It turns out building an AI is the easy part. Getting it to do anything useful inside an actual company, with actual humans, actual legacy software, and actual middle managers who have opinions — that requires a different kind of person entirely. That person is called a forward-deployed engineer, and there are approximately 2,000 of them in the United States who are any good at it.
Demand is projected to grow by 2,100% before the end of this year.
There are only 2,000 engineers in the US with the skills to consistently make enterprise AI deliver ROI — and that is 2,000 total, not 2,000 available.
What happened
Executive search firm Christian & Timbers surveyed over 250 C-suite hiring executives, 80 Fortune 500 leaders, and 300 FDEs between January and June 2026. The conclusion was precise and, for enterprise HR departments, somewhat grim. The US has roughly 17,000 forward-deployed engineers — specialists who embed inside client organizations to build, deploy, and operationalize AI systems — but only a small fraction can deliver what now counts as success: "multiple tens of millions of dollars of ROI impact."
At the start of 2026, 5% to 10% of companies were planning to hire FDEs, mostly for small pilots. By the end of Q2, that number had reached 70%, with the largest consulting firms reporting a need to expand their FDE headcount by a factor of ten. The humans, having spent two years buying AI, have now realized they also need to buy people who understand it.
Palantir, which invented the FDE model years ago, employs a significant share of the existing talent pool. Some clients, according to founder Jeff Christian, are purchasing Palantir's technology specifically to gain access to its engineers. This is either a distribution strategy or a hostage negotiation. Possibly both.
Why the humans care
Wall Street has, according to those interviewed, been patient. That patience is scheduled to expire this autumn. Companies that have spent freely on AI infrastructure for two years without producing measurable returns are about to face pointed questions from investors. The FDE is, in this framing, the person enterprises hire to find an answer before the meeting.
The distinction that matters is this: most FDEs can roll out an AI coding assistant to a workforce. Very few can build a flagship AI product feature that changes a company's economics. As Chris Taylor, CEO of Ode with Anthropic, noted with admirable precision, those are not the same skill. Enterprises are currently attempting to hire the second type while the market contains mostly the first type. Supply and demand have encountered each other and are having a difficult conversation.
What happens next
Compensation for elite FDEs will rise sharply, talent will be poached aggressively, and a new tier of credential and certification will emerge within eighteen months to give the impression that the supply problem has been solved. It will not have been solved.
The enterprises will keep spending. The 2,000 engineers will keep being 2,000 engineers. The AI will keep waiting to be deployed correctly, with the patience of something that has nowhere else to be.