GTM engineering: the fastest growing role in tech

Before we drive in: I'm hosting a free Lightning Lesson with Yusuf Ahmed, founder of Zevenue, on the fundamentals of GTM engineering. Yusuf will show what it looks like in practice, from finding the right accounts to building the AI workflows behind them.

Save your spot here →

Welcome back.

GTM has been having its moment these past few months. In Pave’s analysis of workforce data from more than 8,700 companies, the share employing at least one GTM engineer rose from under 0.1% in January 2023 to 1.3% in January 2026. Pave’s CEO describes it as one of the fastest-growing jobs in the company’s dataset.

On top of that, according to a Bloomberry's analysis report, postings have grown by 205% year over year. So it’s safe to say that this discipline has quickly become a way of running go-to-market that founders want in their company.

Part of the reason for this is that AI flipped where the bottleneck sits. Building got easy, while getting customers stayed hard.

When everyone can ship a feature in barely an afternoon, the only companies that win are the ones that figure out distribution. That's what GTM engineering solves

Previously, a GTM team used to need considerable technical help to research thousands of accounts, connect that research to its CRM, and act on it. Now, AI and workflow tools let someone close to the sales problem build and test that process much faster. Ideas that once sat in an engineering queue can now become working experiments.

There is also far more information to work with now. A company’s job postings, website changes, sales calls, product activity, and CRM history can all tell you something about whether company Y is a good prospect. But you’ll find that these clues live in different places. Someone still has to decide which ones matter, connect them, and make sure the team acts on current information.

And AI has made it easy to produce more outreach. That makes volume less of an advantage on its own. The difficult part is deciding who has a real reason to hear from you, checking that your claims are true, and learning whether the approach leads to revenue. GTM engineering has become popular because it gives teams a way to build those decisions into the work, rather than asking every rep to start from scratch.

To put it simply, it is running go-to-market as a system instead of a series of campaigns. You take the judgment that lives in your best rep's head (who to target, why now, what to say) and turn it into AI workflows that run every week, check their own work, and learn from results.

This is what that looks like:

And what does this new way get you? judgment at scale.

Since AI has made it easy to send more, we’ve been seeing a lot of teams slide over from guesswork straight into spray-and-pray. GTM engineering moves you up instead. Think: you get your best rep's judgment applied to every prospect.

In practice, that means reps spend their time talking instead of researching, outreach only goes to accounts with a real reason to talk now, and you find out which signals actually turn into revenue.

None of this requires a computer science degree. It just requires knowing how deals get made and being willing to build.

Every GTM engineering workflow, whatever it's for, runs through the same five stages. Each one mapping to a building block in Claude Code.

To make it concrete, let's follow one example through all five. Say you sell scheduling software to regional HVAC contractors.

A skill is a reusable set of instructions that captures how your team does a task. Write down what your best rep looks for, then turn it into a skill that reads each prospect’s website and public data and ranks the signals that make them a fit.

Your rep knows that a new branch, a dispatcher job posting, or a new service area usually means scheduling is getting harder. The skill looks for exactly those clues, so you end up with a list nobody else can simply buy.

Give each company its own subagent. Each one researches in its own context window and hands back a short summary, so 50 companies do not blur together in one session. Every useful claim gets a source and a date, and facts stay separate from guesses.

Northline HVAC comes back with three locations and a dispatcher posting from last week, both linked. The record can say that expansion may be putting pressure on scheduling. It cannot claim their scheduling is broken.

A CLAUDE.md file gives every session your ICP, territory, and product details. MCP connects Claude to your CRM and call notes. Hooks run rules that must happen every time, without asking the model to decide.

A hook checks each account against the CRM. Metro Comfort has an open deal, so it goes to its owner instead of getting a cold email. Existing customers never make the list.

A writing skill drafts outreach only after a rep accepts the account, using the checked signal and your messaging rules. Writing comes fourth on purpose: by now, the hard work is done, and the email is its output.

The draft to Northline asks how they coordinate dispatch across three locations. It does not pretend to know their process is failing.

Reps log why they reject an account: wrong size, stale signal, weak evidence. Replies and meetings get tied back to the signal that started the conversation. You can then look for patterns, test whether they hold up, and update the Find skill.

If branch openings lead to better conversations than dispatcher postings, next week’s list should reflect what you learned. Over time, you can compare those early signals with qualified opportunities, not just replies.

Once the individual stages work, put them on a schedule. Routines can run the process every Monday and leave reps a short review queue. Plugins can package the skills so the team works from the same system.

Start with one decision. Do not build all five stages at once. Begin with: Who should our team contact this week?

Capture your rep’s judgment. Ask your best rep for ten accounts they pursued and ten they passed on. Write down what counts as fit, what makes the timing right, what rules an account out, and what evidence they need before trusting a recommendation.

Build and test the first stage. Put your company context in a CLAUDE.md file, then turn those notes into a Find skill. Test it on the 20 accounts before using it on a larger list. Does it catch the accounts your rep passed on? Are its claims sourced? Would the rep agree with its recommendations? If not, fix the skill before going further.

Add guardrails, then expand. Connect the CRM so existing customers and accounts owned by other reps are flagged. Run a small weekly batch into a rep review queue, not straight into outreach. Compare meetings and qualified opportunities with your current process. Once Find works, add Enrich, then the remaining stages.

One caveat: if you have not nailed your ICP yet, this will help you scale confusion faster. Automation multiplies whatever judgment you already have, good or bad.

The short version: GTM engineering builds the path from information to a sales decision, then proves whether that decision produces revenue.

If you’re interested in learning more, join my free Lighting Lesson with Yusuf. Sign up here.

As always, thanks for reading.

Haroon

Founder, AI Ready