Out of 1000 resumes submitted for a single recruitment opening, what kind of FDE are they looking for?

Stark became Iron Man not because an ordinary person put on the armor and then became Iron Man. It is because he was determined to become Iron Man that he was worthy of that suit of armor.

Gan Yifan said a sentence that made me stop and think for a while.

"FDE is a result-oriented position, not a process-oriented one."

When I interviewed six other companies earlier, the topic that most people talked about the most was what FDE "does": some ask FDE to build Agents, some arrange them to work on-site to co-create with clients, and others ask them to collect requirements and design solutions.

But Gan Yifan raised a more fundamental question: What results should all these work ultimately point to? What responsibilities should FDE take for these results?

Gan Yifan and Liu Kai are co-founders of Rolling AI, and both used to work at BCG Digital Ventures. One has a design background, and the other has an engineering background. They are probably the most "rigorous" group of interviewees so far. The two of them are not only strict about how FDE works, but also more rigorous about what FDE essentially is.

Three Common Misinterpretations

Gan Yifan split the three letters of FDE separately, and explained the three most common misinterpretations of this role.

The first misinterpretation is the understanding of E (Engineer).

"Most people in China have a misinterpretation of the word 'engineer'. People think that an engineer is someone who can write code, but that's not true. What an engineer really needs to do is to turn a thing into reusable capabilities, and then make it run systematically in the organization through engineering methods."

Gan Yifan said that he can hardly understand code longer than 10 lines, but he ranks very high in the company in terms of total lines of code written. This seems contradictory, but behind it is the fact that AI has changed the way "engineering capabilities" are realized: A person does not necessarily need to write every line of code by himself, but he needs to know how to use AI to turn business requirements into a running system.

Liu Kai added a remark beside him: "Taking the R&D level of an ordinary software company as a reference, the code produced by Gan Yifan now has higher quality than the average level of the R&D team. To be precise, it is not that Gan Yifan's personal coding level exceeds the R&D team, but that the code generated with the help of AI has quality higher than the team's average level."

The second misinterpretation is the understanding of D (Deployment).

Liu Kai put it very directly: "People usually think that Deployment is to deploy code to servers. But in our view, it refers to the process in which a new set of productivity enters the enterprise and actually starts to operate."

Deploying code to the server is only the launch at the technical level. For AI to actually generate results in the enterprise, it also needs to enter specific business processes, collaborate with the original organization, and align with business objectives.

Gan Yifan said that Deployment in the AI era involves a large amount of business work. In actual implementation, most obstacles do not come from technology, but from organizational collaboration, business processes, and the lack of real alignment of objectives.

The third misinterpretation is the understanding of F (Front End).

The Front End here does not refer to the "front end" in software development, but the real front line where the customer's business takes place.

"There are two definitions of the front line. The front line we refer to is the real business site. For example, if the customer's business takes place in physical stores, we need to go to the stores instead of only staying at the customer's headquarters. The headquarters is not the real Front End."

The three words correspond to three misinterpretations: Engineer is not just writing code, Deployment is not just deploying systems, and Front End is not just sitting at the customer's headquarters to align requirements. Together they point to the essence of FDE: FDE is not a role that completes technical delivery actions, but a role that dives deep into business sites and promotes AI to generate real results.

Beyond Capabilities, Mental Strength and Aspiration Matter More

I asked them how they recruit talents. Gan Yifan gave three words: smart, diligent and "hungry".

"Being 'hungry' is actually a joke in a way. What I mean is that the person should have a sense of hunger when seeing an opportunity, and have the drive to move forward. This original motivation comes from inner desire."

Liu Kai gave a number: for every 1000 resumes they receive, they finally recruit 1 person. There is also a round of AI interview before manual screening.

1000:1. This ratio is the highest among all the companies I have interviewed so far.

But what Gan Yifan really values is not just what capabilities the candidate currently has. He took himself as an example: he has a design background and "extremely poor" coding ability, but he has enough mental strength to do things that he was not good at at first.

"Capabilities are one level, and if you look deeper, it is actually mental strength and aspiration. It does not matter if a person has a small shortboard in capabilities. As long as the general direction matches and he has sufficient mental strength, we are willing to hire him."

This made me think for a long time. Previous interviewees were all talking about "capabilities", including strategic consulting, human-AI collaboration, and hands-on engineering skills. But Gan Yifan dug two more layers below capabilities: Mental strength means whether you dare to do it, and aspiration means whether you want to do it. Capabilities can be supplemented by AI, but mental strength cannot.

What kind of people can hardly stay in the team? Gan Yifan answered very simply:

"As long as a person has too big an ego, he cannot stay in our team, there is no way around it."

Because in the work environment with AI involved, people need to keep trying and making mistakes, learn quickly, and also accept their own capability boundaries. If a person cares too much about proving himself, it is difficult for him to truly integrate into such a collaboration mode.

Gan Yifan also mentioned a preference for employing people that surprised me: he thinks people whose families run small businesses usually "have better common sense". They were exposed to real business and human relationships earlier, and are more used to facing problems on their own and finding solutions. On the contrary, those who have always been very obedient and only good at studying may not adapt well to such an environment.

If Humans Cannot Figure Things Out, How Can AI Do That?

I followed up with a question that many people care about: Can Palantir's FDE model be replicated in China?

Gan Yifan gave a clear answer: No.

In his opinion, there will only be one Palantir in the world. First of all, Palantir serves a group of clients with the strongest payment capacity in the world. These clients have huge budgets (more than one trillion US dollars per year) and can invest regardless of cost when solving key problems. More importantly, the underlying logic of enterprise operation in China and the United States is different, so Palantir's model can hardly be moved to China as it is.

"The underlying of commercial operation in the United States is the contract system, and enterprises operate according to contract texts and job responsibilities. That is not the case in China. Many Chinese enterprises rely on personal connections and trust for management, and important matters are often concentrated on a small number of people they trust, and positions may also be set according to specific people."

This brings a very practical problem to AI implementation: the responsibilities, processes and rules within the enterprise have not been clearly defined. Who is responsible for many things, how to collaborate, and who to turn to when problems occur, all rely on long-formed tacit understanding rather than clear systems and processes.

Gan Yifan continued:

"If humans cannot figure things out, how can AI do that? If even humans cannot clearly state what exactly a person is responsible for, how can AI know that?"

This sentence reminded me of what Sun Wenfeng said: "Connections cannot be passed on". The two people pointed out the same problem from different perspectives: The organizational reality of Chinese enterprises is the real obstacle for AI implementation. It is not a technical problem, but a human problem. Those parts that operate heavily relying on relationships, experience and tacit understanding can hardly be directly handed over to AI.

The Bridge That Will Never Disappear

Near the end of the interview, I asked a question that all FDE practitioners have to face: When AI becomes more and more powerful, is there still a need for FDE to exist? Will it only be a "small crutch" in the stage when the model capability is insufficient?

Gan Yifan believes that as model capabilities differ, the problems that FDE needs to solve will also change accordingly.

"When the model is not strong enough, we need to make up for the shortcomings of the model through engineering methods. When the model is strong enough, it still cannot directly explore many things that are happening in the real world. At this time, what FDE needs to do is another kind of engineering: to provide sufficient context for the model."

In his opinion, FDE is the intermediary that connects AI, the real world and business objectives. The model can process information and generate solutions, but what is happening in reality and what problems the enterprise really needs to solve still needs someone to tell it.

"The day FDE disappears is the day when no one needs to work anymore."

Liu Kai placed this role in a longer timeline:

"In the past 20 years, this role may be called product manager, or digital BP; today we call it FDE; in the future, it may be called context environment provider, or even human mood soother. There will always be such a bridge between the digital world and the physical world."

Product manager, digital BP, FDE, context environment provider, human mood soother, the name keeps changing, but its position remains the same: to understand the problems in the real world, and then convert these problems into tasks that the digital world can understand and process.

After finishing this note, I thought of another metaphor Gan Yifan said:

"Stark became Iron Man not because an ordinary person puts on the armor and can become Iron Man. It is because he wants to become Iron Man that he is worthy of that suit of armor."

AI is just like that suit of armor. No matter how powerful the tool is, it cannot automatically turn a person into an FDE. Whether it is FDE, GAB, or foreman, the name does not matter. What really matters is whether this person has the will to go deep into the site, face those vague and complex problems, and get things done for real.

This article is from the WeChat Official Account "Neuters" (ID: Neuters), author: Cui Qiang, editor: Yanzi, published with authorization from 36Kr.