The Method
A method for redesigning work systems when AI exists.
Seven steps from understanding the real work to running a governed, improvable human–AI system. Not a framework for AI adoption—a discipline for transformation.
Understand the Work
Map work-as-done, not work-as-imagined. Before AI enters, you need to know what the system actually does — the real roles, real coordination, real expertise, real queues, and real exceptions.
What this step includes
- Work-as-done observation and mapping
- Role and coordination analysis
- Expertise and tacit knowledge identification
- Queue and exception mapping
- Decision points and authority paths
- Information flows and dependencies
Outputs
- Work system map
- Role and coordination diagram
- Decision inventory
- Expertise map
Decide What Should Change
Use the AI Work Fit Canvas to decide what should change, what should stay, and what AI should not touch. Not every activity benefits from AI. Not every decision should be delegated.
What this step includes
- AI Work Fit Canvas analysis
- Activity-level fit assessment
- Risk and consequence mapping
- Human judgment requirements
- Change vs. preserve decisions
- Boundary conditions
Outputs
- AI Work Fit Canvas (completed)
- Change/preserve/exclude decision log
- Risk assessment by activity
- Boundary definition
Redesign the Work
Redesign the work system — roles, coordination, authority, and expertise — for human–AI collaboration. Not a tools rollout. A system redesign.
What this step includes
- Role redefinition
- New coordination patterns
- Authority reallocation
- Expertise preservation and augmentation
- Queue and flow redesign
- Exception handling redesign
Outputs
- Redesigned work system map
- New role definitions
- Coordination protocols
- Authority matrix
Design the Human–AI System
Define the boundary between human and AI: what AI does, what humans do, how decisions get made, who answers when something goes wrong, and how escalation works.
What this step includes
- Task allocation (human vs. AI)
- Decision authority assignment
- Override and escalation paths
- Accountability mapping
- Feedback loops
- Error handling and correction
Outputs
- Human–AI boundary specification
- Decision authority map
- Escalation protocol
- Accountability assignment
Prove It Works
Build the evidence that the redesigned system produces better outcomes — not just faster tasks. Evidence survives examination, audit, procurement, and oversight.
What this step includes
- Outcome metrics (not just activity metrics)
- Before/after comparison
- Quality and accuracy measurement
- Human oversight effectiveness
- Exception handling performance
- Stakeholder feedback
Outputs
- Outcome evidence package
- Performance baseline and comparison
- Quality metrics
- Review-ready evidence file
Govern It
Governance as a property of the work system, not a layer stapled on top. The decision, not the system, is the unit of governance. Applicability, controls, evidence, and accountability.
The decision, not the system, is the unit of governance.
What this step includes
- Applicability determination
- Decision-control mapping
- Risk assessment and controls
- Evidence requirements
- Accountability assignment
- Monitoring and review cadence
Outputs
- Applicability profile
- Decision-control map
- Control register
- Evidence checklist
- Governance operating model
Operate and Adapt
Run, monitor, and improve. The human–AI system is a living system that learns. Performance degrades. Requirements change. The work evolves.
What this step includes
- Operational monitoring
- Drift detection
- Incident response
- Continuous improvement
- Adaptation to changing requirements
- Learning integration
Outputs
- Monitoring dashboard
- Incident log and response
- Improvement queue
- Adaptation roadmap
Why this method
AI transformation fails when it starts with the tool and skips the system.
Most AI rollouts bolt a model onto an existing workflow and call it transformation. The task runs faster. But the roles, coordination, authority, and expertise that produce the outcome stay the same — or get worse.
This method starts with the work system, decides what should change, redesigns for human–AI collaboration, proves outcomes, and governs it as a living system. The seven steps follow from one insight:
The unit of AI transformation is the outcome-producing work system — not the task where AI enters.
Next step
Ready to redesign a work system?
Start with the free Field Guide or talk about a work system that already feels the strain.