If the battle over AI "models" was the first half, then the second half—the battle over "deployment"—is only just beginning.
While market attention remains fixed on model parameters and benchmark scores, AI giants like OpenAI and Anthropic have quietly shifted their strategic focus. They are now spending salaries as high as $280,000 to compete globally for a new type of "special forces" soldier: the Forward Deployed Engineer (FDE).
These "amphibious talents"—who can both dive deep into the fabric of a customer's business and write production-grade code with their own hands—are replacing pure algorithm scientists as the decisive force in the "last mile" of AI commercialization. A "ground war" over AI applications and enterprise services has fully begun.
From Lab to Battlefield: The "Deployment Anxiety" of AI Giants
Nearly three years after ChatGPT ignited the tech frenzy, the soul-searching questions from enterprise customers have grown sharper: "The technology is cool, but how does it make me money?" The crux lies in the vast gulf between "proof of concept" and "scaled production." AI giants have realized that top-tier models alone are not enough—they must also help customers solve real problems. The FDE is precisely the key to solving this "last mile" challenge.
The talent market has responded most directly: according to data from the Indeed platform, monthly job postings for FDEs surged by more than 800% in 2025. OpenAI has already formed an FDE team at the start of the year and plans to expand it to roughly 50 people globally next year. Executives at major AI companies generally believe that deploying engineers at the very start of a customer engagement helps build lasting relationships and safely integrate AI models into enterprises' complex workflows.
Oliver Jay, Managing Director of International at OpenAI, pointed out that AI, unlike cloud software, is not easy to test and deploy. When models need to be deployed at scale, enterprises need more advanced techniques to set up safety guardrails, evaluate model accuracy, and integrate them safely into complex internal data and workflows.
Cat de Jong, who leads Applied AI at Anthropic, also emphasized: "A Fortune 500 bank's needs are completely different from those of a startup building AI-native products." The FDE's job is to dig deep into these differentiated needs and bridge bespoke technology with complex production environments.
To hunt down these "amphibious talents," AI companies are offering jaw-dropping compensation. OpenAI's FDE position posted in New York lists a salary range of $220,000 to $280,000—and that's not counting the highly attractive equity incentives. An arms race over scarce talent has fully escalated.
Decoding the Front Line: Turning AI from "Toy" into "Tool"
The FDE is not a new invention of the AI era; the concept originates from military "forward deployment" and was later pushed to its extreme in the tech world by data giant Palantir. Palantir's FDEs are dispatched to complex sites such as military bases and factory floors, with the mission of untangling requirements from chaos and codifying them into solutions.
Palantir's classic internal "Echo-Delta" two-person model illustrates how they work:
· Echo: Acting as the "strategic scout" and "requirements translator," whose core task is to dig deep into pain points that even customers themselves cannot clearly articulate, and convert them into precise product requirements.
· Delta: Acting as the "tactical strike team" and "solution builder," these engineers possess superb coding skills and, based on the "intelligence" relayed by Echo, can rapidly build, iterate, and deploy practical, usable solutions in the real environment of the customer's systems.
"The most valuable software lies not in how elegant the code is, but in whether it solves the customer's problem." Nic Prettejohn, head of AI at Palantir UK, captured the core of the FDE philosophy.
Today, AI upstarts are going all in on emulating this model. OpenAI's collaboration with agricultural machinery giant John Deere is a prime example. The FDE team stationed itself directly on farms, working alongside agronomists and mechanics, observing the entire process of planting, fertilizing, and harvesting firsthand.
Building on OpenAI's technology, they custom-developed an intelligent spraying system for John Deere that, through real-time analysis of crop images, achieved precision strikes on weeds. Ultimately, this system helped farmers reduce chemical usage by 60% to 70%, creating enormous economic value while also delivering significant environmental benefits.
"We experiment and innovate together with our customers. The real needs learned from different industries, in turn, shape OpenAI's own research directions and product roadmap," said Arnaud Fournier, who leads FDEs for Europe and the Middle East at OpenAI. This reveals another strategic value of the FDE: they are the "sensors" that transmit the market's pulse directly back to the company's R&D heart.
"Amphibious Operations": Why Are Enterprises Paying Top Dollar?
Enterprises are willing to pay a steep premium for FDEs because what they are buying is an extremely rare "amphibious" capability.
· Outwardly, they are "value consultants." FDEs must be able to sit in the boardroom and articulate the ROI of AI investments in language that CEOs and CFOs understand, while also being able to stand on the factory floor discussing with veteran technicians how to integrate AI predictive maintenance into existing processes. This "translation" ability to move fluidly between business, technology, and human nature cannot be replaced by pure R&D engineers or traditional salespeople.
· Inwardly, they are "extreme builders." They must, within the customer's complex, outdated, and often "technical debt"-laden IT environment, overcome countless challenges—data silos, compatibility, security and compliance—to "squeeze" the most cutting-edge AI models into the enterprise's real world and ensure they run stably and reliably.
This "generalist" quality has led the FDE role to be seen as a "founder training camp." YC's CEO Garry Tan believes this model can help AI startups achieve an "overtaking maneuver" in competition with giants like Salesforce. The job forces practitioners to simultaneously hone hard product skills and soft business skills—real-world training that no business school can match.
Return to "Heavy Services": The Inevitable Path for AI Commercialization
The rise of the FDE marks the AI industry's shift away from the fantasy of "product-led growth" (PLG) and back to the old enterprise software path of "service-led growth."
Complex enterprise AI applications have proven that an API alone cannot win the trust of major customers with their core business. The data security, system integration, and workflow restructuring involved behind this is immensely arduous "dirty work." Therefore, AI giants have chosen a heavier, slower path—but one that builds a solid moat.
Looking back at history, giants like Salesforce and ServiceNow all faced gross margin pressure early on due to high implementation costs, but it was precisely through deep services that they became the "central nervous system" of customer operations and gained long-term pricing power.
Unlike the previous generation of software, AI is accelerating and reshaping the "heavy services" model itself. FDEs can use AI tools to automate large amounts of repetitive work, compressing integration cycles from months to weeks, making "heavy services" unprecedentedly "light" and efficient.
In the second half of the AI race, the focus has shifted from "benchmark scores" in the lab to "running the business" in the commercial world. Whoever can successfully deliver the most cutting-edge AI capabilities into the heartland of the most traditional industries will truly win this war. And the key to victory lies in the hands of that "mysterious force" of FDEs running along the "front lines" of customers around the globe. (Xiaoxiao)