By | Beyond the Page, Author — Huahua
For the past three years, the most expensive people in the AI industry were model scientists.
Today, the people OpenAI, Anthropic, and Google most want to hire have changed.
They are not researchers, not algorithm engineers, not even large language model experts.
They are a group of people who need to travel, be stationed on-site, attend meetings, and modify workflows.
They have a new name: Forward Deployment Engineer (FDE).
This is a seemingly unremarkable position, yet it likely represents the biggest shift in the AI industry over the past three years: the myth of the model officially fades, and the war for implementation has fully begun.
Silicon Valley's major model giants have finally realized that the model is no longer the problem. Enterprises not knowing how to use it is the hardest kilometer. As a result, a previously overlooked role has seen its value skyrocket overnight.
LinkedIn's 2026 Workforce Report shows that from 2023 to 2025, global FDE job postings grew by 42 times, while AI engineer postings grew by 13 times. The growth rate of the former is approximately three times that of the latter.
This unconventional, frenzied hiring craze has torn off the most unspoken fig leaf the AI industry has worn for the past three years.
I. The Model Delivered, the Organization Didn't Catch Up
Since the birth of ChatGPT, the main narrative in the AI industry has been clear: from who can build a stronger model, to who can build the best Agent.
By 2026, the question changed. Enterprise customers started asking another question: We bought AI, why haven't things changed much?
This is the industry's biggest illusion, equating models with productivity.
The reality is, many enterprises spent big money procuring AI/Agents, employees registered accounts, the IT department made a demo of an internal knowledge base, and everyone was excited for a month.
Then... six months passed, and no one is using it. The way of working remains exactly the same as before.
It's not that employees are uncooperative, or management lacks determination, or the model isn't good enough. The real critical point for enterprises in a production environment is never about how to chat, but where is the historical data, is the format correct, what's the quality? What are the approval rights and responsibilities, who has the lead? How are customer materials imported, how is the ERP system integrated, how are legacy compliance and security systems made compatible?
These are not technical problems; they are organizational problems.
It's like fitting a rocket engine to a horse-drawn carriage. The engine is real, the thrust is real, but the horses are still horses, the track is still a dirt road, and the driver has never learned how to press the accelerator, let alone where the emergency brake is.
Model companies have been selling as tools, giving users the most powerful digital brain and letting them figure out how to install it into the body.
The result, however, is that most enterprises, after two years of 'installation,' still have the brain sitting on the table, with the body completely unchanged.
II. Palantir's Legacy
The one who truly made FDE into a profession was not OpenAI, but Palantir Technologies.
Founded by Silicon Valley guru Peter Thiel, this secretive big data unicorn that once helped the U.S. military kill Osama bin Laden, was ridiculed in Silicon Valley for fifteen years.
The reason was its business model was too heavy—not selling standardized software, but sending engineers to be stationed at client sites, sitting there for six months or more. VCs labeled it: a consulting company masquerading as a software company.
In Silicon Valley's hierarchy of disdain, SaaS is superior; project-based, headcount-driven work is inferior. Palantir stood at the very bottom of this hierarchy.
In 2011, while selling data software to government and defense agencies, Palantir discovered a recurring problem: clients couldn't use the software they bought at all.
But this problem changed everything. The traditional model of sales gathering requirements and engineers developing remotely completely failed in the face of highly secretive, extremely complex clients. The clients themselves didn't even know what they wanted; they only knew what they had wasn't working well.
Palantir's approach wasn't to write better manuals. They directly sent their own engineers to be stationed at client sites. Into the CIA, into energy companies, into banks. Engineers sat next to clients, observed how they worked, studied data flows, understood organizational structures, and then modified the software, the processes, even the ways of working.
This model had never been replicated on a large scale in the era of standardized software. Previously, the product defined the process; if the client wasn't satisfied, it was due to insufficient training.
The large model era completely shattered this logic. AI has no standard usage; its ceiling depends entirely on how it accesses private data, designs workflows, and is adopted within the organization. Each enterprise's siloed systems are completely different; generic products simply cannot solve the deep-water, customized problems.
Thus, the methodology Palantir honed for over a decade suddenly became the textbook for the entire industry.
Today, OpenAI beginning to replicate this model is, in essence, acknowledging that AI has moved from a software development problem to an organizational evolution problem.
III. One Month, Three Giants, the Same Conclusion
If Palantir merely set an example for the industry, then in May 2026, the world's three top AI giants simultaneously used real money to execute a collective gambit targeting application implementation.
May 4th, Anthropic, together with Blackstone, Goldman Sachs, Hellman & Friedman, and several global asset management institutions, launched a joint venture with total committed capital of $15 billion. Its core business is enterprise deployment of the Claude large model.
Following closely, on May 11th, OpenAI announced the establishment of an independent deployment subsidiary, Deployment Company (DeployCo), with total initial cooperative investment exceeding $40 billion. The cooperative alliance comprised 19 institutions, including private investment firms like TPG and Bain Capital, as well as consulting integrators like McKinsey and Accenture.
OpenAI simultaneously acquired the AI on-site consulting firm Tomoro. Post-acquisition, Tomoro will supply DeployCo with approximately 150 forward deployment engineers. Tomoro's existing clients include Tesco, Virgin Atlantic, Red Bull, and Supercell.
Less than two weeks apart, Google Cloud CEO Thomas Kurian publicly announced on LinkedIn a large-scale recruitment drive for FDEs. Google Cloud internally opened over 1,500 AI implementation-related positions, with FDE being a core recruitment category.
Three of the world's top AI companies did the same thing at the same time—not releasing a stronger model, but establishing entities specifically to help enterprises implement AI.
This is a more noteworthy signal than any model release.
OpenAI COO Brad Lightcap even said this:
"AI systems for individuals are already very powerful today, but we haven't truly seen AI penetrate enterprise business processes. Enterprises are structurally complex organizations with fragmented systems, numerous compliance constraints, and cumbersome legacy processes. The biggest challenge now is integrating AI into the core business processes that enterprises rely on to operate."
Simply put, models are good enough. The problem lies inside companies and organizations.
It is precisely because they have seen through this that OpenAI and others are willing to pay any price to acquire the disciples of Accenture and McKinsey, upgrading them en masse into the FDEs charging into battle.
This multi-billion dollar talent battle is directly siphoning off the foundational assets of the traditional consulting and IT implementation industries, also marking the beginning of a revolution in large model delivery models.
IV. The End of Selling Tools is Selling Outcomes
Many thought AI would eliminate the consulting industry. McKinsey is finished, Accenture is finished, large IT implementers are finished.
The result is quite the opposite; AI has made consulting bigger again.
But hidden behind this is an even deeper change: the business model of the entire software industry is undergoing its biggest shift in the past twenty years.
This is precisely the survival rule Palantir honed over a decade ago: Don’t sell software. Deploy outcomes.
This is a fundamental transformation. In the past, Microsoft sold Office, Salesforce sold CRM, Adobe sold suites—what they delivered were tools. How well you used them was your business. What OpenAI and Anthropic are doing today is putting their own people inside client companies to deliver the outcomes.
FDEs are outcome delivery agents. They study the organization, study the processes, study the data, and finally output a system that truly runs in the production environment, not just a beautiful demo.
In the past, consultants delivered PowerPoints; FDEs deliver Agents. In the past, consultants gave advice; FDEs give code. The essence is the same—helping enterprises solve the problem of how to work more efficiently—only the deliverable has changed.
This is also why there's a strange requirement in Anthropic's FDE recruitment: maintain low ego and a collaborative attitude.
This is one of the hardest principles in engineering culture: having enough technical depth to solve any problem on-site, while also adopting a humble posture in front of the client to patiently understand why they might distrust the AI's output.
A salary of $300,000 to $500,000 is not because FDEs are technically stronger; it's because a qualified FDE can replace four roles: product manager, technical architect, project manager, and AI engineer.
On the delivery frontlines, one FDE is an army.
V. The Biggest Obstacle to AI Implementation Has Never Been Technology
Most AI project failures in enterprises today are not technical failures; they are organizational failures.
This is something even the world's top financial empires and retail giants cannot escape.
Goldman Sachs Group encountered classic middle-management compliance defense when pushing for AI migration. The technology department had developed an AI auditing system capable of automatically generating analyst reports and performing initial reviews of IPO compliance documents.
But when the system was ready to be connected to the production environment, middle managers from the risk control and compliance departments jointly pressed the pause button. They submitted thick query reports to management: if the large model's "hallucinations" appeared in listing documents, who would be responsible for potential fines amounting to tens of billions of dollars?
No matter how beautiful the technical prototype, the project was stalled for half a year, unable to cross the deep-rooted culture of liability avoidance within the organization, until the FDE team intervened to redraw the boundaries of human-machine collaboration authority and responsibility, barely passing through.
If Goldman Sachs got stuck on authority, then the famous early debacle between American retail giant Target and Palantir crashed into the wall of organizational interests and culture.
At the time, Palantir sent a large FDE team into Target, attempting to use data models to reconstruct its tens-of-billions-of-dollars annual revenue supply chain and inventory forecasting.
However, Target's most powerful internal team—the veteran buyers—was extremely resistant. They believed their decades of fashion intuition should not bow to an algorithm. Middle management dragged their feet on data interfaces, while frontline employees deliberately ignored the system's replenishment instructions. This multimillion-dollar technological overhaul ultimately ended in Target unilaterally tearing up the contract, a casualty of the power struggle between people and machines within the organization.
Not a single line of code was wrong, but the project couldn't move. This is the most realistic implementation scene: technology accounts for only 20%; the remaining 80% is all about the internal power structure, responsibility distribution, and historical baggage of the organization.
For example, a bank's loan approval process is backed by decades of authority distribution and regulatory requirements. A hospital's scheduling system is linked to the interest structures of all departments. A factory's quality inspection process connects to supplier contracts and quality insurance.
These won't automatically change just because of a GPT account.
These obstacles cannot be solved by an engineer who only understands technology. What's needed are people who can think on both the technical and organizational dimensions simultaneously.
So what FDEs really do is not just deploying AI; the core is helping organizations complete AI migration. If for the past twenty years, IT departments were responsible for digitizing paper-based processes, then for the next ten years, FDEs will be responsible for AI-ifying those digitized processes.
This is the next stage of the same undertaking.
A Word from [Beyond the Page]:
As models become cheaper. Computing power becomes cheaper. Agents become cheaper.
What truly becomes expensive starts to be another capability: understanding the organization, transforming processes, driving change.
This is why FDEs are hot.
It's not that the position itself is so important; the essence is that the entire AI industry has finally admitted one thing:
The hardest part of a technological revolution has never been the technology.
It's people.
