Several pilots, nothing live
The pilots work. None of them is in production, and nobody's job is to stop nine of them and finish one. That decision needs someone with no stake in which nine.
Service · Interim AI Leadership
An experienced AI leader in the seat, for as long as you actually need one
The decisions arrive long before the job advert does.
Experienced AI leadership for exactly as long as you need it, in whatever role fits. The title matters far less than the work. It starts with a strategy that someone then has to deliver: a few advisory days a month to get the direction right, or hands-on leadership through a whole programme. You do not need the budget for a full-time executive to start, and you do not need to wait two quarters for the right one to accept.
Starts in weeksCadence days a month to days a weekMandate 3 to 18 monthsEnds with your own successor
Get in touch See the three ways to work
Why now
Six reasons companies bring someone in
Almost every mandate starts from one of these. Usually two of them at once.
The permanent hire is two quarters away
A credible AI leader takes months to find, longer to free up, and the decisions do not wait for the notice period. We hold the seat while you run the search, and help you run it.
Every supplier now sells AI
The slide decks are excellent and they all promise the same thing. Someone in the room has to know what is a product, what is a prototype with a sales team, and what you already own.
The first AI hires are about to be made
Hiring an AI team without anyone who can judge the candidates is how you end up with the wrong three people and a year of goodwill spent. An interim leader hires, structures and mentors the team, including their own successor.
Someone has to own governance
Article 4 of the EU AI Act has applied since February 2025 and the transparency duties since August 2026. The law does not tell you who signs off. Your auditors, your works council and your customers will still ask.
An investor or an acquirer is asking
Due diligence on an AI claim needs someone who can read the code, the data pipeline and the team, and then say in one page whether the technology is worth what the deck says it is.
The real cost
Cheaper on paper, expensive on both lines that matter
The most expensive AI hire is not the senior one. It is the one who was cheap enough to approve quickly. Inexperience almost never shows up as a bad invoice. It shows up as revenue that never arrived and cost that never left, two or three quarters later, by which time nobody connects it to the decision that caused it.
| The decision | How it goes without experience | What it costs |
|---|---|---|
| Build or buy | Built, because building is what the team knows how to do | Eighteen months and a permanent maintenance burden for something that was available for a licence fee. Or the reverse: a licence for the one thing that was your actual advantage. |
| Which use case first | The one that demos best to the board | A year of engineering on a problem no customer was paying to solve. The top line never moves, and AI loses its budget for the next three years. |
| Vendor selection | The supplier who answered the tender most fluently | A roadmap shaped like that supplier's product, data in a format only they can read, and a renewal you cannot credibly refuse. |
| The data foundation | Skipped, so that something can be shown this quarter | Every model after the first one pays the same tax again. The second use case costs what the first one did, which is the moment most AI programmes quietly stop. |
| Is it good enough to ship | Accuracy measured on the test set and nowhere else | One visible mistake in front of a customer, and the business goes back to the spreadsheet. Winning that trust back costs more than the build did. |
| Who we hire | Three people at the start of their careers instead of one who has done it | Your strongest engineer becomes a part-time manager, your juniors learn judgement on your production system, and the thing you were short of, judgement, is still missing. |
| Governance and the works council | Left until the final review | Go-live slips a quarter because data protection and the works council see the system last, after the decisions that would have been cheap to change are already built. |
A senior day rate looks expensive next to a junior one. It looks cheap next to a wasted year. The point of experience is not that the work goes faster. It is that four or five of the decisions above get made correctly the first time, and each one of those is worth more than the whole mandate.
Why from outside
What an outsider can say, and an insider cannot
Promoting someone internally looks like the safe option. It carries a cost nobody names.
No colleague to sit opposite on Monday
An internal candidate has to live with every recommendation they make. Cancelling a project means cancelling a colleague's project. Moving budget means taking it from someone they will see at lunch. Telling the board the flagship initiative should stop is, for them, a career decision. For us it is an assessment.
No department to build and protect
A permanent AI leader is rewarded for growing a function: more headcount, more platform, more scope. We are not. There is no promotion for us to earn here and no side to take in a disagreement that started before we arrived. Our leaving is the point, not the risk.
Experience from more than one world
Twenty years across industries and across cultures, including delivering measurable impact inside a hyperscaler and inside a traditional German company. Those two places disagree about almost everything: speed, documentation, how a decision gets made, what a pilot is for. Most AI programmes come apart at exactly that seam, and having worked on both sides of it is what lets us tell you which of your constraints is real and which one is a habit.
Our founder has held this seat three times: building an AI consulting practice across EMEA at a global technology leader, as Chief AI Officer at a marketing analytics company, and setting AI strategy for an upcycling business. Different sizes, same job: work out which of the AI ideas on the table are real, build those, and leave behind a team that no longer needs us.
Read what that looked like →How we work together
Three ways in, each with a fixed monthly fee
Pick the one that matches the decisions in front of you. You can move up or down between them with a month's notice, and most clients start at the first one.
Direction
about two days a monthA senior sounding board for the people who have to decide. We keep your roadmap current, sit in the vendor meetings that matter, write down the build-or-buy calls with their reasoning, and tell you when an idea on the list is not worth the quarter it would take.
- Monthly session with the decision makers, plus availability in between
- The roadmap kept honest, with a written decision log of what is out of scope
- Build-or-buy and vendor assessments in writing
- Three months minimum, one month notice
Operating
one to two days a weekWe take the interim Head of AI seat. We run the weekly cadence, chair the steering committee, own one named outcome in your plan, hire and mentor the team, and report to your board in language it already uses. The mandate is written around the outcome, not around the days.
- One accountable name for the AI programme, from week one
- Steering committee, weekly cadence, and the escalations that come with the seat
- Hiring, structure and mentoring for the AI team, your successor included
- Six to twelve months, handed over when your own person runs it
Due diligence
two to three weeks, fixed scopeFor investors, acquirers and boards. We read the code, the data pipeline, the model evaluation and the team, and come back with a report that holds up in an investment committee: what is real, what is a wrapper, what the technology would cost to keep alive, and which three people the value walks out with.
- Technology, data and team assessed against the claims in the deck
- Key-person and dependency risk named, not implied
- A written report plus one session with the committee
- Fixed price and a fixed date, agreed before we start
Each package is a fixed monthly fee, agreed before anything starts, with a minimum term and a month's notice on either side. No day-rate surprises and no charge for the first conversation. We hold at most two leadership mandates at a time, because the seat is not something you can split six ways.
What you get
The work itself
Across the three packages, this is what the seat delivers. Which of it applies depends on where you are, and we agree that in the first two weeks.
- AI strategy and roadmap with named use cases and named owners
- Use-case prioritisation by return, and build-or-buy decisions you can defend
- Communication with the executive team, the board and the advisory board
- AI team building: hiring, structure and mentoring
- Technology stack evaluation and vendor selection
- AI operating model design: central, federated or embedded
- Technical due diligence for investors and acquirers
- A written handover, so the next person starts where we stopped
Straight answers
Three things we will not pretend
We also build. That is a conflict, so here is how we handle it.
We can build what we recommend, which means a build-or-buy call made by us is never neutral by default. So every one of them is written down with its reasoning, the alternatives we weighed and what would change the answer, in a form you can hand to any supplier, including one that is not us. Where we are the wrong builder for something, the document says so. If you would rather we never bid on the work we scoped, put that in the contract and we will sign it.
The EU AI Act does not require you to appoint an AI officer.
There is no AI equivalent of the data protection officer. The European Commission has said that no particular governance structure and no dedicated AI officer is mandated. What Article 4 requires is that the people who use AI are competent to use it. A clear internal owner is still a good idea, and we help you decide who it should be, but anyone selling you a legally mandatory role is selling you something the regulation does not ask for. None of this is legal advice: your legal team keeps the final say.
A part-time leader cannot carry a company-wide change alone.
We can hold the seat, make the decisions, build the team and get the first systems into production. Making AI part of how every department works needs someone who is there every day and who will still be there in three years. That is why the mandate is written to end, with your own successor in the chair and ninety days of us in the background while they run it.
Where the seat has been held
- Building an AI consulting practice across EMEA at a global technology leader: hiring the senior bench, writing down a delivery method and winning the first clients at the same time.
- Chief AI Officer at a marketing analytics company, where the job was to decide which of the AI ideas on the table were real.
- AI strategy for an upcycling business, at the size where the first decision is whether AI belongs in the plan at all.
- In each one the test came later, in what the organisation could still do once the seat was empty again.
3interim seats held
60%duplicated effort removed
20+years in the work
The three mandates and the duplicated effort removed are in the case study. The earlier work was done by our founding team in senior roles, not as ProDataAI client projects. Read the case study
Getting started
Three steps, and the first one is free
30 minutes
A first conversation
You describe the decisions in front of you and who owns them today. We tell you honestly whether this is the right offering, which package fits, or whether you need a build team instead of a leader.
2 to 4 weeks
An AI maturity assessment, if you want the ground mapped first
A fixed-price look at where you actually stand: skills, data, decisions and governance, with a prioritised roadmap at the end. Many mandates start here, and it is just as useful if you then decide to hire permanently.
From week one
The mandate starts
A written scope with one named outcome, the cadence agreed, and the steering committee in the diary. You see progress against the outcome every month, not a timesheet.
Not sure where you stand? The free 3-minute AI Readiness Check gives you a profile and a shareable result in the time it takes to read this page. Or see the AI maturity assessment and the AI Capability Foundation programme, which includes this seat.
Questions before the first call
What is the difference between interim and fractional AI leadership?
Cadence, mostly. Interim usually means near full-time for a fixed transition, for example while a role is vacant. Fractional means a smaller, ongoing share of the week, one to two days, for longer. We do both and the contract looks the same either way: a named outcome, a fixed monthly fee, a minimum term and a month's notice. The German market says Interim, so that is the word we use.
How quickly can you start?
Within weeks, not quarters. The first conversation costs nothing and usually happens the same week you ask. If the fit is right, a Direction mandate can start inside two weeks and an Operating mandate inside three to four, because the scope has to be written properly first.
How many days a month do we actually need?
Two days a month is enough to keep direction right and decisions made, if you already have people who can execute. One to two days a week is what it takes to own an outcome: run the cadence, hire, unblock and report. If you find yourself wanting five days a week from us, what you need is a permanent hire, and we will say so.
Does the EU AI Act require us to appoint a KI-Beauftragten or AI officer?
No. The AI Act has no equivalent of the data protection officer, and the European Commission has said that no particular governance structure is mandated. The duty in Article 4 is AI literacy: the people who use AI systems must be competent to use them. A named internal owner is still worth having, and we help you place it, but it is a management decision, not a legal obligation. Your legal team keeps the final say.
You also build AI systems. Is that not a conflict of interest?
It is, and we would rather name it than hope you do not notice. Every build-or-buy recommendation is written down with its reasoning and the alternatives, in a form any supplier can act on. Where we are the wrong builder, the document says so. If you want us barred from bidding on work we scoped, we will put that in the contract.
What happens at the end of the mandate?
Your own successor runs the operating model and we stay in the background for about ninety days, available but not deciding. You keep the roadmap, the decision log, the team and the handover document. A quarterly check-in afterwards is available if you want one, and plenty of clients do not need it.
Do you do technical due diligence for investors?
Yes, as a fixed-scope engagement of two to three weeks. We assess the technology, the data, the model evaluation and the team against the claims being made, and deliver a written report plus one session with the investment committee. It is the same judgement as the leadership work, applied from the outside and under a deadline.
What does it cost?
A fixed monthly fee per package, based on the cadence and the size of your organisation, agreed before anything starts. Tell us which decisions are waiting and we will quote the smallest package that covers them. The first conversation is free and there is no charge for the proposal.
Which decision is waiting?
Tell us what is on the table, who owns it today and what happens if it is decided wrongly. Thirty minutes, no slides, and an honest answer about whether this is what you need.
Get in touchor email ai@prodata.ai