The AI industry shifts focus to deployment engineers as model hype fades.

Text | Beyond the Layout, Author | Hua Hua

Over the past three years, the most expensive professionals in the AI industry have been model scientists.

Today, the people OpenAI, Anthropic, and Google most want to hire have changed.

Not a researcher, not an algorithm engineer, and not even a large model expert.

But rather a group of people who need to travel for business, work on-site, attend meetings, and adjust processes.

They have a new name: Forward Deployment Engineer (FDE), Frontline Deployment Engineer.

This may seem like an unremarkable position, but it could represent the biggest shift in the AI industry over the past three years: the myth of models has officially ended, and the battle for real-world deployment has fully begun.

The big model giants in Silicon Valley have finally realized that models are no longer the issue—it’s that enterprises don’t know how to use them, which is the hardest final mile. As a result, a once-overlooked role has seen its value skyrocket overnight.

The LinkedIn 2026 Workforce Report shows that global FDE job postings increased 42-fold between 2023 and 2025, while AI engineer postings rose 13-fold during the same period—nearly three times the growth rate of the latter.

This frenzy of poaching talent has stripped away the most unspoken excuse the AI industry has relied on over the past three years.

Since the birth of ChatGPT, the main focus of the AI industry has been clear: shifting from who can build the strongest models to who can create the best agents.

By 2026, the question had changed. Enterprise customers began asking another question: We bought AI, but why has there been so little change?

This is the biggest illusion in the entire industry—that equating models with productivity.

The reality is that many companies spent a lot of money purchasing AI/Agent tools, employees registered for accounts, and the IT department built a demo of an internal knowledge base—only to be excited for a month.

Then... six months passed, and no one used it. The way of working remained exactly the same.

It’s not that employees aren’t cooperating, management lacks determination, or the model isn’t good enough. The real bottleneck for enterprises in production environments has never been how to chat—it’s about where the historical data is, whether its format is correct, and how good its quality is; which paths govern approval authority and who holds primary control; how customer data is imported, how ERP systems are integrated, and how existing compliance and security systems are made compatible.

These are not technical issues; they are organizational issues.

It’s like installing a rocket engine on a horse-drawn carriage. The engine is real, the thrust is real, but the horse is still a horse, 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.

The model company has always sold its products as tools, providing users with the most powerful digital brain and leaving it to them to figure out how to integrate it into their body.

Yet, most companies have had the system in place for two years, but their minds are still on the table, and their bodies haven’t moved an inch.

The company that truly turned FDE into a profession is Palantir Technologies, not OpenAI.

This mysterious big data unicorn, founded by Silicon Valley icon Peter Thiel and once instrumental in helping the U.S. military kill Bin Laden, was mocked in Silicon Valley for fifteen years.

The reason is that its business model is too heavy—it doesn’t sell standardized software but instead sends engineers to work on-site at clients’ locations for months on end. VCs labeled it as a consulting firm wrapped in software packaging.

In Silicon Valley’s hierarchy, SaaS is considered high-end, while projects built on sheer manpower are seen as low-end. Palantir sits at the very bottom of this hierarchy.

In 2011, Palantir discovered a recurring issue while selling data software to government and defense agencies: customers purchased the software but never used it.

But it was this very issue that changed everything. The traditional sales model—collecting requirements and having engineers develop remotely—completely failed in the face of highly secretive and extremely complex clients. The clients themselves didn’t even know what they wanted; they only knew that what they had wasn’t working.

Palantir’s approach isn’t to write better manuals—it’s to send its own engineers directly to client sites. They embed themselves within the CIA, energy companies, and banks. Engineers sit beside clients, observe how they work, analyze data flows, understand organizational structures, and then modify software, processes, and even workflows.

This model was never widely replicated in the era of standardized software, where if customers were dissatisfied, it was assumed to be a training issue.

The era of large models has completely shattered this logic: AI has no standard usage; its ceiling is entirely determined by how private data is integrated, workflows are designed, and adoption is driven within an organization. Every company’s siloed systems are unique, and generic products simply cannot solve the deep, customized challenges.

Thus, Palantir’s decades-long methodology suddenly became the industry textbook.

Today, OpenAI began replicating this model, essentially acknowledging that AI has shifted from a software development challenge to an organizational evolution problem.

If Palantir merely set an example for the industry, then in May 2026, the world's top three AI giants simultaneously invested real capital in a coordinated effort to drive practical applications.

On May 4, Anthropic, in partnership with Blackstone, Goldman Sachs, Hellman & Friedman, and several global asset management firms, launched a joint venture with $1.5 billion in committed capital, focused on deploying Claude large models for enterprise use.

Immediately following on May 11, OpenAI officially announced the formation of an independent deployment subsidiary, Deployment Company (DeployCo), with a combined initial investment exceeding $4 billion and a coalition of 19 institutions, including private equity investors such as TPG and Bain Capital, as well as consulting and integration firms like McKinsey and Accenture.

OpenAI has concurrently acquired the AI on-site consulting firm Tomoro; after the acquisition, approximately 150 frontline deployment engineers will be transferred to DeployCo; Tomoro’s existing clients include Tesco, Virgin Atlantic, Red Bull, and Supercell.

Less than two weeks apart, Google Cloud CEO Thomas Kurian publicly posted on LinkedIn to aggressively recruit FDEs, with over 1,500 internal roles related to AI implementation opening at Google Cloud, with FDEs as the core hiring category.

The three world’s top AI companies simultaneously did the same thing—not releasing a more powerful model, but establishing dedicated entities to help businesses implement AI.

This is a more significant signal than any model release.

OpenAI COO Brad Lightcap even said the following:

AI systems designed for individuals have become highly capable today, but we have not yet truly seen AI penetrate enterprise business processes. Enterprises are complex organizations with fragmented systems, numerous compliance constraints, and legacy workflows; the greatest challenge today is integrating AI into the core business processes upon which enterprises rely.

Simply put, the model is good enough. The issue lies within the company and organization.

Precisely because they understand this, OpenAI and others are willing to go to any lengths to acquire disciples from Accenture and McKinsey, and systematically upgrade them into frontline FDEs.

This multi-billion-dollar talent war has directly drained the foundational assets of the traditional consulting and IT implementation industries, ushering in a revolution in large model delivery models.

Many people believe AI will eliminate the consulting industry. McKinsey is done, Accenture is done, large IT implementation firms are done.

On the contrary, AI has revitalized consulting.

But behind it lies a deeper shift: the software industry's business model is undergoing its largest transformation in the past two decades.

This is precisely the survival principle Palantir developed over a decade ago: Don’t sell software. Deploy outcomes.

This is a fundamental shift. In the past, Microsoft sold Office, Salesforce sold CRM, and Adobe sold suites—delivering tools, and it was up to you to use them well. Today, OpenAI and Anthropic are sending their own teams into clients’ companies to deliver results.

FDE is the Final Delivery Engineer. They research organizations, study processes, analyze data, and ultimately deliver a system that runs in production—not just a polished demo.

Previously, consultants delivered PowerPoint presentations; FDE delivers agents. Previously, consultants provided recommendations; FDE provides code. The essence is the same—helping businesses solve the problem of working more efficiently—only the deliverables have changed.

This is also why Anthropic's FDE job posting includes an unusual requirement: maintain a low sense of self and a collaborative attitude.

This is the most challenging aspect of engineering culture: having sufficient technical depth to resolve any issue on-site, while also setting aside the need to appear more knowledgeable than the client, and patiently understanding why the client distrusts AI outputs.

An annual salary of $300,000 to $500,000 is not because FDE technology is superior, but because a qualified FDE can replace a product manager, a technical architect, a project manager, and an AI engineer.

On the front lines, an FDE is an army.

Most AI projects in enterprises fail today not due to technical failure, but due to organizational failure.

Even the world's most powerful financial empires and retail giants cannot escape this.

Goldman Sachs encountered classic middle-management compliance resistance when advancing its AI migration. The technology team had developed an AI audit system capable of automatically generating analyst reports and conducting initial reviews of IPO compliance documents.

But when the system was preparing to go live, mid-level executives from the risk management and compliance departments jointly hit the pause button. They submitted a lengthy report of concerns to management, asking: who would be held accountable for potential billions of dollars in fines if the large model’s hallucinations appeared in official filings?

Even the most impressive technical prototype was stalled for six months due to deeply entrenched blame-avoidance culture within the organization, until the FDE team stepped in to redefine the boundaries of human-machine collaboration, barely clearing the hurdle.

If Goldman Sachs was stalled by accountability issues, then the U.S. retail giant Target’s early, well-known clash with Palantir hit a wall of organizational interests and culture.

At the time, Palantir deployed a large FDE team to Target, attempting to reconstruct its supply chain and inventory forecasting—generating hundreds of billions in annual revenue—using data models.

However, the senior buying team, the most powerful internal force at Target, strongly resisted it, believing their decades of fashion intuition should not yield to an algorithm. Middle management delayed data integration at every turn, and frontline staff deliberately ignored the system’s restocking instructions. This multi-million-dollar technological overhaul ultimately ended in a bitter collapse, as Target unilaterally terminated the contract amid an internal power struggle between people and machines.

The code was flawless, but the project still wouldn’t move. This is the real-world scenario: technology accounts for only 20%; the remaining 80% is all about internal organizational interests, responsibility allocation, and historical baggage.

For example, a bank’s loan approval process is built on decades of allocated responsibilities and regulatory requirements. A hospital’s scheduling system is tied to the interests of all departments. A factory’s quality inspection process is linked to supplier contracts and quality insurance.

These will not change automatically due to a single GPT account.

These challenges cannot be solved by an engineer who understands only technology. What’s needed is someone who can think simultaneously across both the technical and organizational dimensions.

So what FDE truly does is not just deploying AI; its core mission is helping organizations complete their AI migration. If, over the past two decades, IT departments were responsible for digitizing paper-based processes, then over the next decade, FDE will be responsible for AI-izing those digital processes.

This is the next stage of the same matter.

[Outside the layout]:

As models become cheaper, computing power becomes cheaper, and agents become cheaper.

What’s truly expensive is beginning to become another skill: understanding organizations, transforming processes, and driving change.

This is why FDE has become so popular.

It’s not that this position is particularly important; the essence is that the entire AI industry has finally acknowledged one thing:

The hardest part of a technological revolution has never been the technology.

But people.