Design the AI operating model.Roles, workflows and decision rights, rebuilt around the work.
An operating model is how the work gets done and who decides. AI pays back when that changes. This is the design work that sits between choosing your priorities and building anything.
You probably need this if you are saying one of these.
- “We know where we want to use AI. We don't know how the organisation needs to change.”
- “The tools are in. The org chart, the processes and the meetings are exactly as they were.”
- “Three teams think they own AI. So in practice nobody does.”
- “We need to tell people what their jobs look like after this.”
Start with who does the work
The centre of an AI operating model is a decision most organisations never make explicitly: for each piece of work, who or what does it.
Some work stays with a person because it needs judgement or carries accountability. Some is done by a person with AI alongside. Some is handed to AI inside agreed limits, with an owner who can review and reverse it. And some should stop.
Make that allocation deliberately, workflow by workflow, and the roles, skills, systems and measures that need to change become clear. Leave it to chance and each team answers differently.
What an operating-model design decides.
The same questions are asked of every workflow in scope. The answers, taken together, are the operating model.
- 01
What the work is for
The value each workflow is meant to create, and what should stop before anything is automated.
- 02
Who does what
What stays with a person, what AI assists, what is automated, and what is delegated to AI within defined controls.
- 03
Who decides and who is accountable
Decision rights for each step, a named owner for each outcome, and how a decision can be reviewed and reversed.
- 04
What changes around the work
Roles, skills, systems, measures and governance, so the new allocation holds in practice.
- 05
How the value is realised
A value-realisation model, a roadmap and pilot specifications that link the design to the numbers in the business case.
What tends to go wrong.
Automating the current process
Automating a poor process preserves the wrong work. Redesign comes first.
Allocation by default
When nobody decides what stays with people, the tool decides, one team at a time.
A structure chart and nothing else
New boxes and reporting lines, with the workflows and decision rights underneath unchanged.
This is the Design stage.
Stage 01
Decide
It follows a Diagnostic or Deep Dive, which choose the opportunities the design will serve. The Deep Dive already includes an operating-model outline.
Stage 02
Design
Six to ten weeks of detailed design: journeys and propositions, workflows, decision rights, roles, and data, technology and governance requirements.
Stage 03
Deliver
The design ends in pilot specifications, so the riskiest assumptions can be tested first.
Stage 04
Embed
People are trained on the redesigned work, with owners and measures in place.
Operating Model and Transformation Design.
Six to ten weeks, for a fixed fee agreed once the scope is clear. It suits businesses that have already chosen their priorities. If you have not, the two-week Deep Dive at £12,500 chooses them and produces an operating-model outline as part of its output.
Engagement 04
Operating Model & Transformation Design
Fixed fee, scoped6–10 weeks
- Target customer journeys and propositions
- Workflows, decision rights and roles
- Boundaries between people, AI assistance and automation
- Data, technology and governance requirements
Also worth a look: AI Transformation Deep Dive (£12,500, 2 weeks).
Questions we hear.From people in this spot.
What is an AI operating model?
It is how work gets done and who decides once AI is part of the business. In practice that means workflows, roles, decision rights, the boundaries between people, AI assistance and automation, and the data, technology and governance those depend on.
How is this different from AI governance?
The operating model sets out how work is done and who decides. Governance sets the controls, accountability and oversight around it. They overlap at decision rights, which is why we design them together.
How long does operating-model design take?
Six to ten weeks, depending on how many functions and workflows are in scope. The fee is fixed once the scope is agreed.
Do we need a Deep Dive first?
Only if the priorities are not yet chosen. The design needs a small number of agreed opportunities to serve. If you have them, with value cases, the design can start.
What do we end up with?
Target journeys and propositions, workflows, decision rights and roles, the boundaries between people, AI assistance and automation, data, technology and governance requirements, a value-realisation model, a roadmap and pilot specifications.
Related reading
Use caseAI governance
Accountability, Consumer Duty, UK GDPR and the EU AI Act, built into the design.
TeamOperations & customer service
Redesign before automation, what stays with people, and capacity versus cash.
TeamProduct & technology leaders
The product operating model, the AI-era lifecycle and the build decisions that follow.
Start with a conversation.Thirty minutes, free.
Tell us what you are trying to decide. If we can help, we will say how and what it would cost. If we can't, we will say that too.