The Forward Deployed Growth Engineer

The AI industry has a new obsession, and for once it isn't a model.

It's a role. The forward deployed engineer — invented at Palantir in 2005, now the position OpenAI, Anthropic, and Google are hiring for as fast as they can. TechCrunch calls it the industry's latest talent obsession. At the start of 2026, maybe one company in ten was planning to hire one. By summer, seven in ten.

Why? Because the industry finally admitted something uncomfortable: 95% of enterprise AI pilots show no measurable business impact (MIT, 2025). Not because the technology fails — because nobody stays to ship it. The demo works. The deck is beautiful. The pilot dies in a staging environment, and the consultants who recommended it are three clients away by then.

The forward deployed engineer is the antidote. One person who embeds inside the customer's team, commits to the customer's repo, builds against the customer's real data and real constraints — and is still there when it breaks in month six, because they own the outcome, not the engagement.

Growth has had this exact disease for twenty years.

Swap "AI pilot" for "growth initiative" and the failure mode is identical. The agency pulls levers from the outside and owns nothing downstream. The consultant diagnoses beautifully and leaves a roadmap nobody builds. The tracking lies quietly the whole time, so even the wins can't be proven. Most growth problems look like marketing problems; underneath they're systems problems — and systems problems don't get fixed by people who leave.

I've worked forward deployed for 25 years. I just didn't have the name for it.

At Memoria Holdings I didn't advise the membership business into existence — I built it: server-side measurement, the composable CDP, the ML decision layer that steers acquisition, a production AI voice agent recovering carts. When a checkout failure silently killed an acquisition channel, I found it, and I was the one who fixed it — because I was inside the system, not reviewing it from outside.

That's the whole difference. It's the difference between a report about your revenue leak and standing in the water with a wrench.

What forward deployed means when the product is your growth engine

At the AI labs, forward deployed engineers feed what they learn back into the mothership's product. Independent, there's no mothership — the product is your growth engine itself. The loop is the same one on my front page: embed and diagnose, instrument so decisions run on truth, build and automate, evaluate, and stay on it while it compounds. The engagement ends when the number has moved and your team owns a running system — not when the deck is delivered, because there is no deck.

Who this is for: operators and PE portfolio companies with a growth engine that leaks — revenue, signal, or time — who want the fix built, in production, by the same person who diagnosed it.

Who it isn't for: anyone shopping for a strategy document. There are excellent firms for that. They will leave you a deck.