Google Says Its AI Can Do the Work of Forward Deployed Engineers

In an interview published by The Information on August 18, Andi Gutmans — vice president and general manager of Google’s Data Cloud — made a claim that cuts against the defining talent trend of the AI industry’s current phase: Google’s AI, he said, can do the work of forward deployed engineers.

The forward deployed engineer, or FDE, is the engineer who embeds inside a customer’s organization to build, integrate, and ship AI systems on the customer’s own messy data and processes. The role was invented by Palantir years ago, and in 2026 it has become the single most sought-after specialist in enterprise AI. Google’s argument, per Gutmans, is that the manual context-gathering work FDEs do — mapping schemas, wiring together data pipelines, encoding business logic — can instead be generated by AI itself, in the form of knowledge graphs and semantic models that Google’s platform constructs automatically.

That “AI-generated context,” Gutmans told The Information’s Kevin McLaughlin, is a competitive advantage versus rivals that rely on a more manual approach. It is a direct shot at the services-heavy playbook of Palantir, OpenAI, Anthropic, and the big consultancies — all of whom are betting billions on humans in the loop.

The irony: Google was just hiring FDEs by the hundreds

Three months ago, Google was on the other side of this argument. In May 2026, The Information reported that Google planned to hire hundreds of engineers to help customers adopt its business AI products. Cloud CEO Thomas Kurian had announced a restructured AI-focused organization built specifically around investing in additional forward deployed engineers. Google Cloud’s applied-AI FDE postings advertised base salaries of $127,000 to $183,000, with senior bands in the low $200,000s; Google DeepMind’s FDE roles reached $174,000 to $253,000 plus equity, with observers pegging average packages around $238,000.

Now the company’s Data Cloud leadership is signaling that the endgame is software doing that job. The two positions are not necessarily contradictory — hire humans to deliver today, automate the delivery tomorrow — but the sequencing matters. Google is effectively telling enterprises that the FDE model its rivals are scaling is a transitional artifact, not the destination.

Why the FDE market is so hot right now

To understand why Gutmans’ claim is provocative, consider the state of the market. A C&T study covered by TechCrunch in late July projected demand for FDEs to surge by 2,100% by the end of 2026. At the start of the year, only 5% to 10% of companies planned to hire FDEs, mostly for small pilots. By the end of Q2, that figure had jumped to 70%, with the largest consulting and services firms reporting they need to increase FDE headcount tenfold, building teams of 20 to 100.

The supply side is brutally thin. The study counted roughly 17,000 FDEs in the US market, a large share of whom already work at Palantir — and estimated that only around 2,000 US engineers have the expertise to deliver meaningful AI ROI. Salaries have climbed 10–25% into the $250,000–$300,000 range. “This is all happening at a speed I’ve never seen. Enterprises are hiring in the middle of summer,” C&T founder Jeff Christian told TechCrunch.

The frontier labs have responded by building their own FDE armies: Anthropic backs “Ode with Anthropic,” while OpenAI runs its “Deployment Company,” both staffed with engineers whose sole mission is embedding their employers’ models deep inside enterprise workflows. Accenture and Google Cloud themselves announced a partnership in April to field “thousands” of AI-skilled engineers and FDEs.

The economics behind the claim

The pressure driving all of this is return on investment. Enterprises have spent hundreds of millions — in many cases billions — on AI, and Wall Street is losing patience. Christian’s warning to TechCrunch was blunt: this fall, investors will start “punishing those that have spent hundreds of millions, maybe even billions on this, and aren’t generating ROI, and rewarding those that have.”

Human-delivered deployment doesn’t scale and doesn’t margin. Every FDE team is a services cost center that caps how fast a vendor’s AI can spread through the enterprise. If Google can genuinely generate the contextual layer — schemas understood, relationships mapped, business semantics encoded — automatically from a customer’s data, the delivery bottleneck collapses. That is the strategic logic of Google’s Agentic Data Cloud and its universal context engine, unveiled at Cloud Next in April, and of the broader shift from Vertex AI toward the agent-first Gemini Enterprise platform.

There’s also a trust angle that plays in Google’s favor. Enterprises are increasingly wary of handing proprietary business processes to OpenAI’s and Anthropic’s deployment teams, fearing those processes could inform competing products. Christian’s clients are building internal FDE teams precisely to keep that knowledge in-house. An automated context engine that lives inside a customer’s own cloud tenant offers a third path: deployment without disclosure.

Skepticism from the people who do the job

The FDE world’s response to Google’s claim was immediate and unimpressed. “Clearly Google hasn’t met our FDEs,” quipped data platform vendor Chalk on X, quote-tweeting the story within hours.

The skeptics have a point grounded in the same C&T research. Chris Taylor, CEO of Ode with Anthropic, drew a sharp line between routine rollout work and real transformation: “Many FDEs are well equipped to help you roll Claude Code out to your workforce. Very few are capable of building your flagship AI product feature.” AI-generated knowledge graphs may well absorb the first category — the schema mapping, the pipeline plumbing, the boilerplate integration. The second category, which involves organizational trust, political navigation, and judgment about what to build at all, is where both human FDEs and autonomous agents tend to stall.

Even Christian, the loudest bull on FDE demand, concedes the role’s shelf life. “Maybe in two years, everything’s automated, and agents are automating agents as opposed to humans automating agents,” he said, suggesting the FDE need will migrate toward physical AI before the role potentially disappears within five to ten years.

What to watch

Gutmans’ claim is, for now, a positioning statement backed by real technology, not a demonstrated replacement. The test will be empirical: can Google deploy agents into large enterprises that have never hosted an FDE team, and get them productive on native data without the humans? If yes, the economics of the entire enterprise AI layer — Palantir’s premium, the labs’ deployment subsidiaries, the consultancies’ 10x hiring plans — get repriced quickly. If no, Google will keep quietly hiring the same scarce humans as everyone else, at the same escalating salaries.

Either way, the industry’s most important experiment just got a public hypothesis: that the hottest job in AI is also the first one AI should be able to do.