How I would design a GTM team in the age of agents

I recently re-read parts of Leopold Aschenbrenner’s “Situational Awareness” essay.

One line has been stuck in my head:

“AGI by 2027 is strikingly plausible.”

You do not need to agree with the exact timeline for the implication to matter.

Because if even a softer version of that trajectory is directionally right, many companies are still designing their GTM teams for the wrong world.

I have spent the last 10+ years in and around fast growth startups and scaleups, often close to the CMO seat.

And for most of that period, the basic GTM operating model has looked surprisingly similar.

You build a team around functions: product marketing, content, performance, design, web, marketing ops, sales, RevOps and customer success.

Then you try to make the handoffs work.

That model made sense when execution was expensive.

If you wanted to turn a market insight into something live in the market, you needed specialists, timelines, briefs, reviews, agencies, dashboards and a lot of coordination.

The team was built around scarcity.

Production was slow. Technical work sat with a few people. Distribution knowledge was fragmented. And time was always the bottleneck.

But that scarcity is changing fast.

Marc Andreessen wrote that software is eating the world. Dario Amodei has written about a world where AI systems compress years of intellectual work. Jensen Huang talks about AI factories. Karpathy has been pointing at the shift from hand-written software to systems shaped through data, feedback and training.

Different lenses, same direction:

Capability is scaling.

And when capability scales, the question for GTM is not only:

“How do we use AI tools?”

That is too small.

The better question is:

“How should we build the team if execution is no longer the scarcest resource?”

Because I think many companies will make the obvious move.

They will keep the old org chart, the same functional roles, the same handoffs, the same planning cycles, and simply add AI tools on top.

The content marketer gets ChatGPT. The designer gets Midjourney. The developer gets Cursor. The sales rep gets AI email drafts. Marketing ops gets better automation.

Useful, yes.

But not a new operating model.

It is the old model with faster parts.

And I think that will become a real disadvantage.

The old team was built around scarcity

Most GTM teams were designed for a world where execution was hard to scale.

If you wanted to turn a commercial insight into something live in the market, you usually needed a chain of people around it.

Strategy had to become research. Research had to become copy. Copy needed design. Design needed web or ops. Distribution pulled in paid, CRM, reporting and sales coordination.

This made sense at the time.

The tools were harder, the work was slower, and very few people could move across the full stack without losing quality.

So the team became a relay race.

And the problem with relay races is that every handoff loses context.

The customer insight gets simplified. The positioning gets softened. The weird but important nuance disappears. The campaign becomes a task, the task becomes an asset, and the asset ships after the original market signal has already cooled down.

In complex B2B, that is especially painful.

The hard part is rarely “make more stuff.”

The hard part is understanding the buying committee, finding the commercial bottleneck, shaping a narrative that matters, getting the right people to see it, and learning quickly from the market.

That loop is where growth happens.

And most teams run it too slowly.

What changed over the last decade

This is the part I keep coming back to from my own work.

Over the last decade, a lot of GTM progress came from improving the same basic machine.

CRMs got stronger. Attribution became cleaner. Paid social platforms became more precise. Marketing automation, intent data, sales engagement tools, BI dashboards, agencies, freelancers and fractional experts all made the system more sophisticated.

And honestly, a lot of that worked.

A modern GTM team in 2024 is obviously more advanced than one in 2014.

But the operating model underneath did not change as much as we like to think.

Insight still moved through too many people. Strategy still got separated from execution. Commercial context still got lost in briefs. The market still had to wait for the internal machine to catch up.

AI changes that.

Not because it magically creates strategy.

Because it lets a small number of high-judgment operators compress the distance between thinking, building, distributing and learning.

That is the shift.

The best teams will not just produce more.

They will learn faster.

The 2026 team should be built around learning speed

If I were building a GTM team in 2026, I would not start with functions.

I would start with loops.

What are the loops that actually create learning and revenue?

One loop might start with a market signal, turn into a commercial hypothesis, become a positioning angle, get translated into an audience or account selection, ship as an asset, move through distribution and sales follow-up, then come back as feedback for the next iteration.

Another loop might start with a founder insight, become a category narrative, get supported by customer proof, turn into a landing page and paid test, create pipeline signal, sharpen the sales conversation, and feed the next version of the narrative.

These loops matter more than the job titles.

The question is not:

“Do we have a content person?”

The question is:

“How fast can we turn a good commercial insight into something live in the market, and how fast can we learn from it?”

That is where AI changes team design.

Not because AI replaces the team.

Because it changes which parts of the team need to be people, which parts need to be systems, and which parts need to be agents.

The new core role: AI-native commercial operator

I think the best 2026 GTM teams will have a role that many companies do not know how to hire for yet.

Call it an AI-native commercial operator.

Not a prompt engineer. Not a classic growth marketer. Not a marketing ops person. Not a fractional CMO in the old sense.

More like a commercial systems builder.

Someone who can sit with a founder and talk about category, ICP, market timing, buying committees, offer design and sales narrative.

Then open the laptop and start turning that thinking into a working system.

Maybe that means a landing page, a research pipeline, an account list and a paid social angle.

Maybe it means a CRM workflow, a founder-led content system, a sales follow-up sequence and a reporting loop.

Maybe it means building a small internal tool because the team needs a better way to see the market.

The important part is not the asset.

The important part is the distance between thinking and shipping.

Not six weeks later.

Now.

Not because the first version is perfect.

Because the learning loop starts earlier.

That is the point.

I think this is one of the biggest changes in commercial work.

The best operators are not just doing tasks.

They are shaping systems around how they think: how they research a market, judge customer language, prioritize, decide what should be automated, and know when the output is not good enough.

The AI-native operator is not just a better marketer.

It is a marketer with an execution engine.

Hire for judgment before tool usage

A mistake I see coming:

Companies will hire for AI tool fluency too literally.

They will ask if someone knows Clay, Cursor, agents, prompt engineering and outbound automation.

Those questions are fine.

But they are not enough.

Tools change too fast.

The better hiring questions are more annoying.

Can this person diagnose why growth is not working?

Can they tell the difference between a distribution problem and a positioning problem?

Do they understand messy buying committees?

Can they work from customer calls, Slack threads, CRM data and founder intuition without needing a perfect brief?

Can they turn ambiguity into a testable commercial hypothesis?

Can they ship a credible first version without waiting for a full team?

Can they judge when AI output is strategically weak, even if it looks polished?

Can they build workflows instead of just assets?

And maybe most importantly:

Can they explain what should not be automated?

That last one matters.

The best AI-native operators are not the people who automate everything.

They are the people who know where human taste, judgment and trust need to stay close to the work.

The team model I would use

If I were building a GTM team from scratch in 2026, I would design it less like a set of departments and more like a set of layers.

The first layer is strategy and judgment.

This is the senior commercial brain of the team: positioning, category, ICP, market timing, buying committee, demand creation, offer design and sales narrative.

You need people here who can think.

Not just manage.

This is where many teams underinvest because they confuse motion with progress.

AI will make motion cheaper.

That makes judgment more valuable.

The second layer is the operator layer.

This is where AI-native commercial operators connect strategy to execution.

They do not need to be the best designer, developer, copywriter, analyst and ops person in the company.

But they need enough range to move across the system.

They should be able to build first versions, connect tools, brief agents, inspect outputs, run tests and keep the commercial logic intact.

This is the layer I think many companies are missing.

Then you have specialists.

They still matter a lot.

A great designer will beat a generalist with AI. A great performance marketer will see things a generalist misses. A strong product marketer can sharpen a narrative in ways a tool will not. A serious lifecycle or RevOps person can prevent chaos.

But I would use specialists differently.

Not as the first stop for every idea.

I would bring them in where the loop has signal and the craft needs to scale.

First prove the direction.

Then apply specialist depth.

The fourth layer is agents and automation.

This is the execution surface: research agents, scrapers, CRM workflows, content repurposing, account enrichment, meeting summaries, competitive monitoring, reporting, drafting, QA and internal tools.

This layer should remove coordination work, not create more noise.

If agents only produce more tasks for humans to review, the system is badly designed.

And finally, you need an approval and taste layer.

This is the part many AI maximalists underplay.

Someone still needs to decide if the work is true, sharp, on-brand and commercially useful.

Someone needs to know whether it will matter to the buyer.

Someone needs to feel when the work creates trust, and when it quietly destroys it.

The goal is not full automation.

The goal is faster execution with better judgment density.

What I would stop hiring for too early

In 2026, I would be careful with roles that mainly exist because the old system had too many handoffs.

I would be careful hiring pure coordinators before there is enough complexity to coordinate.

I would not rush to hire channel owners before the positioning is clear.

I would avoid scaling content production before the company has a strong POV.

I would not add ops too early if the workflow itself is still unclear.

And I would be careful bringing in specialists before the company knows which loop is actually working.

Not because those roles are bad.

Because hiring them too early can freeze the wrong operating model into the company.

A lot of startups do this.

They feel pain, hire a function, the function creates a process, the process creates dependencies, and the dependencies slow learning.

Then everyone wonders why the team got bigger but not faster.

I would rather start with fewer people who can own larger loops.

Then add specialists around proven bottlenecks.

The new hiring bar

If I were interviewing a senior growth hire today, I would not only ask about channels.

I would give them a messy commercial problem and watch how they think.

For example:

“We are getting interest from CFOs, but our messaging was built for RevOps. Pipeline quality is mixed. Sales says the best conversations happen when prospects already understand the cost of the problem. What would you do in the next two weeks?”

A classic marketer might propose a campaign. A strategist might propose a positioning project. A marketing ops person might propose workflow changes. A content person might propose thought leadership.

None of those answers are automatically wrong.

But the strongest AI-native commercial operator should be able to connect the whole loop.

They would pull customer language, analyze the buying committee, find the CFO pain, shape the narrative, build a test page, create the content angle, define the account list, draft the sales follow-up, set up the workflow, launch a small test, read the signal and decide what to improve.

That is the difference.

Not more activity.

More connected execution.

The uncomfortable part

This changes how we evaluate talent.

A person who looks “too generalist” on an old org chart may be extremely valuable in the new model.

And a person who looks senior in the old model may struggle if their leverage came mostly from managing handoffs.

The 2026 GTM team needs fewer people who only move work through the machine.

It needs more people who can redesign the machine.

That does not mean everyone becomes technical.

But it does mean commercial leaders need more systems fluency.

They need to understand APIs, agents, memory, data flows, automation, context windows, evals, workflows and approval gates well enough to design with them.

Not as a hobby.

As part of the job.

The question I would ask

If you are building a GTM team now, I think the most useful question is not:

“How do we use AI?”

It is:

“What would this team look like if execution was no longer the scarcest resource?”

Because that forces better questions.

Where is judgment scarce?

Where is context trapped?

Which handoffs are killing speed?

Where do we need deep craft?

What can agents do without creating review debt?

Where should humans approve?

And where do we need an operator who can connect the system end to end?

That is the conversation I think more founders, CMOs and VCs should be having.

Not because AI makes teams irrelevant.

Because it changes what a great team is.

And in 2026, the best GTM teams will not just be teams with AI tools.

They will be teams designed around AI-native operators, specialist depth, human judgment and fast learning loops.

That is the new model.

And I think it will become the new hiring bar.