all of it. You can't automate around organizational dysfunction.

Start by understanding where you actually show up in AI-mediated research. Most marketing teams are still optimizing for Google while their customers are asking ChatGPT, Claude, and Perplexity for recommendations.

Run a systematic audit:

This forces a mindset shift from "are we ranking?" to "are we part of the answer?" That's the new battleground.

Next, get clear on your layer strategy. Are you building agents that guide customers through your owned experience? Are you optimizing your brand presence so consumer agents recommend you accurately? Or are you preparing your systems for direct agent-to-agent interaction?

Each layer requires different capabilities. Brand agents need strong decision frameworks and clear escalation paths. Influencing consumer agents requires structured, authoritative content that models can parse and trust. AI-to-AI intermediation demands robust APIs, data governance, and real-time monitoring.

You don't need to execute all three immediately. But you need to know which matters most for your business and what infrastructure that requires.

Governance has to be built in from the start, not bolted on later. Think about what data your agents can access and share. Define clear boundaries around what decisions they can make autonomously. Build monitoring so you can see when things go sideways. And recognize that governance isn't a one-time policy. It's an ongoing discipline that evolves as your agents get more capable and the ecosystem gets more complex.

Finally, rethink how you build authority. The old SEO playbook was about ranking for keywords. The new game is about being synthesized correctly. That means your content can't just answer questions. It needs to tell a story that connects your brand to the solution in a way that's clear, consistent, and contextually relevant across every surface where information about you exists.

Most companies are still in the early stages. They're using AI to generate content or automate simple tasks. Agent-to-agent negotiation isn't widespread yet. But the infrastructure is being built right now. The models are improving. The agents are getting more autonomous.

The companies that figure out which layer they're competing in, and build the systems and governance to compete there effectively, will maintain relationships with customers even as those relationships get mediated by AI.

The ones that don't will keep seeing traffic that doesn't convert, leads that don't close, and deals lost to competitors they never knew they were up against.

The question isn't whether this shift is coming. It's whether you'll be visible when it arrives.

Stay Hungry,