Here is the statistic nobody’s LinkedIn feed predicted: SDR headcount at AI-native companies has more than doubled year-over-year — making it the second-fastest-growing GTM role at exactly the companies building the AI that was supposed to eliminate it. The finding comes from Kyle Poyar’s Growth Unhinged (May 20, 2026), built on Sumble job-post data across roughly three million verified companies; it is a single dataset with undisclosed sample sizes, and we attribute it as such. But the named hirers are checkable and telling: Cursor, OpenAI, Decagon, LangChain, Fireworks AI — the fastest-scaling AI companies on earth are staffing human SDR teams while selling the automation story to everyone else.

The two curves that don’t match

The same data shows overall SDR/BDR job posts down 21% year-over-year, and total GTM postings down 15%. So the market is not “SDRs are back” — it is bifurcating. The broad middle of B2B is cutting the role; the AI-native tier is doubling it, and posting ~50% more SDR roles as a share of GTM openings than digital natives at large. Whatever the AI-native companies know about selling AI products, it apparently requires more humans at the top of the funnel, not fewer.

Why the builders are buying humans

Three explanations fit the evidence. First, their pipelines are on fire — when demand doubles yearly, you hire everything, and SDRs are the fastest headcount to convert into meetings. Second, AI products are consultative sells: an SDR selling Decagon is qualifying deployment complexity, security posture, and executive sponsorship — judgment work that the churn data says packaged agents still fumble. Third — the uncomfortable one — the people with the best information about AI-SDR capability are revealed-preference bearish on it for their own revenue. They automate the research and personalization underneath the rep (the composable pattern our playbooks cover) and keep the human on the conversation.

The role is recomposing, not dying

Read alongside the GTM-engineering hiring boom (doubled YoY on the same dataset), the picture is a re-division of labor: the list-building, enrichment, and first-draft work migrates into Clay-class systems run by GTM engineers; the SDR title increasingly means someone who works signals a machine surfaced and runs judgment-heavy conversations a machine can’t. That’s also why AI-native support headcount sits at just 2.2% of GTM staff in the same data — support conversations automated first because their floor for error is lower. Outbound trust-building is proving to be the stickiest human job in the funnel.

What to do with this

If you’re planning 2027 headcount: benchmark against the tier you compete with, not the aggregate — if your rivals are AI-native, they are out-hiring you at the top of funnel while you read obituaries for the role. If you’re an SDR: the winning move is visible in the postings — learn the orchestration layer (our learning path is the syllabus) and become the person the machine reports to. If you’re buying AI SDR software: ask every vendor how many human SDRs they employ. The answer is public, and it is the most honest benchmark in the category.