Who leads your AI while the seat is empty?
AI programmes rarely fail on technology. They fail because no one owns what sits between the board decision and production. I take that ownership on an interim basis: with decision authority, a reporting line to the executive team, and the ability to defend the same decision in the boardroom and in an architecture review.
From boardroom to code.
For organisations that need operational AI leadership before the next architecture, vendor, or investment decision is made.
Discuss your situationThe market produces new promises every day.Only someone who has seen them in production can judge them.
Architecture, vendors, and governance are decided in separate rooms, under pressure from those with something to sell. Those decisions take effect only together, and only when someone owns them.
Three situations where leadership on an interim basis is the right answer
The AI leadership seat is vacant
The role is open, or a permanent search is under way. Active programmes, vendor contracts, and board dates do not wait for the new hire to start. I keep leading, hand over in an orderly way, and leave the role in a state a successor can take on.
Transformation has stalled after the pilot phase
There are prototypes, tools, and motivated teams, but no operating model. What is missing is the link between architecture, governance, evaluation, and ownership that turns something demonstrable into something that runs in production. I build that foundation and staff it until it holds.
The board and engineering read the same situation differently
In the boardroom a programme is on track; in delivery it is blocked. That divergence is a leadership problem, not a communication problem. I establish a shared factual basis and drive the decisions that commit both sides.
All three have one thing in common: they do not call for a recommendation. They call for a decision and someone accountable for it.
Interim Head of AI or Chief AI Officer
The scope follows the size and maturity of the organisation, not the title. Titles are secondary. What matters is that the function is filled effectively.
Interim Head of AI
- I own system architecture, operating model, and technology selection, including what runs in-house and what is bought.
- I build the team, develop its capability, and fill open key roles.
Interim Chief AI Officer
- I assess and negotiate platform and model vendors, and design exit capability deliberately rather than leaving it to chance.
- I prepare decision papers and report to the executive team, supervisory bodies, and auditors.
Governance and handover
- Ownership, approval paths, evaluation, and evidence, aligned with your regulatory context.
- An operational state, documented decisions, and a successor who can carry on without me.
Three mandates where I led
Anonymised, with industry and context reported accurately. The full accounts are published as case studies.
2020–2023, external lead, steered directly with the CTO and the founder
Fifteen years of accumulated C# and SAS silos separated research, portfolio management, and engineering. Releases every three months, no ESG capability, and the business dependent on IT for every report.
My decisions: Consolidation onto Python and open source, a central data hub with standardised models, replacement of proprietary reporting tools, agile delivery. That included the decision to replace positions within the ten-person engineering team where the shift was not carried.
At handover: Deployment cycle down from three months to three weeks. 44 people trained across four cohorts. A stack that supports traceability and regulatory requirements in a financial-services environment.
Full case study →2022–2024, three project phases, accountable for scope, prioritisation, and outcome
Three decades of accumulated data, 70 percent of processing manual, no reliable single source. Vendors were setting expectations no one internally could assess.
My decisions: A data strategy along governance, architecture, capability, and data culture. 48 individual interviews as the basis rather than theoretical assumptions. Selection of an AI use case on merit and feasibility, not on vendor promises.
At handover: A 120-page implementation roadmap with timelines, ownership, and dependencies. A forecasting model at 90 percent accuracy that moved workforce planning from annual to continuous. The basis was 48 individual interviews across the organisation.
Full case study →2019, strategic concept and architecture, iterative over several months
10,859 research documents spanning three decades, unstructured and confidential. Cloud and API use were ruled out.
My decisions: Classical NLP methods rather than a language model, deliberately: reproducible, traceable, no running costs. Fully on-premises, with an architecture that keeps later extension open.
At handover: Nine thematic clusters, research in seconds instead of archive work. Duplicate research became avoidable. The client went on to extend the solution with its own budget.
Full case study →How to tell early whether this fits
- At least three days a weekFrom three days a week, leadership accountability can be held effectively: present in decisions, in the team, and in the bodies that carry them. For smaller engagements, advisory work is the better format.
- A reporting line to the executive teamWith decision authority and direct access to leadership, the role becomes effective. That is what counts, more than the title.
- On site, particularly at the startI work on site in the early phase. Bringing people with you happens in the room, not on a video call. The technical shift is manageable; habit, relearning, and acceptance take longer than a project plan allows. Later the on-site share follows the phase and the confidentiality involved.
- Availability and startNew mandates are possible at short notice. We settle the earliest start and the exact scope in a first conversation.
Leadership accountability in recent mandates, combined with technical substance
Units of five to fifty people led
Across recent mandates I have led teams of this size functionally and as a people manager, including replacing positions where a shift was not carried. The reporting line ran directly to the executive team, the CTO, or a division head.
Ownership in delivery
Scope, prioritisation, and outcome of transformations, together with architecture and vendor decisions taken alongside the CTO and the executive team.
From boardroom to code
I defend the same decision in the supervisory meeting and in the architecture review. In interim mandates that dual fluency is the real bottleneck: speaking only one of the two languages costs time in translation.
Sovereignty is the ability to switch
Whether a component runs in-house or is bought is a pragmatic call; where proprietary components fit better, I make the dependency explicit. What matters is that the alternative stays real: a credible exit option changes every negotiation, from licence price to operating terms. Open source keeps that option open, and brings auditability and data sovereignty with it.