The impact of AI on growth and marketing teams is often framed in two ways: either AI changes everything and we’ve entered the era of the one-person team, or AI can’t replace taste so it’s merely a co-pilot. Both arguments are distractions.

AI reduces operational complexity, slashing the cost of producing “Marketing”. This is a net positive for all marketers, but more marketing doesn’t mean good marketing. Customer attention remains finite, and the way to win customer attention doesn’t change because of AI. 

Scaling marketing has historically required a lot of headcount, software, and coordination. That complexity was a barrier to entry. Now the barrier is gone. A lean team can compete with an org ten times its size. The only differentiator is whether your marketing is actually good, and this comes down to taste, judgment, and systems thinking.

What I find most interesting is how AI enables a new team design. Teams can now organize around how people think, not just what they’ve historically done, and this is the upshot missing in a lot of what I’ve read about AI transformation in marketing. 

No one has ever wanted operational complexity. It was a tax on understanding the customer, generating quality ideas, experimenting, and ultimately shipping good work. With fewer bottlenecks, meetings, hand-offs, and less paperwork, how do you build a team?

Here’s how I’d build one now.

Org Design

Previously growth teams organized around channel specialization: channels were complex, production burdensome, and optimization largely idiosyncratic. Channels typically required dedicated ownership, and dedicated ownership required coordination between dedicated owners. That coordination between those silos created bloated management layers, hierarchy, and onerous processes which diluted the compounding impact of a unified approach to reaching customers.

Specialization was the widely-accepted organizing principle. It was also a trap. Marketing teams were a boiling frog, and AI is the rude awakening.

AI can now streamline operations, channel-specific knowledge is no longer scarce, and tools and media platforms increasingly have AI built in, so marketers get stronger performance without pulling as many levers. 

The reductionist interpretation of this operational offload is that people can now own more channels. This is true, but misses the real point. You can now build around how people think (we’ll call this “cognitive modes”) not what they’ve done before.

You don’t need to hire roles that map exactly to these archetypes: pick the ones that fit your needs. The modes can be staffed with multiple employees, or not. The specialization framework has always been sub-optimal but necessary. However, some specialization is still necessary: you don’t want your systems designer creating motion graphics, so there are roles that need to be specialized, but far fewer than previously.

Cognitive modes lead to better:

Energy allocation - People focus on the highest-leverage activity within their mode rather than being confined to a single surface area or channel.  Cognitive modes align how someone thinks directly to the work they’re doing.

Cohesion - Channel owners optimize for their channel. Cognitive modes optimize for the customer. The difference shows up in the output: a unified message rather than an expression of your org structure.

Talent density -  Specialization narrows the hiring pool to people who’ve worked a specific channel while cognitive modes expands it. Your next Architect might come from product or engineering, your next Anthropologist from UX research. Senior operators are also more attracted to roles with real breadth and ownership than to channel management.

Three objections come up when I describe this to people: How do you maintain accountability without channel ownership? Doesn’t cross-functional often lead to diffuse responsibility? And perhaps most common, how can I hire someone to manage a channel when they’ve never actually done it? This structure changes how collaboration works, how performance gets measured, and how to view experience.

On accountability and collaboration: when organizing by specialization, there are perverse incentives to make a channel work even when it cannot, or to excuse poor channel performance with things like seasonality or increased competition. 

Cognitive modes are interdependent by design: the Storyteller needs the Architect’s instrumentation, the Architect needs the Anthropologist’s customer insight. Some surfaces sit across modes and few modes succeeds in isolation. A landing page is a creative problem and a conversion problem simultaneously. The structure encourages collaboration, not coordination between two siloed surfaces. The result is collective problem-solving, not passing a problem over a wall.

Performance evaluation looks less like ‘paid search grew’ and more like how a PM is assessed: what did you ship, how did the systems you build perform, and what was your contribution to the metrics that actually move the business?

Experience still matters, but it’s more about pattern recognition and judgment than specific channel knowledge. Information that used to be inside baseball is now widely available, and someone with the right cognitive predisposition can get up to speed on a channel faster than you’d expect. The real question is what that predisposition looks like, and how you hire for it. 

People

Hiring mistakes are the biggest mistakes you can make and I’ve made my fair share. Those mistakes happened when I biased my decision toward what someone had done rather than how they thought. I whiffed on employees who were technically competent but cognitively narrow.

Designing around cognitive modes requires a trait-based rubric that mirrors how PMs have historically been hired. Cognitive modes tell you how to organize the work, traits tell you who to hire.

In Practice

Compensation: The artificial pay distinction between technical and non-technical functions is collapsing. You’re asking someone to solve big problems, optimize for impact over single channel performance, and own a broader mandate than a channel owner. If you don’t want just a channel owner, don’t pay them like one.

Junior Employees: Much has been said about how AI will make junior employees a thing of the past. This is wrong. Without a premium on specialization, a junior with the right core traits can be valuable despite limited experience. There’s still a lot of work to do, and you don’t want your senior operators spending their time on less intellectually demanding work. The old entry-level hire was essentially an operations coordinator relegated to trafficking ads, reporting, scheduling and so on. That work is largely gone, but juniors now get exposure across the full growth stack, real ownership earlier, and AI as a learning accelerant. The junior who would have spent two years siloed in a single channel can now develop taste, pattern recognition, and judgment faster than any previous generation.

Screening: Interviews should be less interrogative, and more observational. In-person and take-home cases should now make up the majority of an interview loop. You’ll want the candidate to show they can build, context switch, learn on their own, make decisions with limited information, and display a deep understanding of the customer. Taste, the hardest thing to screen for, comes through in how they explain why they did what they did.

This isn’t an assault on experience: Intuition, taste, and judgment are earned, not inherited from a job title or years in a single channel. What I’m questioning is whether channel specialization is the right proxy for those things. There will be current channel owners who fit well into a cognitive mode and have the right traits, but you’re no longer boxed in by specialization requirements.

AI didn’t suddenly break marketing teams. It revealed that most marketing orgs were designed around the limitations of software, platforms, and human coordination rather than around how exceptional operators create leverage. This is ultimately good news. The removal of the ever-expanding operations overhead is a net positive for good marketers and ultimately, the consumer.

I’m not advocating knee-jerk downsizing. Many teams have done a great job of hiring and already have the right people in place, they just need to be pointed towards the cognitive mode that aligns with their strengths. If your team feels too big for this model, it may have always been too big, but there was a structural constraint and subsequent operational tax that required more people.  

The future of growth teams is not necessarily fewer marketers. It’s fewer coordination layers, broader ownership, and teams organized around judgment rather than channels.

What’s next

This piece lays out a new goalpost for how to design marketing to adapt to AI rather than just adopt it. Most teams are somewhere in the middle of adoption and adaptation. They’re running a hybrid of the old model and the new one, with AI accelerating execution but org design still based on specialization.

Adaptation is hard. I’ll cover how to do it in a follow-up post.