Summary of Key Points
The FDE (Forward Deployed Engineering) model originated with the American company Palantir, which involves sending engineers to the customer's site to understand business challenges, customize solutions, and then incorporate that experience into products for scaled profitability. However, this model does not work in China for several reasons: customers are unwilling to pay for the exploration process, manufacturers cannot make a profit, there is a scarcity of specialized talent, AI companies are eager to generate revenue rather than invest in long-term development, and there is resistance to AI within companies. As a result, FDE in China has essentially become a form of "high-end outsourcing with a different name."
1. Customers: "I Only Want Finished Products; Exploration Is Your Problem"
The core of the FDE model is to understand the business first before customizing solutions, but Chinese customers have a completely different mindset:
- In the U.S.: The process is valuable. American companies and governments are accustomed to charging by the hour (for example, Accenture charges per man-day), and Palantir's on-site time can be included in the contract for payment. Government procurement even allows paying a premium for better technology (for instance, Palantir won a large contract after suing the Army).
- In China: Results matter most. Customers sign contracts for ready-to-use products, and understanding the business or making trial-and-error iterations is considered the vendor's cost. During bidding, only the feature list and the lowest price are considered (72% of government projects in Guangdong are won at the lowest bid, sometimes below cost). For example, when a hospital purchases a system, three bids were 10 million, 8 million, and 5 million—although the 5-million bid was invalid, it reflects the prevailing trend of "low price wins."
- Misalignment in expectations: FDE believes it takes three months to understand the business, but customers think three days are enough; FDE wants to iterate and improve, while customers demand a perfect first version. This contradiction kills the viability of the FDE model.
2. Manufacturers: "Can't Make a Profit with FDE"
Palantir took 20 years to become profitable, relying on high-profit government contracts and a subscription-based model (with regular renewals from existing clients). However, Chinese manufacturers do not have these advantages:
- Profitability model does not apply: Chinese companies prefer a one-time purchase (the SaaS penetration rate is only 15.8%), and customers are not seen as long-term assets. The cost of having FDE on-site for three months is recorded as an expense with no corresponding revenue, making it difficult for sales managers to justify and for finance departments to approve the budget.
- Capital market constraints: Chinese investors expect companies to profit quickly; no one can afford to sustain a FDE team for 20 years. Companies like UFIDA and Kingdee are still losing money and cannot afford to pay engineers with salaries in the hundreds of thousands of dollars (in China, implementation positions earn only 5,000-8,000 RMB per month).
- Regulatory restrictions: The Government Procurement Law requires that projects with unified technical standards be awarded to the lowest bidder, forcing manufacturers to compete on price. With thin margins, they cannot afford to support FDE teams.
3. Talent: "Specialized FDE Professionals Do Not Exist in China"
FDE requires a combination of technical skills, industry knowledge, and business acumen (for example, a logistics FDE needs to understand logistics processes, code, and interact with customers), but such talent is scarce in China:
- Low professional status: In the Chinese B2B hierarchy, on-site support is considered low-level (research > algorithms > product development > pre-sales > implementation). The ultimate goal for professionals is to move to a higher position within the company, so few are willing to work in this role for long.
- Unattractive salaries: The median annual salary for FDE professionals in the U.S. is 210,000 USD, with senior roles reaching up to 800,000 USD; in China, new graduates in implementation positions earn only 100,000-150,000 RMB per year, and even more experienced professionals at UFIDA earn 5,000-8,000 RMB per month, which is much lower than in larger companies (300,000-550,000 RMB). The most talented young people prefer to work in algorithm development.
- Lack of career pathways: FDE roles at American companies like Palantir serve as a stepping stone to entrepreneurship (with over 100 startups founded and 11.6 billion USD in funding), but in China, top founders often come from large companies' R&D departments, and having worked for a vendor is not seen as a valuable asset.
4. AI Manufacturers: "Eager to Generate Revenue, Lack Patience"
The primary motivation for AI manufacturers to adopt FDE is to quickly sell their models, but FDE requires long-term investment:
- Desire for quick profit: Companies like OpenAI lose 17 billion USD annually, and Anthropic spends 1.25 billion USD on computing resources each month; both want to sell their models immediately. However, FDE requires several months to years of on-site work, which contradicts their goal of rapid revenue generation.
- Mismatch in capabilities: AI companies are focused on model research, not on implementation and refinement. Palantir spent 20 years turning field experience into products; AI manufacturers have not even developed good models and lack the resources for such detailed work.
- Unrealistic promises: Companies like OpenAI promise investors annual returns of 17.5%, but FDE models rely on long-term subscriptions, which are unachievable in the short term—this is more of a marketing gimmick.
5. Internal Company Resistance: "Bosses Don't Believe, Employees Resist"
Even if FDE teams are on-site, they face resistance from within the company:
- Bosses want to replace employees: Many use AI to cut costs, not to transform their organizations. FDE requires changes in processes and knowledge transfer, but bosses prefer a top-down approach, believing that AI is just a tool to be used.
- Employee resistance: 29% of employees oppose AI strategies (with Generation Z accounting for 44% of the workforce), fearing that AI will take their jobs. They prefer to use personal tools like ChatGPT rather than contribute to company systems, fearing being criticized as lazy.
- High failure rate of pilots: A NANDA report shows that 95% of AI initiatives fail to deliver measurable benefits, mainly due to organizational resistance. Even with FDE teams on-site, these issues cannot be resolved.
In Conclusion: FDE in China Is Just "Outsourcing under a Different Name"
Palantir succeeded with its FDE model thanks to the U.S. business environment (hourly billing, subscription models, and talent pathways). However, China lacks these conditions: customers are unwilling to pay for the process, manufacturers cannot make a profit, there is a shortage of talent, and there is resistance within companies. FDE in China is merely a form of outsourcing with a fancy name. Unless these fundamental issues are addressed, it will not become a successful model.
"We don't lack new terms (such as central platforms or the metaverse) but rather the patience to allow things to develop slowly and the willingness to invest in exploration."