The get-rich-quick myth of AI headhunters: Find one person and get a commission of 3 million RMB?

Can an AI headhunter earn nearly 3 million RMB from a single deal?!

Not long ago, a screenshot went viral across numerous groups: a humanoid intelligence company was recruiting AI algorithm experts, with a total annual compensation package as high as 10.5 million RMB; the headhunter fee is 26%, amounting to 2.73 million RMB.

As soon as the news came out, some envied the market value of core AI talents, while others sighed that the AI wave has not only raised the salaries of practitioners, but also enabled relevant roles in the industrial chain to find new profit opportunities. The scarcer the talents are, the fiercer the competition will be, and headhunters standing between enterprises and candidates have also been pushed to the forefront of the trend.

This kind of competition does not only exist in rumors. According to Euronews, giants including OpenAI, Meta, Google and Anthropic are scrambling for the extremely limited number of top AI researchers. Among them, Meta once offered 24-year-old AI researcher Matt Deitke a four-year compensation package worth about 250 million US dollars in total. The domestic competition is also pushing up the remuneration of a small number of core positions, and the annual salary of some core positions in ByteDance's Top Seed program has exceeded 6 million RMB.

In this escalating high-price competition, a thought-provoking paradox has also emerged.

From the perspective of technical logic, headhunting seems to be exactly an occupation that AI can easily replace. Enterprises can use AI to search resumes, screen candidates and match positions, with far higher efficiency than traditional manual recommendation. At the same time, real AI talents have no shortage of job offers, they often hold multiple offers and only need to make a choice, so it seems that they do not need headhunters to help them find jobs.

However, the reality has gone in the opposite direction: the headhunting business in the AI industry is getting better and better. This occupation, which can be improved in efficiency or even replaced by AI, has earned more money in the AI talent war.

With all kinds of doubts, we found Peter, an employee of another subsidiary of the headhunting group mentioned in the widely spread screenshot. He confirmed that the order in the screenshot does exist, but 2.73 million RMB is the service fee at the company level, which does not mean that a single headhunter can take all of it.

"There are multi-layer distributions for the company and the team in the middle, and the amount that actually goes to an individual is about 400,000 to 500,000 RMB."

What is more worth exploring than this commission is: when AI can already screen resumes and match positions, why does the recruitment of top talents rely more on headhunters? Between enterprises and candidates who hold multiple offers, what exactly do headhunters provide?

The answer is hidden in the real workflow of AI industry headhunters. The following is Peter's account —

AI headhunters on the trend do not "fly" that easily

I switched to headhunting because I didn't want to work on the enterprise side anymore.

I have been doing recruitment work on the enterprise side since my internship in 2022. In those years, I was looking for people in the same industry almost every day, and the positions and candidates I met were very vertical. But headhunting is different, it's like pushing open a wall in the office: I face different companies and different tracks at the same time, and the choices I have are suddenly much more.

So in 2025, I decided to switch to the other side, and began to look for talents in the market for enterprises.

After I joined the industry, I found that headhunting is not as simple as getting a JD and sending it to candidates. The work is mainly divided into two parts: one side is called business development, which means taking the initiative to find companies with recruitment needs, and negotiating cooperation with founders, HR or other persons in charge; the other side is called delivery, which means sending suitable candidates to enterprises after receiving the position. I do both parts myself, I need to figure out exactly what kind of talents the client wants, and also bear the pressure of whether I can successfully deliver suitable candidates in the end.

Although I am based in Ningbo, my AI clients are mainly concentrated in Beijing, Shanghai, Shenzhen and Hangzhou. Most of the demands can be clarified through online communication. If there is a situation that requires face-to-face communication, I can buy a ticket the next day and rush there. In fact, after working for a long time, you will find that for headhunters, the distance between cities is not a big problem.

What is really difficult to cross is the gap between different AI businesses.

Even for AI companies, some develop software, some make hardware; some serve enterprises, some directly target consumers. Even for product roles, some are for selling efficiency, while others are for providing emotional value. Different businesses require completely different talents. So every time I contact a new client, I have to first figure out what the company is doing at present, what the founder's background is, what stage the enterprise is in, and so on.

Only when these questions are answered can I know where to find the right people.

Looking along these demands, the heat of the market will gradually show. World models, humanoid intelligence and robotics tracks receive relatively more capital, hardware is roughly in the second echelon, and software depends on what specific products are being developed.

Where the money flows, the recruitment demand will naturally follow.

When it comes to specific positions, the demand for front-end roles is relatively small, while back-end, full-stack, algorithm and product positions are very popular. After the product is actually developed, the enterprise has to find a way to sell it, so BD has also become an indispensable part of the recruitment chain.

However, more positions do not mean that everyone can get a high salary. Most job hopping cases still get a 10%-20% increase on the original salary, and the remuneration is usually mainly in cash. Whether there are options and what proportion they account for depends on the position and the candidate's title.

The income of headhunters also follows the same reality. The common service rate in the industry is 20%-25%, and the common annual performance of traditional headhunters is about 700,000 to 1 million RMB. With base salary and commission, the personal annual income is usually 200,000 to 400,000 RMB. However, when you master talents that enterprises cannot find and other headhunters cannot easily reach, you also have the opportunity to negotiate a higher service rate.

There are many AI positions and talents flow fast, so the chance of closing a deal is indeed more than that in some traditional industries. But a hot track does not mean that every headhunter can earn more money. The income is ultimately determined by delivery: finding candidates is one aspect, and only when the candidates are actually hired by the enterprise and successfully join the job can the deal be counted as valid performance.

Moreover, even after the candidate joins the company, the money is not actually in your pocket.

Enterprises usually set a probation period of three or six months, and the headhunter fee will be paid in two or even three installments. If the candidate leaves the job during the probation period due to mismatched capabilities, the headhunter needs to re-supply a suitable candidate within the agreed time; if no suitable candidate can be supplemented, the headhunter will refund the fee already received.

Therefore, what headhunters earn is never the money for "finding a person", but the money for "finding a suitable person and making him stay stably".

I moved from the enterprise side to headhunting, and then from traditional recruitment to the AI industry. I thought I just switched to a popular track. Later, I realized that the trend only brings more positions, and whether you can seize them starts with finding the right person.

The fastest time to compete for talents is after others get off work

Many candidates can only be reached by phone at night.

During the day, they are in meetings and rushing for projects, and they will not stop working for unfamiliar numbers, so my working hours will also be postponed. While others work seven or eight hours a day, I may work three or four more hours; sometimes I continue to communicate on weekends, and sometimes I rush to a candidate's city to meet him for a cup of coffee.

The headhunting talent poaching imagined by the outside world is always a little dramatic with "chasing and blocking", but in fact it is not so exaggerated. I won't stare at the location on Moments to create a chance encounter, that's too deliberate. If I want to meet a person, I will contact him first, and then formally make an appointment. When we sit down for the first time, we don't necessarily talk about positions, I will get to know him first, listen to what he is doing, and then judge whether he wants to change his working environment.

Before that, there is an invisible preparation process.

When an enterprise sends a position, I will not immediately rush into the resume library, but first ask: Why are you recruiting now? How long have you been looking for? What stage is the team at? What kind of talents do you prefer? What salary and benefits can you offer? After the enterprise clarifies the demand, the talent search work will actually start.

I will first draw a "talent map": which cities and companies the core talents in this field are concentrated in, and through which channels I can contact them. Recruitment platforms and LinkedIn are the most commonly used entrances, and offline hackathons, forums, salons and exhibitions cannot be missed either. When I find a suitable resume but cannot contact the person himself, I may continue to look for him through his alma mater and alumni relationships.

However, even if you find the right direction, you may not find the right person immediately. The AI industry is updating so fast that new concepts keep emerging, and the same capability may be packaged into completely different position titles in different companies.

FDE is a typical example.

Different companies have inconsistent understandings of FDE, and the candidate profiles they give are sometimes very vague. If you only search according to the position title, it is easy to go further and further off track. In this case, I can only first select a group of people who meet the approximate requirements for recommendation, and then continuously adjust according to the enterprise's interview feedback. After several rounds, the originally unclear demand will gradually show its true outline.

The titles can keep changing, but there are many common standards for enterprises to select talents. Being smart, able to think independently, and having experienced the 0-to-1 process are the most frequently mentioned requirements. Startups will also ask one more question: is this person willing to grow up with the company?

In addition to soft requirements, there is a more direct screening line.

Some enterprises clearly require candidates to graduate from 985, 211 or Double First-Class universities, and even specify the school, education level and experience in large factories. Age is also a threshold: some companies hope the candidate is under 30 years old, and some will relax to 35 years old. But people over 35 are not completely out of opportunities. If their project experience is outstanding enough, and they have experience in large factories or startups, they can still apply for product positions, while there are far fewer opportunities for technical positions.

After finding a candidate who matches the profile, I still have to go through the salary negotiation stage.

Excellent candidates often hold more than one offer, and some will ask for a 30%-40% salary increase. If there is an inversion between the original salary and job level, and the interview result proves that the candidate's ability does match the position, I can fully understand this request. From the perspective of a headhunter, I naturally also hope that the candidate can get a higher salary, because the higher his annual salary is, the higher the headhunter's service fee will be.

But the enterprise sitting on the other side of the negotiation table is also calculating its own employment cost. Most companies can accept a 10%-20% salary increase, and only when the position is urgent enough, the direction is new enough, or the number of people in the market who can actually do this job is very limited, the budget may be further increased.

However, what really makes people nervous is often not salary negotiation, but time.

The same position may be handed over to multiple headhunters at the same time, and the enterprise HR is also building their own candidate list. Whoever completes the effective recommendation first will usually get the opportunity. If a candidate has already entered the enterprise's talent pool, or has been recommended by other headhunters, even if the later headhunter finds him, this performance will not be counted under his name. Therefore, before every communication, I will ask first: "Have you ever been in contact with this company before?"

Our group shares candidate resources internally, we will not snatch orders from each other, nor will we poach people from client B for the position of client A. But outside the group, all headhunters are often facing the same group of candidate names.

Therefore, the "AI talent war" seen by the outside world may not be such a fierce scene in our eyes. More often, it is like a quiet race: when others stop in front of an unanswered resume, I will find one more layer of relationship; when others get off work, I will make one more call.

One step earlier may mean a closed deal; one step later, that person has already been included in someone else's talent pool.

AI can retrieve resumes, but cannot find talents "hidden under the water"

Since the victory or defeat of talent competition is often only one step apart, AI has naturally become the most convenient accelerator. The work that used to take a long time to search and organize can now be run through AI first.

For example, after receiving resumes, I will put them all into my own resume library. When a new position comes in, I will first let AI do a round of matching according to the JD, and circle out the people who may be suitable.

Now there are already many Agents on the market that are specially designed to find talents for enterprises. They can search for candidates according to position requirements, and some can even complete the first contact with candidates through voice calls. For positions with an annual salary of less than one million RMB and relatively clear requirements, these tools can indeed take over a large number of repetitive tasks.

However, after the efficiency is improved, the gap between headhunters becomes even more obvious.

Once, the same candidate was contacted by another headhunter first. That headhunter copied and pasted the JD and company introduction to him, and quickly completed the recommendation. Later when I communicated with him, I didn't rush to make a recommendation, but first explained the company clearly: the basic situation of the company, what stage it is in, the structure of the team, the background of the founder and co-founder, the candidate's past experience and his motivation for leaving the current job.

Although the candidate did not pass the enterprise's initial resume screening, he remembered the difference between the two types of headhunters. At least he would think that the latter way of communication is more reliable.

This experience made me see more clearly: the threshold of headhunting is not adding the word "AI" to the business card, nor is it copying a JD faster than others, but whether you can clearly tell a specific person where he is going.

AI can quickly find a group of people who "look suitable", but it may not be able to find the exact one that the enterprise really wants.

In particular, those core AI talents almost never post their resumes on recruitment platforms publicly. They have no shortage of jobs, let alone no shortage of offers, so their resumes will never "float on the surface". Moreover, these people have been in contact with a large number of founders and CEOs on weekdays, and can even flow directly between enterprises, without going through headhunters at all.

AI can screen out a hundred suitable resumes from existing materials, but it is difficult to find the person who has not left any clues in public channels. To find these "underwater people", we rely on talent maps, referrals from the industry circle, and relationships accumulated through in-depth communication again and again.

Even if you finally get in touch with them, the work is only half done.

The problem for top AI talents is