On this page
GTM engineering is the discipline of building and operating go-to-market as a system: the data, automation, scoring and agents that connect marketing, sales and customer success into one pipeline. A company can practise the discipline with nobody holding the GTM engineer title, and it can hire the title without practising the discipline. The practice is what decides whether the job works, and this piece is about the practice rather than the role.
The work is about twenty years old, the title is three
Marketing automation created the first version of the job. Marketo was founded in 2006, and by the early 2010s every B2B marketing team of any size had someone whose job was to keep it running, wire it to Salesforce and explain why the lead counts did not match. That person was called marketing ops. Becca Wilson, who runs marketing operations at Nasuni, said on a call with RevSure that the tools are still relatively new and that attribution, and marketing ops as a function, remain fields with a lot left to learn. She was describing a function that had existed for fifteen years, which says how little of the practice had ever been written down.
Sales got its own version around 2014, when sequencing tools such as Outreach turned outbound into software and sales ops grew a technical wing to manage it. Sean Ellis had coined "growth hacker" in 2010, and by the mid-2010s product-led companies had growth engineers writing code against the product. Three teams, each with its own stack and its own definition of a qualified account.
RevOps was the first attempt to unify them, and the title spread through B2B from roughly 2018. It unified the reporting line, and in most companies it stopped there, because RevOps governed the stack but rarely had the people to rebuild it.
Then the money ran out. The 2022 and 2023 SaaS downturn cut marketing and sales headcount at the same moment AI made it possible to automate the work those people had done: list building, research, first-touch outreach, scoring. A company that could no longer afford three ops teams and a data engineer needed one person who could do enough of each. Clay, whose product had become the workbench for that person, says it coined the title in 2023. Openings passed 3,000 in 2025, up from about 1,400 the year before, and in September 2026 Clay raised $115 million at a $7.1 billion valuation and announced a fund to train more GTM engineers. The origin story is convergence under budget pressure, not a grand plan.
The state those teams merged from is visible in RevSure's own customer-call corpus. Across the 22 enterprise GTM teams whose calls RevSure has indexed, 21 ran Salesforce and 17 named a spreadsheet as part of their revenue stack, sitting alongside Marketo, 6sense, Outreach and a warehouse. The spreadsheet is where the discipline lived before it had a name.
What it borrowed from software engineering
The vocabulary arrived before the practice. Job postings ask for people who ship workflows, run sprints and treat the funnel as a system, and some of that borrowing is real and has made the discipline better.
Version control is the clearest example. Salesforce metadata can be tracked, compared and deployed through DevOps Center or a Git-based pipeline, so a team that does this can answer what changed on Tuesday, which is the first question in every outage. Environments are the second borrowing: a sandbox that mirrors production lets an engineer break a routing rule without a prospect noticing. The third is the idea of the incident, the recognition that a scoring model producing nonsense at nine in the morning is an outage with an owner and a fix, not a line item for the Thursday meeting.
The most important borrowing is the least adopted. A software team knows its service is down within a minute, while most GTM teams find out a form stopped creating leads when a rep asks why it is quiet. One RevSure customer lost fifteen days to a cookie-consent misconfiguration before anyone noticed, a story told in full in the companion piece on what a GTM engineer does day to day. That gap between the break and the noticing is where the discipline has the most to learn from the one it named itself after.
Where the analogy breaks
It breaks in four places, and the discipline has solved none of them.
There is no staging environment for a buyer. A software team can promote from staging to production with some confidence, but a GTM engineer's production is a human being's inbox, the first real test of a sequence is a real prospect, and the failure mode is a reply asking to be removed from the list, or silence.
There is no rollback for an action that reached a person. Code can be reverted, and a bad writeback to Salesforce can be undone with effort, but an email cannot be unsent, an alert that told a rep to call the wrong account cannot be recalled, and an agent that changed hundreds of opportunity stages overnight has already changed the forecast the CRO read at breakfast.
Testing is barely started. Validation rules confirm that a field has a value; almost nothing in wide use confirms that the funnel behaves, that a director-level lead from a webinar at a 500-person company lands in the right queue with the right score within minutes. RevSure has written those tests for its own platform and can report that they are hard, because the correct answer depends on definitions that change every quarter.
And the security review has not caught up with the connectors. Jason Beddall, who runs growth marketing at Ncontracts, described his IT team running a thorough security review of a new integration to Salesforce, the right response from IT and a real cost to the GTM team, and nobody has settled who owns the middle ground. Software engineering solved this with platform teams and paved roads. GTM engineering is still arguing about who owns the road. RevSure's view, which some practitioners will dispute, is that the discipline is a craft with an aspirational name, and that a test suite for the funnel and a rollback for actions are the two things that would make the name accurate.
What a GTM engineering agency actually does
A few hundred searches a month ask this, and it deserves a plain answer. A GTM engineering agency builds outbound and enrichment systems for companies without someone in-house to build them. In practice that means turning an ideal-customer profile into a list, building the enrichment tables that pull firmographic and intent data from a waterfall of providers, writing the sequences and the personalisation logic, connecting the output to Outreach or Salesloft and writing results back to the CRM, and at the better shops, building the reporting that shows whether any of it produced pipeline. Retainers in early 2026 ran roughly $5,000 to $15,000 a month.
It works well for a defined project with a clear end, such as launching an outbound motion or migrating between CRMs, where the agency brings a pattern it has built many times and leaves a working system behind. It works badly as a substitute for the discipline. The agency leaves, the system stays, and six months later a provider changes an API, the ideal-customer profile shifts, the sequences go stale and nobody in the building knows how the tables were built. A bought platform decays the same way. A marketing leader at a software company that had just cut headcount told RevSure, about RevSure, that the team saw the value and had all their data in one place but did not have the bandwidth to use it as they should. A system with no owner decays at the same rate whoever built it, and a vendor that pretends otherwise is selling shelfware.
In-house or agency
Hire in-house when the motion changes often enough that the system needs continuous work, which past $20 million in annual recurring revenue means most of the time. Use an agency for a bounded build with a handover plan that names the in-house owner on day one. The one option to avoid is using an agency to postpone deciding who owns the system, because that decision does not disappear; it gets made by default, badly, when the contract ends. The third path, which did not exist in 2023, is to run the build-and-run part of the discipline on a platform and keep the judgment in-house, and RevSure has published the time, cost and risk math for that choice.
Where the discipline is heading
The unit of work is moving from the workflow to the decision. For three years a GTM engineer's output was a Clay table or a routing rule; over the next three it becomes a decision policy, what an agent may do, on what evidence, with what approval, and how its choices get reviewed. MIT research reported by Forbes in August 2025 found that roughly 95% of enterprise generative AI pilots produced no measurable return, with the cause traced to business context rather than model capability, and the GTM engineer who thrives is the one who owns that context.
The integration work is going away. Model Context Protocol and its equivalents are standardising how agents connect to Salesforce, HubSpot and the warehouse, so the custom connector that filled a GTM engineer's 2024 was a temporary job, and what replaces it is the harder question of what an agent should know and touch. The safety net will be built by platforms, because no single company can afford to build rollback for go-to-market actions on its own. RevSure's GTM Harness runs every agent action through a propose, approve, commit and roll back loop with a decision trace; writebacks can be rolled back, and anything that would reach a person waits at propose until a human approves it, since the only rollback for an email is not sending it. That answers the rollback problem, not the testing one, which nobody has answered.
The discipline will also get a curriculum, because the current way in, described in one RevSure engineer's account of falling into the role, is accident, and a practice people fall into cannot staff the thousands of open roles the job boards show. The title will keep changing, but the discipline, now that it has been named, will not go back to being three teams that do not talk to each other. The build-and-run part of it is what RevSure's AI GTM Engineer exists to run, and the post on what an AI GTM Engineer is covers how that differs from the human role.