HandsOn AI Operating Model. Two layers. Six domains. One core.
The model describes how an organization has to be built for AI to work in it. It carries Galbraith’s Star Model, the center-led structures of Kates and Kesler and Worren’s axiomatic design into the context of AI.
Click on a domain. The model in detail.
Leadership sets the foundation. Employees experience the guardrails every day. At the center sits the Human-AI Interface, connected to every domain.
Strategy & Value
Strategy & Value Architecture
Where does AI pay off?
Where does AI create competitive advantage, and where only efficiency? Use case portfolio, build, buy or partner, the cost of autonomy and what stays at the core.
Organization
Organizational Structure
Who is accountable for AI?
Central center of excellence, federated model or center-led hybrid with an AI hub and embedded AI leads. Reporting lines, AI owner and AI steward.
Governance
System Governance
Which rules apply?
The right governance framework, risk tiers under the EU AI Act, named accountability per system and governance across the whole lifecycle. Built on proportionality.
Decisions
Decision Architecture
How are decisions made?
The decision rights registry: for each decision type, which role decides at which autonomy level and who may change that assignment. Escalation paths are documented.
Processes
Process & Workflow Architecture
How does the work flow?
Three approaches side by side: AI overlay, AI-integrated redesign, AI-first. Which process gets which treatment, in which order, with which budget. The design tool for this is CLEAR.
Capabilities
Capabilities & Culture
What does the team need to know?
Learning paths for five audiences, from the C-suite to the front line. Deciding on a high-risk system takes different skills than everyday Copilot use.
Human-AI Interface
The organizational architecture for decisions and accountability when people and AI systems share responsibility for outcomes. For every AI tool it defines who decides, who is accountable, how the system learns and within which limits it works.
- Who decides?
- Who is accountable?
- How does the system learn?
- Within which limits?
The Human-AI Interface. The line between people and AI.
Every AI deployment creates a line: the point where human judgment ends and AI acts. Where that line sits, how it is governed and who answers for the outcome is a question of organization design.
- Q1
Who decides?
Who may accept AI output as a basis for action, at which autonomy level and under which conditions.
- Q2
Who is accountable?
Even when AI has acted, accountability stays with a person. It cannot be handed to a model.
- Q3
How does the system learn?
Whose judgment improves the system, how errors are detected and signals are captured.
- Q4
Within which limits?
Under which conditions AI may work without review, and which triggers bring a person back in.
Four autonomy levels. A design decision per decision type.
Every decision type involving AI sits on one of four levels. The level has consequences for cost and capability, so it is set deliberately.
AI recommends, people decide every time.
AI executes, people review samples and handle exceptions.
AI runs continuously within policy, people monitor at system level.
AI orchestrates multi-step workflows, people step in on escalations.
- AI executes
- Person reviews
- System-level monitoring
- Escalation
Four maturity stages. Measured per domain.
An organization’s maturity follows from how all six domains are designed. The white paper also describes which domains move first on the way to the next stage.
- Stage 0Unstructured
- Stage 1Augmented
- Stage 2Embedded
- Stage 3Agentic
Technology second. Deliberately without a technology domain.
Technology is a precondition for everything the model describes: strategy needs data, governance needs monitoring, decision rights need logs. We address these preconditions in every project, together with our partners. They belong in procurement, IT architecture and platform strategy.
Which LLM platform do we use?
How do we govern the decisions this system makes on our behalf?
Three questions. For your next board meeting.
- Q01 · Decision architecture
Are your autonomy levels formally set?
For which decision types is it defined at which level AI may operate, and who approved that assignment?
- Q02 · Governance
Do you know who is accountable for which system?
Who is registered as AI owner and AI steward for each AI system in production, and how often is this reviewed?
- Q03 · Process architecture
Which core process was redesigned?
Which core process was fully redesigned for AI, and where does AI only sit as an overlay on the existing flow?
