A GTM Engineer req crossed your desk last quarter. You approved the band, maybe stretched it, and six months later you still can’t say what the person built. That isn’t a hiring miss — it’s a definition miss. AI didn’t just add titles to the go-to-market org chart. It unhitched titles from the work underneath them, so the same four words on a job post now describe four different jobs at four different companies. If you hire off the title, you’re buying a label. If you hire off a documented build — the tools, the code, the outputs, the before-and-after number — you’re buying a person who has already done the thing.
The title inflation is real, and it’s fast
GTM Engineer postings grew 205% year over year — which is why you’ve seen the title five times this month when you’d never seen it two years ago. That kind of curve doesn’t come from a new discipline maturing. It comes from existing work getting relabeled, because the label attracts better applicants and bigger budgets.
Here’s the part that should change how you write reqs: 9 out of 10 responsibilities listed in GTM Engineer postings also show up in RevOps Engineer postings. Strip the title off the top of the page and most of these documents are indistinguishable. Which means the market is paying a premium for a word, not a skill set — and you’re on the paying end of that.
AI skill requirements have also migrated out of engineering entirely. 51% of job postings requiring AI skills now sit outside IT and computer science occupations. Those requirements landed on commercial titles: demand gen, marketing ops, sales development, product marketing. The title didn’t change. The job did. That gap is where bad hires live.
Four titles, one actual job
The clearest way to see the unhitching is side by side. These are the four labels most commonly used for overlapping work right now, and what a candidate holding each one has usually actually done.
| Title on the resume | What they usually built | What you may be assuming |
|---|---|---|
| GTM Engineer | Enrichment pipelines, outbound sequencing logic, API glue between CRM and data warehouse | A full-stack engineer who also owns quota |
| RevOps Engineer | Same pipelines, same CRM architecture, routing and attribution logic | A reporting and forecasting administrator |
| Marketing Ops / AI-enabled | Lifecycle automation, scoring models, LLM-assisted content and segmentation ops | A campaign builder who can also write Python |
| AI SDR owner | Prompt and sequence design, agent QA, deliverability and reply triage | A quota-carrying rep whose agents do the dialing |
Read the second column against the third. The assumption is almost always broader than the build. That’s not candidate dishonesty — it’s title drift. The market handed them a word that implies more than the role allowed them to do.
The resume says what they were called. Only the build says what they can do.
The “AI SDR” tells you exactly how early this is
2025 marked the first year that “AI SDRs” appeared as a distinct category in sales development benchmarking — at 1% of respondents. One percent. The title is everywhere in your LinkedIn feed and on every vendor’s homepage, and it represents a sliver of actual teams.
That gap matters for two reasons. First, there is no established playbook a candidate could have learned — so anyone claiming five years of it is claiming something that didn’t exist. Second, comp benchmarks for the title are being set by a handful of outliers, which means the number a candidate quotes you has almost no market behind it. When you’re hiring into a sales development function and someone asks for a 30% premium on the strength of an AI-forward title, ask what they shipped. Agent QA process? Deliverability recovery after a domain burn? A measured lift in connect rate? Those are answerable questions. “AI SDR” is not.
Agents multiply, productivity doesn’t follow
Gartner projects that by 2028, AI agents will outnumber sellers 10 to 1 — yet fewer than 40% of sellers will say the agents improved their productivity. Sit with the second half of that. The tooling arrives at scale and the majority of the humans using it report no gain.
The reason is almost never the model. It’s that nobody owns the orchestration — the data hygiene, the handoff rules, the escalation logic, the monitoring when an agent silently stops working. Those responsibilities get stapled onto an existing title during a reorg, and the person holding it was hired for something else entirely.
So when you write a req that says “AI-enabled,” decide which of these you’re actually buying:
The builder
Writes the code. Owns the API integrations, the enrichment waterfall, the warehouse queries. Comfortable in a repo. Usually the person a GTM engineering search should target, and usually the most expensive of the three.
The operator
Configures and maintains. Lives in the CRM, the MAP, the sequencing tool. Can build a scoring model and debug a broken routing rule, won’t be writing services. This is a marketing operations profile with an AI layer, not an engineering one.
The orchestrator
Designs the motion, QAs the agents, owns the number. Less code, more judgment about where automation helps and where it torches your domain reputation. Often the hardest to assess because the output is a decision, not an artifact.
One title covers all three in today’s market. One band does not.
Why marketing reqs get rewritten mid-search
In marketing and PR, 8% of job postings now require AI skills, growing 50% annually. Small base, steep slope — and that combination is precisely why reqs mutate in flight. A search opens as “Demand Gen Manager.” Four weeks in, someone on the team sees a competitor’s post, and the req becomes “AI-first Growth Engineer.” The band moves. The interview loop doesn’t.
You can feel this in your pipeline before you can name it. Candidates who looked strong in week two now look underqualified against a spec they never applied to. Your interviewers disagree about what “good” is because they’re each holding a different version of the role. Cycle time stretches, the best people take other offers, and you reopen in Q3 — which is the expensive version of a definition problem.
The fix is unglamorous: freeze the scope before the search opens, and write it as outputs rather than adjectives. Our marketing hiring guide and the GTM engineer hiring guide both exist for this moment, because the scorecard you write in week zero is what keeps the req from drifting in week six.
How to hire off a build, not a title
Title-blind assessment isn’t ideology, it’s cost control. When 9 out of 10 responsibilities are shared across two different titles, the title carries almost no signal — so put your signal elsewhere.
Ask for four things, in this order:
- The artifact. A repo, a workflow screenshot, a Looker board, a doc describing the pipeline. Not a deck about the strategy — the thing itself.
- The stack, named specifically. Not “AI tools.” Clay, n8n, Snowflake, HubSpot, Marketo, dbt, whatever it was. Vague stacks mean borrowed credit.
- The before-and-after. What was the metric when they arrived, what was it when they left, and what else was changing at the same time. Candidates who volunteer the confounders are the ones who actually owned it.
- The failure. Every real build has one — an agent that went quiet, a domain that got burned, an enrichment source that degraded. People who’ve shipped can describe it in detail. People who’ve watched can’t.
Then pay against the build. If a candidate’s documented work is RevOps work, benchmark it as RevOps work regardless of the word on the resume. If it genuinely spans code and commercial ownership, pay for that — those people are rare and worth it. Salary benchmarks are how you keep that decision anchored to the market instead of to the last candidate who negotiated hard.
What this means for your next req
Three practical moves, none of which require a reorg.
Write the scorecard before the title. List the five outcomes you’d fire someone for missing. Then pick whichever title the market will respond to — the title is marketing copy for the req, and that’s fine, as long as the scorecard is the thing you interview against.
Decide builder, operator or orchestrator on day one. Put it in writing. Tell your interviewers. The single most common cause of a stalled GTM search is three interviewers evaluating three different jobs.
Treat AI claims as a skills question, not a seniority question. With 51% of AI-skill postings now outside technical occupations, “works with AI” describes most of your commercial org inside two years. It is not a differentiator and it should not carry a premium on its own.
The companies filling these roles in days, not weeks aren’t the ones with the most fashionable title on the post. They’re the ones who defined the work precisely enough that a qualified person recognizes themselves in it — and specific reqs beat spray-and-pray every time. Define the build, and top-tier talent stops being a search problem.
