Head of AI is four different jobs. Decide which one before you hire.
Written by Miki Blasko, veteran recruiter and headhunter for Deconstructor of Fun Talent.
A founder of a mid-size developer (+$50M) called me back last week wanting to hire a Head of AI. Reasonable ask. So I asked the obvious follow-up: to do what?
20 minutes later, we had established that what he needed was a production engineer who could wire generative tooling into an existing art pipeline. The six-figure leadership hire they had budgeted for would have likely spent a year with nothing to lead, writing strategy decks for a team that did not exist and benchmarking vendors in a market where new leaders emerge on a monthly basis.
We have some version of that conversation almost every week. The volume of studios wanting to hire for AI has exploded. No other way to describe it. Yet the clarity about what that means has not moved an inch.
Titles tell you nothing
Director of AI. Head of AI. AI Chief of Staff. Forward Deployed Engineer. Sometimes these are simply seniority variations on the same job. Often the focus, internal versus external, is completely different. More often, the title means nothing at all because the job description was written by AI, which is a poor authority on what your company's current stage actually calls for.
For example, Netflix advertised a Director of Generative AI for Games asking for a decade-plus of gaming experience plus deep AI/ML expertise, with a disclosed band of $430,000 to $840,000. Elsewhere, roles with generative AI in the title get posted as hourly contract work.
This matters more than it sounds. Games-industry AI talent largely does not advertise itself with clean, searchable titles. The best candidates are usually inside another function: a technical director who built the pipeline tooling, an analytics lead who shipped the moderation model. They are not labelled experts. That’s why a keyword search finds you people who optimized their LinkedIn headline. And that’s a different population from people who did the actual work. It happens to be the same population that demands a bombastic AI title.
Head of AI is four different jobs
The responsibilities problem is even worse than the naming problem since surprisingly many companies offer almost a blank slate for their AI hires. In my experience, the misalignment of expectations by not setting expectations leads to costly terminations within a year.
I’ve learned that a “Head of AI” can mean roughly four things:
A research leader pushing model capability
An internal transformation leader changing how the company operates
A creative-pipeline leader embedding generative tooling into art and content
A live-ops leader applying ML to player data, moderation and matchmaking.
If you’re confused, you should be. That’s four different people. Four different backgrounds. Four different definitions of success. And a high chance of hiring the wrong person. A CEO who has not decided which mandate they are buying can end up with a leader optimizing the wrong thing. And they will realize this in month eleven.
Research, transformation, creative pipeline or live ops. Pick one before you post the job.
We actually had a case like this: A public gaming company hired a head of AI to take their studios into the new era. The leadership believed they were hiring an internal transformation leader, yet what the studios needed was a live-ops leader. They were all bought in. They needed direct execution support. That journey ended inside the first year with the head of AI looking for another job.
Here’s a thing to keep in mind: the adoption data says the value is landing first in the operational layer. This means crafts like coding, QA, marketing, analytics, and art. Square Enix has publicly committed to automating 70% of its QA and debugging by the end of 2027.
The role you need first is usually the unglamorous one that makes your existing pipeline faster. Not the moonshot hire that you hope will make your board stop asking about “how you’re using AI in your company?”
Head or Troops First?
Based on what I just said, you must think that I’m always suggesting individual contributors (ICs) over managers and heads. So let me insert a twist.
Troops first for a clear use case. A head first when AI cuts across every team.
Hire troops first when you have a clear, bounded use case. Automate the QA pass. Accelerate the texture pipeline. Build the live-ops analytics layer. If you already have engineering or production leadership credible enough to direct that work, a senior AI leader is premature. Get ICs in, ship something real, learn what the actual problems are. Create proof before a strategy function.
Hire the head first when AI is cross-cutting and politically contested across teams, when the failure mode you are avoiding is fragmentation, or when you are at a scale where a transformation mandate needs an owner with authority to make tradeoffs and kill bad projects. At that point a strong leader is the thing standing between your company and an expensive mess.
Both hires share one root error: buying capability with no judgment attached or buying judgment with nothing to apply it to. The right sequence depends entirely on which one you are currently short of. Most companies do not know which that is until someone maps it out for them. And they think that buying judgment is the same as buying mapping and capability in one package. It’s not.
When It Goes Wrong…
EA is the most instructive public example of strategy without operational ownership. The strategy was defensible and the lack of execution hit every failure mode at once.
The kick-off was perfect. They had board-level commitment to scaling generative AI. Partnerships with top tier vendors. Strong top-down intent. Then the internal reality, first reported by Business Insider: ReefGPT, the in-house chatbot executives described as a thought partner, producing faulty code and hallucinations that developers had to stop and fix. A tool that created work instead of removing it.
At the same time, creative staff reportedly expected to train AI on their own output while fearing it would erode demand for if not their jobs then at least their co-workers’ jobs. And when the leadership is pressing roughly 15,000 employees to use AI for nearly everything, widely read as cost-cutting after the $55 billion debt infused acquisition.
That is strategy without operational ownership. The lesson is not: don't do AI. It’s that the gap between a board-level commitment and a working capability gets filled by people in correctly defined roles, hired in the right order. When you leave the definitions vague, that gap is where the money goes.
There is a slower-burning version of the same mistake.
An IDC survey found 66% of enterprises expect to slow entry-level hiring, and 91% report roles already changing or disappearing because of AI. In tech, recent-graduate hiring at large firms fell 25% year-on-year in 2024 and sits roughly 50% below pre-pandemic levels.
In games, junior roles have always been how senior talent gets made. And with AI, the logical step I see many studios taking is cutting the hiring of junior talent. An agentic workflow with a senior engineer removes the need for extra hands. But hollowing the bench off of juniors is a risk.
Cut them and your most expensive people spend their days checking machine output. Judgment comes from making bad work and getting corrected. Take that away and in six years your senior people are the ones who only ever edited AI. They will approve work that looks right. And nobody wants to train a junior who gets poached at year four. So everyone free-rides, and nobody trains anyone.
Personally, I encourage companies to hire AI-enthusiastic juniors for junior roles rather than building often very expensive agentic systems in an attempt to save on junior hires. The data we see from companies is that junior employees with AI tools and agentic workflows level up in months to a position where they commit almost the same level of code as their more senior counterparts.
Where, why, how long, and how much will it cost?
Every founder, executive, and hiring manager I talk to about AI hiring believes their problem is finding the person. It almost never is. Trust me, the market has people. What I find lacking are companies that can articulate what the person is and isn’t for.
This makes AI hiring a mirror. The clarity of your job description is the clarity of your strategy. If you cannot describe the role without using the word AI, chances are, you have not decided anything yet. You have just noticed that other people are moving and want to be seen moving too. And so you hire someone to take you somewhere. But you don't quite know where, how long it will take, and how much it will cost you. But the fear of missing out is strong.
The hires that we’ve seen work are boring to describe and specific to death. The hires that fail sound impressive in a board update. That’s actually always been true of hiring. AI just made the failure mode more expensive.
If you are about to hire for AI and want to pressure-test whether it is the right hire, that conversation is what the DoF Talent does. Reach us at talent@deconstructoroffun.com.
Related reading
Why Great Hires Fail at Startups
Hiring as a Competitive Advantage
How Game Studios Should Allocate AI Budgets
