Background
Where this thinking comes from.
Five years across AI strategy, operating model design and delivery in regulated environments.
Photography · Tim van de Koppel
Scroll
01Independence
No platform to defend, no license to resell, no implementation practice to keep busy.
That is not a positioning line. It is the reason my advice on vendors, and on buy versus build versus redesign, can be trusted: there is no answer I am paid to prefer. I have no build team of my own. If you need one, I will help you select and brief it, contracted directly by you and with no fee to me from either side.
An AI-native practice, in the only sense of that phrase that means anything: it was built for this work rather than turned toward it. I have been in AI strategy since 2021, before the market arrived, and there is no legacy offering here that had to be repositioned.
The practice runs on the premise it brings to a client: people own the judgement, machines carry more of the work. It is a small demonstration of the thing I am hired to design.
02Where this thinking started · 2021
The line between assisted and agentic is drawn by identity.
Twelve enterprise interviews, four years before this became a board topic. What the research found has not aged.
What I added
A sub-facet of role identity: what an organization believes it has to keep doing itself in order to still be itself. In 2021 I called it safety identity.
Today it comes out of a stakeholder’s mouth as: we will do copilots, we will not do agents.
Architectural change is the one that moves the cost base. It is also the one identity refuses first.
MSc thesis, Strategy and Organization. Building on Kammerlander et al., 2018 and Henderson and Clark, 1990.
03The record
Five years, four things worth knowing.
-
1
Capgemini Invent, Global AI Team
- Co-architected the global AI Target Operating Model offering
- Worked directly with Director and VP-level stakeholders
- Authored internal work on agentic AI architecture
-
2
Client delivery
- Co-led a multi-phase AI engagement at a large Dutch insurer
- Automation strategy and TOM for a European IT services group
- Public sector, insurance, pharma and semiconductor
-
3
Capability building at scale
- Ran the weekly internal AI upskilling program
- Hundreds of colleagues; topics and speakers curated personally
- The 70% is not a theory to me
-
4
Foundation
- MSc Business Administration, Strategy and Organization
- 2021 thesis on strategic AI adoption; 12 enterprises interviewed
- Ongoing published point of view on operating model design
Client work as part of Atos
UWV · CIZ · VGZ · Vlaamse Overheid · European Commission
Client work as part of Capgemini Invent
ASML · AbbVie · a large Dutch insurer
Delivered as an employee of these firms, not as relationships of my own practice. The insurer stays anonymous and no engagement of theirs is described here.
04What I bring
Two things that rarely sit in the same person.
Operating model design as a discipline, and enough hands-on currency that the design is not two years stale.
Business foundation
The design discipline
- Operating model design as a craft, not as a slide format
- MSc Strategy and Organization; thesis on enterprise AI adoption, twelve companies interviewed
- Five years of target operating model work in regulated environments
- Starts from your P&L and your baseline, never from a use case catalog
Hands on the technology
So the design survives contact with it
- In AI strategy since 2021, before the market arrived
- Knows what each kind of model can actually carry, and what it cannot
- Builds with these tools daily: drafting, analysis, code and design, not only advising on them
- Co-architected a global AI target operating model offering
Most people have one of these. The engagement is worth having because you get both.
05My principles
Three rules I do not bend.
-
1
Business first
Start with the work that matters most.
-
2
Baseline first
You can’t improve what you don’t measure.
-
3
Measure what changes
Track value to its destination, not just activity.
The operating model is where a strategy stops being an intention.
06Why this work
The hours come back. And everything gets faster.
Most work holds more repetition than anybody would choose. Automating it properly is worth doing, and not because it is cheaper. It is worth doing because the hours come back, and what somebody does with returned hours is their own business.
The second half gets less attention and matters more. Somebody with these tools can test an idea in an afternoon that used to take a quarter. Multiply that across a research group, a company, an industry, and it stops being a productivity gain. It changes the rate at which anything gets tried at all, and that is the part that compounds.
None of which is a reason to be enthusiastic in advance. I know what AI fatigue looks like, and I have watched people quietly stop using tools nobody trained them for. Some things are still better made by hand.
My grandmother, Sita van de Koppel, painted this one.
The operating model is the difference
Between AI ambition and enterprise value.
A useful first conversation
Tell me which part of the business has to change next year, and I will tell you which of four things is standing in the way: a process problem, a data problem, a governance problem, or a sponsorship problem. That is usually enough to know whether it is worth working together.
Thirty minutes. No obligation.
Bring the target you have to hit next year.