Marketing is broken. What’s next?
This piece is the first in a three-part series about AI and marketing. It’s speculative, so I’d love to hear what you think.
Marketers, you're right to be worried about AI's impact on your job. You should probably be more worried than you are.
You've heard the take: "AI will automate the execution, but strategy will always need a human." The entry-level jobs go, the senior ones stay. Content writers are at risk, CMOs are fine. You might even believe it, because it's reassuring.
It's not wrong, but it's not right either.
Here's what I actually know, because I'm doing it: I can take ten years of my frameworks (the way I think about ICPs, positioning, messaging architecture, channel strategy), upload them, and build an agentic version of myself that runs a competent marketing operation of ~6 agents with an operator agent. Not perfect. Maybe 80% as good. But 80% at a fraction of the cost and none of the salary, and it ships on Sunday at 2am without complaint. That's possible right now. Not in five years. Now.
So when someone tells you strategy is safe, ask them: safe from what, exactly? Safe from a version of you that runs 24 hours a day and never needs a performance review?
How We Got Here
Marketing has moved through distinct phases:
Push. The Mad Men era. We owned the channels (television, print, radio). We told people what was good and they more or less believed us. Distribution was the moat.
Respond. Inbound, content marketing, SEO. The consumer got the internet and started searching instead of waiting. We stopped shouting and started answering. We got data, saw what worked, and did more of it.
Predict. Programmatic, algorithmic, social signals. We stopped waiting for people to come looking and started anticipating where they'd be before they got there. Reading signals, modeling behavior, showing up automatically and at scale. (More on that later.)
Right now, most of us think we're moving into AI-optimized marketing: AI content, AIO. Making sure your brand shows up when someone asks their assistant a question instead of typing it into Google. If this feels like SEO with a facelift, you're not wrong. Optimize, but for the algorithm's algorithm.
But that’s scratching the surface.
AI doesn't just collapse marketing's execution layers. It collapses marketing as a function itself. Here's why.
Every tactic we've ever used (ads, content, SEO, social, influencer, email) is a variation on "reach the human before they make up their mind." But what if humans aren't the ones making up their minds?
Agentic Decision-Making
We're moving into territory where AI makes buying decisions for us. It's already moved from unlocking awareness to enabling consideration, and it's coming for actual purchases.
When someone asks an assistant "what's the best CRM for a 50-person company" and gets three recommendations, the entire funnel collapses. The awareness campaign, the comparison blog post, the retargeting ad, the nurture sequence: gone. An agent made the shortlist. If you're not on it, your marketing might as well not exist.
We're seeing this today in early retail data. The infrastructure is being built now: Amazon's Rufus AI, PayPal's Agentic Toolkit, Visa's Intelligent Commerce.
When agent-mediated purchasing expands, agents won't respond to marketing the way humans do. They don't have feelings about your brand. They don't remember the clever campaign. They don't care about cool.
Agents evaluate on structured signals: performance data, verified reviews, compatibility, reliability metrics, ecosystem integration. You don't win by making people feel something. You win by being the option the system trusts. That's the shift from a persuasion economy to a verification economy.
B2B is not safe. B2B purchasing has historically been protected by relationship complexity, long sales cycles, and organizational politics. That protection is eroding. An exec who has suspected for years that their martech stack underperforms, that their agency isn't delivering, that the vendor they've been renewing out of inertia could be replaced: now they have something that confirms that suspicion, evaluates alternatives, models switching costs, and produces a shortlist before the QBR happens. The relationships that protected B2B incumbents were always partly a substitute for better information. AI is the better information.
On timing: today's AI systems are impressive on the surface and often unreliable underneath. Anyone working closely with these tools knows they produce plausible-sounding answers that fall apart under scrutiny. Full agent-mediated purchasing is 2 to 3 years out for most categories. But that gap is closing. The bet here is on trajectory, not current state.
For the decisions that are still human - and many are, for now - there's a different problem for marketers.
AI compresses marketing’s daily function. AI can create infinite content, in infinite variations, for infinite individuals. This content isn’t all high-quality, but let’s be honest: most businesses don’t have elite strategy, clear differentiation, or even good marketing. Average AI is an improvement. And AI’s quality problems are (I think) temporary. You can learn how to pressure-test it today, and it will learn how to pressure-test itself tomorrow.
And that volume is changing human behavior. People aren’t just overwhelmed, they’re becoming immune. (Ironically, some solve this overwhelm with their AI assistants. It’s driving and solving the problem for a whole cohort of buyers.) We see this most clearly in email marketing: how many AI outbound requests do you get every day versus a year ago? And what has that done to the time you spend in your inbox? This channel has been gutted, and anyone who tells you different is trying to sell you something using email.
Most content about marketing in the age of AI is either optimistic ("AI is you but better, embrace it") or tactical ("here's how to adopt AI"). But these three shifts mean one thing…AI erodes the value marketing used to have.
And I’m left asking one question:
If humans outsource more decisions, what role does marketing really serve?
Distribution to Trust
Distribution architecture is dead as a competitive moat. Trust architecture replaces it.
Distribution architecture is the old game: game channels and reach, and spend enough to stay top of mind. It worked when attention was capturable and people made decisions based on familiarity.
Trust architecture is the deliberate construction of the signals, relationships, systems, and proof that make your brand the one someone (or something) turns to when it matters. Not about being everywhere. About being believed somewhere.
We have early, human versions of trust architecture today.
Yelp, Google, and G2 Crowd were built to verify trust at scale, but what they captured was sentiment at a moment in time. A snapshot of how someone felt after a transaction. Backward-looking, easy to game, disconnected from durable performance. They aggregated human opinion and called it trust. And they're unsatisfying even to humans: how often have you bought something based on reviews, only to be disappointed?
Agents need something different.
Not how someone felt, but how something performed: durably, repeatedly, measurably. Return rates, renewal rates, churn, support ticket volume, time-to-value. These metrics actually predict whether a product does what it claims. They've historically been private because they're competitively sensitive and because no one was asking for them in a structured way. But agentic buying changes that. If agents prioritize brands that can demonstrate performance at the structural level (not just reviews but actual outcome data) then brands face a new incentive: disclose or be deprioritized. That's a market mechanism, not a cultural shift. We already see the early version on Amazon, with return rate signals built into product listings.
The companies building genuine performance records and machine-readable proof today are making a bet that will pay off as that infrastructure matures. The companies optimizing for today's AI visibility through content are betting the current surface stays relevant. Given the pace of change, that seems like the riskier position.
What This Means for Marketers
Marketers, I want you to be able to do the job you love. I want that to still exist for you. I want it to exist for me.
But your options are going to change. The middle layer (execution, optimization, strategy built on frameworks that can be uploaded and replicated) is compressing. Not eventually. Now. But if you love doing that kind of work, you'll be looking for companies whose value system aligns with paying a human to do it, rather than bringing in an AI architect.
What remains is harder to name and harder to hire for: judgment about what's worth building, relationships that can't be automated, the conviction to stand for something specific over time and pay the real cost of that commitment. Whether that still gets called "marketing" or dissolves into something else, I genuinely don't know.
There’s some truth that stategists will be more insulated. Those who work as bridge between operations, growth, customer, product, company, and competitors will likely hold value longer than most, particularly in legacy B2B.
The people who navigate this best will be asking the hard questions now, not after the transition has already happened.
Only you can answer what you want your future to look like. Luckily, we're human. We get that chance.
Next up, my take on trust architecture. And lastly, what the optimization loop is doing to human judgement and why that matters for brands.
News I’m Watching
AI Marketing Skills Directory: Train your AI (and agents) to operate like a specialist across marketing disciplines.
Trust Signals for eComm: Worth paying attention to what the agents will pay attention to.
