In the past three years, the most expensive person in the AI industry has been model scientists.
Today, OpenAI、Anthropic、 The people Google wants to hire the most have changed.
Not a researcher, not an algorithm engineer, not even a big model expert.
But a group of people who need to travel, stay on site, attend meetings, and change procedures.
They have a new name: Forward Deployment Engineer (FDE), Front 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 model myth has officially retired, and the landing war has begun comprehensively.
The big model giants in Silicon Valley have finally realized that models are no longer a problem. Not knowing how to use it is the most difficult kilometer for enterprises. So, a job that used to be unpopular skyrocketed in value overnight.
Anthropic entry-level FDE base salary ranges from $170000 to $200000, with a total package of $300000 to $500000;
The minimum salary for OpenAI FDE positions is 210000 yuan, and both types of positions come with four-year equity.
The monthly salary of domestic ByteDance Doubao/Feishu FDE posts is 35000 – 70000, 15, and the top salary is converted into an annual salary of 1.05 million.
There are already headhunters in the market offering a special offer of $400000 per year, fully remote, to senior FDEs with two years of landing experience.
According to LinkedIn’s 2026 Workforce Report, global FDE recruitment positions will increase by 42 times from 2023 to 2025, while AI engineer positions will increase by 13 times during the same period. The former’s growth rate is about three times that of the latter.
This unconventional frenzy of snatching people has pulled down the most unspoken fig leaf of the entire AI industry in the past three years.
1、 The model was achieved, but the organization didn’t keep up
Since the birth of ChatGPT, the main line of the AI industry has been very clear. From who can make stronger models to who can make the best agents.
By 2026, the problem has changed. Enterprise customers are starting to ask another question: why hasn’t there been much change since we bought AI?
This is the biggest illusion in the entire industry, thinking that models are equivalent to productivity.
The reality is that many companies have spent a lot of money on purchasing AI/Agents, employees have registered accounts, and the IT department has made a demo of an internal knowledge base, which excited them for a month.
And then… half a year has passed and no one is using it. The way of working is the same as before.
It’s not that the employees are uncooperative, it’s not that the management lacks determination, and it’s not that the model is not good enough. The real destiny of enterprises in the production environment is never how to chat, but where is the historical data, is the format correct, and how is the quality? Which is the approval authority and who has the lead? How to import customer information, integrate ERP systems, and ensure compatibility with existing compliance and security systems?
These are not technical issues, they are organizational issues.
It’s like installing a rocket engine on a carriage. The engine is real, the thrust is real, but the horse is still a horse, the track is still dirt road, and the driver has never learned how to step on the accelerator, let alone where the emergency brake is.
Model companies have always sold in the form of tools, providing users with the strongest digital brain and encouraging them to find ways to fit it into their bodies.
However, the result is that most companies have been installing for two years, but their brains are still on the table and their bodies remain motionless.
2、 Palantir’s Legacy
The one who truly turned FDE into a profession is not OpenAI, but Palantir Technologies.
This mysterious big data unicorn, founded by Peter Thiel, the godfather of Silicon Valley, who helped the US military kill Osama bin Laden, was ridiculed in Silicon Valley for fifteen years.
The reason is that its business model is too heavy. It does not sell standardized software, but instead sends engineers to customer sites for almost half a year. VCs have labeled it as a consulting company disguised as software.
In the hierarchy of disdain in Silicon Valley, SaaS is considered advanced, while projects created by a pile of people are considered low-level. Palantir stands at the bottom of the contempt chain.
In 2011, Palantir discovered a recurring issue when selling data software to government and defense agencies, where customers were unable to use the software after purchasing it.
But it was this issue that changed everything. The traditional sales model of collecting requirements and remote development by engineers is completely ineffective in the face of highly confidential and extremely complex customers. Customers themselves don’t even know what they want, they only know that the existing things are not useful.
Palantir’s approach is not to provide better instructions, but to send their own engineers directly to the customer’s site for on-site deployment. Enter CIA, enter energy companies, enter banks. Engineers sit next to clients, observing how they work, studying data processes, understanding organizational structures, and then modifying software, processes, and even work methods.
This model has never been widely replicated in the era of standardized software. It used to be a product defined process, and if customers were not satisfied, it was due to insufficient training.
The era of big models has completely broken this logic. AI has no standard usage, and its ceiling depends entirely on how to access private data, design workflows, and implement them within organizations. The chimney system of each enterprise is completely different, and universal products cannot solve the problem of customized deep water areas.
So Palantir’s methodology, which had been accumulated for over a decade, suddenly became a textbook for the entire industry.
Today OpenAI started replicating this pattern, essentially acknowledging that AI has evolved from a software development problem to an organizational evolution problem.
3、 One month, three giants, one judgment
If Palantir only set an example for the industry, then in May 2026, the top three giants in the global AI arena simultaneously completed a collective conspiracy against application landing with real money.
On May 4th, Anthropic, in collaboration with Blackstone, Goldman Sachs, Hellman&Friedman, and multiple global asset management firms, launched a joint venture with a total committed capital of $1.5 billion. Its core business is to deploy the Claude model for the enterprise.
On May 11th, OpenAI officially announced the establishment of an independent deployment subsidiary, Deployment Company (Deploy Co), with a total initial investment of over 4 billion US dollars. The partnership includes 19 institutions, including private equity investors such as TPG and Bain Capital, as well as consulting integrators such as McKinsey and Accenture.
OpenAI synchronously acquires AI on-site consulting company Tomoro, and after the acquisition is completed, it will provide approximately 150 frontline deployment engineers to Deploy Co; Existing customers of Tomoro include TESCO Group, Virgin Atlantic Airlines, Red Bull Supercell。
Less than two weeks apart, Google Cloud CEO Thomas Kurian publicly announced on LinkedIn the large-scale recruitment of FDEs. Google Cloud has opened up over 1500 AI landing related positions internally, with FDEs as the core recruitment category.
Three of the world’s top AI companies did the same thing at the same time, not releasing stronger models, but establishing entities specifically to help businesses implement AI.
This is a more noteworthy signal than any model release.
OpenAI COO Brad Lightcap said the following:
Nowadays, AI systems targeting individuals have become very powerful, but we have not yet truly seen AI penetrate into enterprise business processes. Enterprises are complex organizations with fragmented systems, multiple compliance constraints, and complex legacy processes; The biggest challenge at present is to integrate AI into the core business processes that enterprises rely on for operation.
Simply put, the model is good enough. The problem lies within the company and organization.
It is precisely because OpenAIs have seen through this that they are willing to buy disciples of Accenture and McKinsey at any cost, upgrading them in bulk into FDEs that charge forward.
This multi billion dollar battle for talent has directly taken away the underlying assets of the traditional consulting and IT implementation industry, and has also sparked a revolution belonging to the big model delivery model.
4、 The end of selling tools is selling results
Many people believe that AI will eliminate the consulting industry. McKinsey is finished, Accenture is finished, and large IT implementers are finished.
On the contrary, AI has made consulting bigger again.
But behind it lies a deeper change, as the business model of the entire software industry is undergoing the biggest shift in the past twenty years.
License era: Customers buy software and solve problems themselves.
Subscription era (SaaS): Customers buy services and solve problems themselves.
In the era of AI, customers buy results, and suppliers are responsible for solving problems.
This is exactly the survival rule that Palantir settled down 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, and all they delivered were tools. Whether they were used well or not was up to you. What OpenAI and Anthropic are doing today is to bring their own people into the client company and deliver the results.
FDE is the result delivery specialist. Research organization, research process, research data, and finally output a system that truly runs in a production environment, rather than a beautiful demo.
In the past, consultants output PPT and FDE outputs Agent. In the past, consultants provided advice and FDE provided code. The essence is the same, helping companies solve the problem of how to work more efficiently, but the deliverables have changed.
That’s also why Anthropic’s FDE recruitment has a strange requirement: to maintain a low sense of self and a collaborative attitude.
This is the most difficult aspect of engineering culture, which requires having sufficient technical depth to solve any problem on site, while also putting down a more understanding attitude in front of customers and patiently understanding why customers do not trust the output of AI.
The annual salary ranges from 300000 to 500000 US dollars, not because FDE technology is stronger, but because a qualified FDE can replace four people: product manager, technical architect, project manager, and AI engineer.
On the front line of delivery, an FDE is like an army.
5、 The biggest obstacle to the implementation of AI has never been technology
The vast majority of AI project failures in enterprises nowadays are not technical failures, but organizational failures.
Even the world’s top financial empires and retail giants are not immune to this.
Goldman Sachs has encountered the classic mid-level compliance defense when promoting AI migration. At that time, the technical department developed an AI audit system that could automatically generate analyst reports and conduct initial review of IPO compliance documents.
But when the system was ready to connect to the production environment, the middle management of the risk control and compliance departments teamed up to press the pause button. They submitted a thick inquiry report to the management. If the illusion of the big model appears in the listing documents, who will be held responsible for the potential billions of dollars in fines?
No matter how beautiful the technical prototype is, the project was stuck for half a year due to the inability to overcome the deeply ingrained culture of accountability within the organization. It was not until the FDE team intervened and redefined the boundaries of human-machine collaboration that the project barely passed.
If Goldman Sachs is stuck in power and responsibility, then the famous setback between American retail giants Target and Palantir in the early days collided with the walls of organizational interests and culture.
At that time, Palantir dispatched a large FDE team to Target, attempting to reconstruct its supply chain and inventory forecasts with data models that generate billions of dollars in annual revenue.
However, the most powerful senior buyer team within Target strongly opposes this, believing that their decades long fashion acumen should not bow down to an algorithm. Middle level employees procrastinate on data interfaces, while frontline employees intentionally do not execute the system’s replenishment instructions. This technology cleaning, which cost millions of dollars, ultimately ended tragically with Target unilaterally tearing up the contract due to power struggles between people and machines within the organization.
I didn’t miss a single line of code, but the project just can’t be moved. This is the most realistic landing site, where technology only accounts for 20%, and the remaining 80% is all about the internal interests, power distribution, and historical burdens of the organization.
For example, the loan approval process of a bank involves decades of allocation of responsibilities and regulatory requirements. The scheduling system of a hospital is related to the interests of all departments. The quality inspection process of a factory is linked to supplier contracts and quality insurance.
These will not automatically change due to a GPT account.
These obstacles cannot be solved by an engineer who only understands technology. What is needed is someone who can think from both technical and organizational dimensions simultaneously.
So what FDE really does is not just deploy AI, the core is to help organizations complete AI migration. If the IT department has been responsible for digitizing paper-based processes in the past twenty years, then in the next ten years, FDE will be responsible for AI based digitization of these processes.
This is the next stage of the same thing.
If it’s outside of the page:
As models become increasingly affordable. Computing power is getting cheaper and cheaper. Agents are getting cheaper and cheaper.
What is truly expensive is starting to become another ability: understanding organizations, transforming processes, and driving change.
This is the reason why FDE has become popular.
It’s not that this position is important, the essence is that the entire AI industry has finally acknowledged one thing:
The most difficult part of technological revolution has never been technology.
But people.
相关文章
