You've Just Been Named First CIO (or Head of AI). You Have No Team. Now What.
The press release is out. «The company appoints its first chief technology officer / head of artificial intelligence / digital transformation director to lead its AI strategy.» Congratulations, sincerely — it's a good job at the best possible moment to have it.
Now the part the press release doesn't cover: you have a remit, a budget to defend and a board expecting results this year. What you don't have is a team. Not one AI engineer, not one data engineer, nobody who has ever put a model into production. Often there isn't even a technical department to lean on — your company manufactures, insures, distributes or builds, and technology was always a vendor's job.
I talk to people in this exact situation every week. It isn't the edge case: it's the case. And the pattern of who survives year one and who doesn't is clear enough to write down.
Your job is the new norm, not the experiment
If you feel like a pioneer, the data disagree. According to an IBM survey reported by Rework, 76% of organizations worldwide had named an AI lead by 2026, up from 26% in 2025. The Chief AI Officer role is growing at triple digits, and the Gartner figures circulating in the industry suggest organizations with a dedicated AI lead deploy three times faster and get 2.5 times more return on their AI investments.
That last number is your budget argument — and your trap. Because the return doesn't come from the appointment: it comes from what gets built afterwards. An AI lead with no capacity to execute is a press release with a LinkedIn profile.
The clock starts on day one
The first months of a new role run on borrowed credit. The board that hired you defends you, the budget is fresh, and nobody expects miracles yet. That credit expires — usually at the first budget cycle.
Do the math with me. If your plan is "first I build the team": a senior hiring process in Spain takes three to six months per profile when the profile is AI, and that's assuming your employer brand competes well for AI talent — which, if you're an industrial or services company with no technical history, is assuming a lot. Three sequential hires and you're presenting your first internal demo right when you should be defending the year-two budget.
I've watched more than one new appointment die exactly this way: not from bad strategy, but from a dishonest Gantt chart.
The rule I give them: by the end of the first quarter, something has to be in production, used by someone who isn't you. Small, measurable, real. That early win turns your mandate from theoretical into operational, and everything else — budget, hires, the board's patience — hangs from it.
Heads and hands: the sequencing mistake that shapes everything else
Almost every new AI lead makes the same inversion: they try to hire the heads first (a head of data, an architect, a lead) and the hands later. It sounds reasonable. It goes wrong for three reasons:
Good heads don't move to an empty lot. A great AI architect wants to land where there's already something to architect. With no live project, your offer competes on salary alone — against banks and big tech.
Without production, you don't know which heads you need. The org chart you draw in month one almost never survives the first real use case. Hiring permanently against a theoretical org chart is locking your hypothesis errors into payroll.
The cost of being wrong is asymmetric. Unwinding a bad senior hire on payroll costs months and money. Adjusting an external team costs a notice period.
The sequence that works is the reverse: borrowed hands first, your own heads later. An external senior squad builds the first use case while you learn what capabilities your company actually needs — and you hire the permanent team with a live project to show candidates in interviews, not a deck.
The three ways to get build capacity (seen from your chair)
In the AI adoption guide for mid-size and large companies I walk through the three paths in detail; here's the version for your specific seat.
Hiring on payroll. Right as a destination, lethal as a starting point — see the Gantt above. Real cost: 1.4-1.5 times gross per profile, two to three months per process.
A large consultancy. It will give you a magnificent plan and legitimacy in front of the board. But you already have the mandate — you don't need to buy legitimacy, you need to buy code in production. And the fine print of execution (a team you don't choose, rotation, implementation rates) gets paid by you, in credibility.
A deployed technical squad. Senior external engineers embedded with you from week one, on your backlog and your repository. It's our model at Conectia, so here are the numbers to compare against: seniors at 16-22 €/h, senior leads at 22-28 €/h, first candidate within 72 hours, a 3% intake filter. The MintID case study — which renewed at double the initial budget — shows what this looks like at the six-month mark.
It isn't a purity decision, it's a sequencing one: squad for quarter one, your own hiring from quarter two or three, a consultancy if the day comes when the problem is redesigning the whole organization.
The five first-year mistakes (AI-lead edition)
1. Starting with the platform. «First we build the data lake / the MLOps platform / the reference architecture, then the use cases.» That's pilot purgatory with an infrastructure budget. The platform earns its keep at the third use case, not before the first.
2. Accepting the political use case over the viable one. The one the CEO asks for isn't always the one with accessible data and measurable ROI. Your first project can't afford to be the hard one: negotiate the order, not the destination.
3. Signing with someone whose interview profiles won't write your code. Always ask who will actually be in your repository. If the proposal team and the delivery team are different people, you're buying a raffle ticket.
4. Not locking down the handover. Everything an external team builds has to be documented for internalization — architecture decisions, data pipeline, runbooks. Demand it in the contract from day one. A partner who resists the handover is optimizing their monthly invoice, not your company.
5. Ignoring governance until the audit. With the EU AI Act rolling out in phases, use cases touching people or regulated sectors need the risk analysis before they're built. As the AI lead, the fine carries your signature even if the code was someone else's.
The minimum viable team for year one
What it actually takes to reach production in a quarter, without aspirational org charts:
- You, with the mandate and the business priority clear.
- An external senior lead who translates business case into architecture — the expensive-decisions profile, not the hours-on-a-calendar one.
- Two or three senior engineers (AI + data + integration) building from week one.
- An internal product owner — someone from your own house who knows the process you're transforming. This role doesn't get outsourced: it's where the business knowledge lives.
With that structure, the realistic range for a first use case in production is one quarter and 60,000-120,000 €. From the second case on, the same team delivers at twice the pace: the infrastructure, the patterns and the judgment already exist.
FAQ
Shouldn't I hire a CTO first to help me decide?
If there's nobody technical at the table, a fractional CTO can cover the architecture decisions of the launch phase — I've written about when that makes sense and when it's a mistake. Just don't use it to postpone execution: decisions without construction expire.
My company has never had a technical team. Can it absorb an external squad?
It's the most common case, not the hardest. The squad brings its own practices (repository, CI/CD, backlog management) and leaves them installed. The one piece your company must supply, no exceptions, is the internal product owner.
What do I tell the board when they ask why we don't just hire in-house?
The two Gantt numbers: two to three months per senior hire against 72 hours per deployed engineer, and a total payroll cost of 1.4-1.5 times gross against an hourly rate that switches off when the project no longer needs it. And the deeper argument: you'll hire better in six months, with a live project to show candidates.
Have the mandate but not the hands?
Talk to a technical partner and deploy CTO-vetted engineers in 72 hours.
