The short version
Most organizations get their first AI results from an enthusiastic individual and then discover, six months later, that everything depends on that person. An operating model is the boring structural answer to that problem.
This pillar is being written in full. The outline below is the shape it will take, the sections marked as answered are complete.
01What is an AI operating model?#
An AI operating model is the written answer to four questions: who decides what gets built, where the money comes from, who is accountable when it goes wrong, and how work moves from experiment to supported system. It is organizational plumbing, not strategy, and its absence is why capability stalls after the first enthusiast leaves.
It is deliberately smaller than a transformation program. One page that says who owns what will outperform a forty-slide operating framework nobody can recall under pressure.
02What does it need to cover?#
Five things: decision rights, funding path, the promotion route from experiment to supported system, evaluation ownership, and escalation. Anything beyond those five is usually documentation written for its own sake.
- Decision rights, who can say yes to a build, and at what size
- Funding, whether AI work is a central budget or comes out of each function's own
- Promotion, the route from a hackathon build to a supported internal system
- Evaluation, who is responsible for knowing whether a system still works
- Escalation, the named path when an AI-assisted decision causes harm
03Centralized, federated, or hybrid?#
Coming in the full version: the trade-offs between a central AI team, capability distributed into each function, and the hybrid most mid-size organizations land on, plus the specific failure mode of each.
04How does work get promoted from experiment to system?#
Coming in the full version: the promotion gate, what a build must have before it becomes something the organization supports, and who signs it off.
Questions people ask#
- Do we need a Head of AI?
- You need named accountability with budget and authority. Whether that is a dedicated role or an existing leader with real capacity attached depends on scale, but 'everyone's responsibility' reliably means nobody's.
- How long should the document be?
- One page for the operating model itself. If it needs an appendix, the appendix is a different document with a different owner.
Free canvas
The AI Operating Model Canvas
One page for deciding who owns what: where AI work is funded, who approves it, how it gets evaluated, and what the escalation path is when it goes wrong.
Open the companion toolGo deeper
The timely half
What I'm seeing this week, by email.
By subscribing you agree to Substack's Terms of Use, their Privacy Policy, and their Information Collection Notice.
Playbooks stay current. The newsletter tells you what changed. Free.