Annual Salary of 1.7 Million: This AI Role Suddenly Became Fiercely Contested

Recently, an AI role called FDE has suddenly taken off.

On recruitment platforms, ByteDance's "Doubao AI Large Model FDE" offers up to 70,000 RMB per month with 15 months' pay; Ant Digital's B2B FDE offers up to 60,000 RMB per month; Zhipu's FDE lead offers up to 80,000 RMB per month.

In Silicon Valley, the median annual salary for FDE-related roles exceeds $250,000 (close to 1.7 million RMB), with companies like OpenAI, Anthropic, and Salesforce investing heavily.

FDE, short for Forward Deployed Engineer, refers to people who go deep into customer sites and are responsible for integrating AI technology into enterprise business processes. They not only write code and build systems, but also communicate with bosses and employees, turning a vague request like "we want to use AI" into a workflow that actually runs.

But there is also considerable controversy about this role. Some say it is the most valuable new profession of the AI era, while others believe it is nothing more than "premium outsourcing."

Founded recently, valued at 94.5 billion

FDE is not a role that just appeared.

As early as around 2010, the American data company Palantir established this role. After taking orders from the military and intelligence agencies, Palantir sent engineers directly to customer sites — even to the front lines in Iraq and Afghanistan. When a soldier said "this road looks suspicious," the engineer would turn it into an annotation tool on the map on the spot, and useful features would then be folded back into the platform. To this day, about half of Palantir's employees are in FDE roles.

So why did FDE suddenly take off in 2026?

Zhan Bingqiang, founder of AIGCLINK, believes "the hype is actually the bigger factor." He told Qianbidao that the FDE group has existed for years — it just wasn't called that before, but rather "premium outsourcing." There's even a joke in the industry: FDEs call themselves "premium outsourcing," while OPCs call themselves "freelancers."

But behind the hype, there is indeed genuine demand driving it.

The core reason is: the AI industry has moved from "selling models" to "selling results."

Over the past two years, enterprises buying models and AI agents became all the rage, but many projects stalled at the demo stage and hit all kinds of walls once they entered production — legacy systems that couldn't connect, data scattered across WeChat and Excel, bosses who couldn't articulate specific requirements, and compliance and permission hurdles. Large models are standardized products, but enterprises' data, processes, and systems vary enormously. Someone has to be on-site to bridge the "last mile" from demo to something actually usable.

In a report by Dingjiao, an FDE named Lawted walked into the office of a freight forwarding company in Shenzhen and saw more than thirty employees all repeating the same task: constantly opening shipping PDFs sent by customers via WeChat and email, finding the port and sailing schedule, replying with confirmation, and then manually entering it into the system. He single-handedly automated this entire process using AI tools.

This is the value of an FDE. In Zhan Bingqiang's words: "In the past, when people went into enterprises to implement AI, there was never a particularly clear identity. Now that FDE exists, it's like giving a clearer identity to this group of people who help enterprises with AI transformation."

The moves of overseas giants have further fueled FDE's popularity.

In May 2026, OpenAI announced the establishment of The Deployment Company, valued at $14 billion (about 94.5 billion RMB), with initial funding exceeding $4 billion, while acquiring the UK consulting firm Tomoro (about 150 seasoned FDEs) as the cornerstone of its deployment team.

Almost at the same time, Anthropic, together with Blackstone and Goldman Sachs, established a $1.5 billion joint venture for enterprise AI services, focusing on implementation in heavily regulated industries like finance and healthcare. Salesforce publicly recruited 1,000 FDEs to serve its Agentforce product line. AWS, the world's largest cloud computing company, is preparing to invest $1 billion to build an FDE team of several thousand people.

Compared with industries like finance and the internet, manufacturing — with thin margins and difficult transformation — is precisely the "main battlefield" where AI implementation has the greatest room for efficiency gains and where FDE value is most evident. Frontline FDEs' practices in industrial manufacturing, construction engineering, cross-border e-commerce, and other fields have already proven: once enterprises see the efficiency leap with their own eyes, their attitude shifts from wait-and-see to actively seeking more AI transformation opportunities.

Palantir AIP product interface Source: gateex.com

There is also policy-level push domestically. Beijing and other places have incorporated concepts like OPC, FDE, and Token into relevant government programs, pushing an already warming topic completely into the mainstream. Li Yachong, founder of Grapevine Technology, also observed: "Because AI genuinely has implementation needs, no matter what this role is called, the underlying logic is that enterprises will increase their use of AI. The name doesn't affect the demand itself."

With two forces combined — on one side, the real pain points of enterprise AI implementation, and on the other, the push from capital and policy — FDE has gone from a niche engineering role to a hot profession discussed by everyone.

A new opportunity for IT outsourcing companies

Zhan Bingqiang has a metaphor: if we see the past information age as the agricultural age, then AI is like the industrial age. Today's FDEs are a group of "people driving harvesters to farm."

In the past, writing a piece of software took ten people several months; now a few people with Claude, Codex, and AI agents might be able to get it done.

"In the past, on-site development meant writing code yourself; now AI writes the code, or you help clients use AI." Li Yachong also believes the essence hasn't changed — only the tools have.

Indeed, when no one else has a harvester yet, the person who can operate one is naturally extremely valuable.

"But this window period is actually quite short — probably until around the first half of next year, and then the opportunities will get narrower and narrower," Zhan Bingqiang said.

This also leaves a window open for traditional IT service providers. "FDE is actually the easiest direction for practitioners in various industries to transition into," Zhan Bingqiang said.

Traditional IT outsourcing personnel already know how to deal with clients, know how to deliver projects, and are familiar with enterprises' legacy systems. If they then add AI capabilities and transition into FDEs, that is far more realistic than having a pure model researcher relearn the realities of factories, finance, or ERP.

Anthropic is taking this path to the extreme.

Claude by Anthropic Source: winbuzzer.com

In June this year, it announced a partnership with the established IT service provider DXC. DXC plans to train tens of thousands of Claude-certified FDEs and then place these people in banks, airlines, insurance companies, manufacturing enterprises, and government agencies.

DXC itself has 115,000 employees and has long operated critical systems for these clients. Before partnering with Anthropic, it had already involved Claude in developing its own AI-native operations platform OASIS, with over 95% of the code generated by Claude.

When 95% of code can be generated by AI, it hasn't made some far-sighted IT service companies disappear — on the contrary, they have begun training FDEs at scale, because what clients buy has never been just the code.

Zhan Bingqiang believes this is the most noteworthy change in the FDE business model.

Traditional outsourcing likes to "sell people" — ten people working for half a year means charging for ten people's half-year. In the future, FDEs may see two completely different pricing models.

The first is helping an enterprise deploy an "AI employee."

The FDE enters the enterprise, sets up the AI employee, and teaches the client's own staff how to use it — equivalent to "helping them onto the horse, then seeing them off a stretch."

The second is more radical: selling results directly.

"You don't need to worry about what the process in between looks like," Zhan Bingqiang said by way of example.

Suppose an enterprise used to spend 30,000 RMB to complete a task; now an FDE with their own AI employee can do it for 10,000 RMB — then the client only buys the final result.

As for whether it was done by three people, or by one person plus ten AI agents, the client doesn't care.

The next step: some companies may directly sell a "completed task."

This is also why Zhan Bingqiang believes small and mid-sized IT companies must transform quickly. "In the future, every boss must themselves be an FDE."

If the owner of a software or AI company doesn't even know what AI can currently achieve for the business, and keeps operating under the old headcount- and project-based logic, he believes such companies will sooner or later lose their competitiveness.

Going one step further, companies should become AI Native; at the very least, managers themselves must be capable of using AI to reorganize delivery.

Business skills matter more than technical skills

"I have a fairly certain conclusion," he said. "In the future, FDE will be a guaranteed role at AI companies. The internet era gave rise to roles like product manager, operations, and engineering; in the AI Native companies of the future, FDE will likewise be a role that is certain to emerge, just like those."

His logic is this: no matter how AI agents develop, enterprises undergoing AI transformation need customization tailored to specific scenarios, which naturally requires a group of people doing this "dirty, tiring work" on the front lines. AI Native companies can focus on models and data, but reaching customers on the ground and completing the last mile of work still has to be handed to FDEs.

But he believes Vibe Coding (AI-assisted programming) ultimately won't become a real barrier to entry. "Slowly, everyone will probably be able to do it."

What truly sets people apart are two things: industry expertise and business skills.

Zhan Bingqiang listed the capabilities a qualified FDE needs: first, industry know-how; second, the ability to do vibe coding; third, customer-facing delivery and implementation skills; and fourth, a certain degree of business acumen.

"In the end, when it comes down to it," Zhan Bingqiang said, "what FDEs really compete on is two capabilities: one is expertise—that is, how deep your industry know-how actually is; the other is business skills."

OpenAI's recent hiring changes already show signs of this trend.

It isn't just hiring generalist FDEs; it has begun subdividing into fields like healthcare, life sciences, and government.

Healthcare FDEs need to understand hospitals, insurance, electronic medical records, and the regulatory environment; the life sciences team works directly with pharmaceutical companies, biotech firms, and CROs, participating in drug discovery and R&D processes; government roles even require the corresponding security clearances.

FDEs are looking less and less like a technical job category and more like "industry expert + engineer."

This also means that not everyone flooding into FDE roles today will land a million-yuan salary.

The Chinese hiring market has already become stratified. A survey by Jiemian News found that some ordinary FDEs earn a monthly salary of around 20,000 to 30,000 yuan, while annual packages for senior-level personnel mostly exceed 400,000 yuan; the high-paying positions offered by ByteDance, Ant Group, and Zhipu are concentrated mainly on senior talent.

The rise of the FDE, on the surface, is the birth of a new profession; what it reflects underneath is actually an important pivot in the AI industry.

How far this role can go ultimately depends on how deeply AI gets implemented. If AI truly permeates every industry and every enterprise the way electricity did, then demand for FDEs will only grow—just as the electrification era needed large numbers of electricians, and the information age needed large numbers of programmers.

The people who will truly remain in the future are those who understand technology, understand business, and can also build strong rapport with customers.

This article comes from the WeChat public account "Qianbidao" (ID: pencilnews), author: Qianbidao, published by 36Kr with authorization.