← Blog

What will go-to-market look like in 2027?

Our view of go-to-market in 2027. AI agents do the execution, building is cheap, and the teams that win are the ones with someone who owns the testing.

4 min readBy WarmUpHow we work

Key takeaways

  • We expect AI agents to do nearly all of the execution in go-to-market by 2027, and a working system to be the baseline, not the advantage.
  • Building will be cheap. Finding what works in a given market will still take months of testing with real buyers.
  • We think the teams that succeed will be the ones with a person who owns that testing, and clear rules for what agents may do.
  • A firm will be paid for directing the agents and answering for the results, not for the hours of building.

By 2027 we expect AI agents to do nearly all of the execution in go-to-market: the research, the lists, the first drafts, the data work. A working system will be the baseline. The teams that succeed will be the ones with someone who owns the testing, and clear rules for what the agents may do.

That is a forecast, not a fact, and the rest of this post is marked the same way. It is Mike's view, as our founder, from running these systems for clients.

What do the teams that succeed in 2027 look like?

We think the teams that succeed in 2027 will look ordinary from the outside. They will have the same agents and the same tools as everyone else. What sets them apart is a person who owns the question "what is working?", reads the results on a set rhythm, and changes the plan when the answer changes.

This question gets asked whenever someone predicts that go-to-market will be run by agents, and it tends to go unanswered. Our answer has three parts.

  • Someone owns the testing. Not as a side task. A named person who reads the numbers every couple of weeks and decides what to drop and what to try.
  • The work lives in tools the company owns. Lists, rules and results sit where an agent can act on them and a person can check them, and they stay when a vendor leaves.
  • The rules for agents are written down. What an agent may change on its own, what waits for a yes, and where the record is kept.

None of these is a technology. All three are decisions a team can make this year.

What will AI agents do in go-to-market by 2027?

By 2027 we expect AI agents to do almost all of the execution in go-to-market. In our own work they already do the research on each company, enrich and check contacts, score companies against past wins, sort replies, draft copy and do the bulk work of a CRM clean-up. We expect that list to grow, and the cost of each item to keep falling.

The consequence is that execution stops being a reason to choose one team over another. Mike's view is that every firm in this field will need these systems as a baseline, and that what a client pays for sits on top of them: deciding what the agents work on, watching what they produce, and answering for the results.

We describe how our agents are run today on the method page, and what they do for each service on pages such as GTM engineering.

If building is cheap, what will still take time?

What will still take time is finding out what works. We expect the build of a growth system to take days in 2027, and the search for the signals, messages and buyers that pay off to take months, as it does now. Neither a person nor an AI knows the answer in advance. It has to be tested with real buyers.

Buyers reply on their own schedule, and a market only holds so many people to test on. An agent can prepare the next experiment in an hour, and then it waits for the replies like everyone else. We set this out in why a growth system takes months when AI can build it in days.

So we do not expect cheap building to make growth fast. We expect it to move the work from building to testing, and to reward the teams that keep testing after the first version is live.

What will companies pay a growth firm for in 2027?

In 2027 we expect companies to pay a growth firm for people, not for output. The agents make the output cheap. What stays scarce is someone who directs them, checks them, sits in the review and takes responsibility when a play fails. That is what our clients tell us they value now, and we expect it to matter more.

Two things we see today point this way. A lot of our clients do not have someone who can be fully dedicated to building and maintaining the system, which is why they hire us. And most keep us running it afterwards. A handover to their own team is rare. Neither of those depends on how hard the system is to build.

We also expect the work to widen. Clients who have seen agents at work in their growth system are already asking us to build them for other jobs in the business, which we take on as AI engineering projects. We wrote more about the fee itself in what you pay a growth firm for when agents do the work.

What could make this forecast wrong?

This forecast could be wrong in two directions. Agents could become reliable enough to run without approvals sooner than we expect, which would shrink the role of the person in the loop. Or buyers could tire of agent-written outreach faster than the tools improve, which would make the human parts worth more, not less.

We hold the second view more strongly than the first. Sending more has never fixed outbound, and agents make sending more very cheap. We expect inboxes to get noisier, and a real reason to write, said by a real person, to stand out more in 2027 than it does today.

On approvals, our rule now is that anything a buyer will read waits for a yes, and a person makes every call. We set out why in how we manage AI agents inside a client's tools. We would relax that one job at a time, where an agent has shown it can be trusted, and not before.

If you want to prepare for 2027 this quarter, do one thing: pick the person who will own the testing. Give them one group of buyers, one reason to write, and a review every two weeks. The tools will keep changing. The habit of reading the results and changing the plan is the part that will still be yours.

Questions, answered.

Is this a forecast or a fact?

A forecast. Everything in this post about 2027 is our view, formed from running growth systems for clients and from watching how the tools have changed this year. We have tried to mark each claim as a view. We may be wrong about the timing, and we will say so here if we are.

Will companies still hire a growth firm in 2027?

We think many will, for a different reason than before. Today most of our clients have nobody who can give the system their full attention, and that does not change when the building gets cheaper. What changes is what the fee is for, which is the people who direct the agents and own the results.

Will AI agents replace sales and marketing teams?

We do not think so. We expect agents to take over the research, the data work and the first drafts, which they already do in our own work. Deciding who to go after, approving what a buyer reads, and holding the conversation when a buyer replies still need a person, and we expect that to hold in 2027.

What should a team do now to be ready?

Name one person who owns testing what works, and give them the time to do it. Put your lists, your rules and your results in tools you own, where an agent can work and a person can check. Write down what an agent may change on its own and what waits for a yes.

Part of

GTM engineering. Outbound aimed by what buyers do.

Bring us your last ten deals.

Thirty minutes. We will tell you which signals exist for your market, what we would build, and what would be yours from the first day.

Or skip the form