A year ago, when I opened a vacancy, I roughly knew who I was looking for: there was a job title, a typical profile and a market I understood. Now, more and more often, a manager comes to me with a task, and together we realise that this profession does not exist yet. Not on job boards, and not in candidates' heads.
I am the Chief People Officer at GPTunneL. Here is why hiring in an AI company needs new specialisations, told through one role that grew inside our company on its own.
Roles appear faster than their names
The AI market changes every week, and the company keeps getting tasks nobody did yesterday. At first someone takes them on "on the side". Then it becomes clear that this is a separate job. And there is no ready-made person for it on the market.
New roles are born where old ones meet: a procurement specialist who understands tokens and APIs, or an account manager who negotiates with suppliers in the US and China. If you search for these people by the old job title, the wrong candidates show up.
An example: the role that no job board lists
At first we were looking for an AI partnerships manager. Then we looked at what our colleague who already handled similar tasks actually did day to day, and we understood: this is one big role, not three different ones. This person owns the whole journey of a model, from the supplier to the customer:
- Suppliers. Finds new providers and negotiates prices, limits and SLAs.
- Platform. Makes sure that 200+ models run without failures: if one provider goes down, the request is instantly routed to another. New models appear on our platform within a day of their release.
- Money. Calculates the cost of every request, reconciles supplier invoices with our own statistics, and notices when a model starts losing money.
- Model honesty. Checks that nobody is selling us a cut-down copy under the name of a flagship model.
- Customers. Grows the product for companies: access and budgets for employees, corporate API keys, terms for large contracts.
- Analytics. Builds metrics and dashboards so the whole company looks at the same numbers.
In the past these would have been three vacancies: a buyer, a technical product manager and a B2B product manager. In an AI company they are hard to separate. The price you negotiated with a supplier immediately becomes your margin and the customer's rate. And a model connected a month later than competitors means users who have already left.
We argued about the title longer than about the duties. The options were "AI Platform Product Manager", "AI Supply & Monetization Lead" and "AI Model Marketplace Lead". The last one is closest to me: in essence, this is the person who keeps the balance between those who sell neural networks and those who use them.
Who we are really looking for
The ideal candidate with "5 years of managing an AI platform" on their CV does not exist. So I stopped looking for a full match on paper. Now I look at a few qualities that matter more than the line about a previous job title.
Curiosity about the market. I always ask: "Which neural network did you try most recently, and what did you like about it?" A person with genuine interest answers at once and with details. Someone who came "just for an IT job" starts recalling ChatGPT.
Fundamentals at the "explain it in your own words" level. I ask candidates to explain what an API or a token is, as if we were explaining it to a grandmother. Many know the terms. Few can explain them simply.
Being friends with numbers. Here, every request to a model costs money. I ask the candidate to estimate on a napkin how much we earn on one request. The exact answer does not matter. What matters is the angle from which the person approaches it.
Independence. In a role like this, nobody will bring you a ready-made list of tasks. In the interview I ask the candidate to tell me about a deal they found on their own, from scratch. And I listen for where the story says "I" and where it says "we as a team".
A calm attitude to change. Supplier terms change, and so do priorities inside the company. If a person needs a plan for a year ahead, they will find it hard. That is neither good nor bad. It is just important to say it honestly at the start.
One more thing. Our co-founder Ilya Trikoz wrote recently that AI has made development itself much cheaper. What stays with the human is taste, decisions and responsibility. I see the same thing in hiring: a model will write the email to a supplier, but deciding whom to negotiate with and on what terms is something only a person can do. That is exactly the skill I test.
How hiring itself is changing
New roles change not only who we look for, but how we look. Here is what we changed in our own process.
Tasks first, title later. When a manager brings a request, we do not start with the job title. We write down what the person will do every week and which numbers will tell us they are doing well. The title comes last.
We check whether a person is needed at all. Ruslan Kharsiev described how, after a manager left, a bot took over their work. It turned out that of seven steps, only two involved real decisions, and the rest was moving data around. Now, before we open a vacancy, we ask ourselves the same question: where are the decisions here, and where is the routine? Sometimes automation is better than hiring. And sometimes it is the other way round: a bot takes the routine, and the vacancy becomes one strong role instead of two weak ones.
We test by doing, not by telling. For this role we give a live case: "A supplier has raised the price and the model is now losing money. What will you do?" A good candidate goes to the numbers first and to the negotiating table second. We only assess English in a live conversation. C1 on a CV and C1 on a call are often two different things.
We are honest about the inconvenient parts. Suppliers in the US and Europe mean evening calls. A new role means the duties will keep being built out on the go. If you stay silent about this at the interview, the person leaves during the probation period.
Onboarding is longer and deeper. A newcomer should not just learn where the documents are. They need to understand the product well enough to demo it with confidence. So the probation plan includes time to learn the platform, and we check it at onboarding meetings.
What I would advise
To colleagues in HR and recruiting
- Use your company's product yourself. It is hard to hire people in AI if you have never compared two models on the same task.
- Look for people at the intersection of fields. Someone for a role like ours may come from product management, financial analysis or IT procurement.
- Break the request down together with the manager. In a new role, the manager often does not fully know who they are looking for. Your job is to help them put it into words.
To candidates
- Do not be afraid of vacancies with unfamiliar titles. Read the duties: it often turns out you have already done most of them.
- Show that you use AI in your work. Not as a list of tool names on a CV, but as an example: which task you solved and how much time you saved.
- Talk about your personal contribution and specific numbers. That is valued everywhere, and in fast-moving teams especially.
Instead of a conclusion
Jobs of the new market do not appear out of nowhere. They grow out of old ones when you add an understanding of AI and the speed at which the market changes. AI Model Marketplace Lead is one example. We have others in the works: prompt engineer, research analyst, product marketer for an AI product.
For HR this means one thing: you can no longer take an old vacancy template and just change the title in it. You need to understand the business as well as the managers do and help them invent roles that do not exist yet. To me, this is the most interesting part of our profession right now.
If you recognised yourself in the description above, get in touch: we are always glad to meet.