Hire a Head of AI when your company has enough AI work to keep that person and at least one engineer busy all year. Until then, we think a company of 20 to 500 people gets further with a fractional AI department, an outside team that sets the strategy, trains the staff, and builds the automations. When you do hire, that team can hand the work over or keep building for your new Head of AI.
We sell the second option, so read this with that in mind. We have also talked a prospect through making the hire, and the profile we gave them is further down.
Someone has to own AI
Every company is going to need a person or a team that owns AI, the same way someone owns finance and HR. Whether to hire for it is one of the questions we hear most on sales calls. The usual backdrop is one person using ChatGPT, another using Claude, and nobody deciding anything.
Companies already rent other functions this way. A business keeps a law firm on retainer and has an accounting firm close its books every month, long before it hires a general counsel or a CFO. We think AI is at that stage for a company of this size.
Why one hire struggles
Read a Head of AI job posting and count the jobs inside it. The person is expected to set the strategy, write the AI policy, train every department, connect the systems, then build and maintain the automations. Those are different skills, and the last two need engineers.
When a large real estate company asked us in May what kind of AI hire to make, Julian told them the person they were picturing might not exist as a single hire. People who can build with these tools mostly don't want a corporate seat right now. The tools also change so quickly that, as he put it on the call, you almost have to be unemployed to keep up.
There is a timing problem too. A Head of AI who joins a company with scattered data, no policy, and staff on personal accounts spends their first months on setup, and nothing gets built until that's done.
What a fractional AI department does
Ours starts by learning the business. We spend the first 30 to 60 days interviewing everyone on the team for 30 minutes to an hour each. A survey goes out beforehand asking how people use AI today and what makes them nervous about it. We pair the interviews with a couple of quick wins so the early weeks produce something people can use.
The interviews turn into a ranked roadmap. It says what to automate first, what can wait, and what to leave alone because automating it would waste your time. An outside team is most useful at this point, because your staff can't ask for automations they don't know are possible.
From there the work runs on three tracks. The first is training every role, alongside an AI policy and with your IT team involved from the start. The second is building the automations and agents on the roadmap. The third is connecting your systems into one data layer, so each new build has your company's information to work from.
Then we stay. When Anthropic, OpenAI, or one of your software vendors ships something new, we tell you whether it changes the plan. We can review proposals from other vendors, including the AI features your software vendors are adding. If you're about to hire for a role and AI can cover the gap for now, we'll say so.
You own what gets built. The code, prompts, and data sit in your accounts, and we build so that another developer could pick the work up without us.
What it asks of you
A fractional team has no authority inside your company. Someone senior has to back the roadmap when two departments both want to go first, and that person needs to join a 30-minute call most weeks. Your staff give up time for interviews in the first month.
If nobody inside can spare that, the work will drift. A full-time hire with a seat at the leadership table will do better in that situation.
The two side by side
| Head of AI | Fractional AI department | |
|---|---|---|
| Who does the work | One senior employee, plus any engineers you hire for them | A team of strategists and engineers |
| What you pay | Salary, benefits, and recruiting | A monthly fee |
| Who owns what gets built | You | You |
| Authority inside the company | Their own, if leadership gives it | Borrowed from an executive sponsor |
| Fits when | AI work fills a year for a leader and an engineer | The work is real and doesn't yet fill a team |
Signs you're ready to hire
We'd make the hire once these four things are true.
- Your data sits somewhere AI can read it.
- You have a ranked list of what to build, and staff already work in one approved AI tool under a written policy.
- There is a year of work for a leader and at least one engineer.
- Someone at the top will give that person authority across departments.
If the first two are missing, your new hire starts with the setup work described above, and you wait for the builds.
Who we'd hire first
The profile Julian gave that real estate company was specific. He suggested someone a year or two out of school with startup experience, strong communication skills, and a background outside computer science. That person works full time and pulls in outside help for particular projects.
We'd set it up that way. The internal person holds the relationships inside the company, and the outside team supplies the engineering.
After you hire
Our role changes when a Head of AI arrives. Your leader can set direction while we keep building, or we can step down to training and strategy. Some companies will bring on an in-house engineer to maintain what we built. We've told clients that all of those outcomes are fine with us.
A way to decide this week
Write down the AI work you want done in the next 12 months and put a name beside each item. If the list would fill a year for a senior leader and an engineer, start recruiting.
If most lines have no name, or the same name sits beside every line, you're asking one person to be a department. A fractional team is built for that situation.
If you want a second opinion on the list, book a free audit and we'll go through it with you.