Turn your Human + AI operating model into a living system.
Design explicit workflow contracts. Run governed transformation experiments. Use evidence to decide what should scale.
- Vendor-neutral
- Applicable across functions
- Designed for measurable outcomes
Four outcomes we hold the work to.
Every engagement is measured against the same small set of results, so progress stays comparable across teams and functions.
Faster Workflow Performance
Reduce cycle time and improve throughput.
Higher Decision & Outcome Quality
Improve accuracy and consistency of decisions.
Controlled AI Cost
Right-size usage with guardrails and visibility.
Evidence for ROI
Tie outcomes to decisions with verifiable evidence.
Four parts that turn an operating model into something you can inspect.
Each part is deliberately small and legible. Together they describe how people, agents, knowledge and decisions actually produce value — and what evidence is required before anything scales.
Human + AI Work Graph
Map how people, agents, knowledge, decisions and systems create value together.
Workflow Contract
Define intent, roles, decision rights and boundaries.
Experiment Engine
Run governed experiments with clear hypotheses.
Evidence & Value Ledger
Record results and decide what should scale or stop.
The Workflow Contract, field by field.
Ambiguity is where Human + AI work goes wrong. Eight questions, answered explicitly, before an agent is allowed anywhere near production work.
- IntentWhat are we trying to achieve?
- Human OwnerAccountable human role
- AI AgentsAgents and tools involved
- Decision RightsWho decides what and when
- Autonomy BoundaryWhere agents may operate independently
- Evidence RequiredWhat proof is needed for completion
- Cost GuardrailLimits for cost, time and resources
- FallbackHuman fallback and escalation path
A pilot runs in four stages.
Short enough to finish, structured enough to trust. Each stage produces an artefact the next one depends on, so nothing advances on opinion alone.
- 1
Baseline & Work Graph
- 2
Human–Agent Workflow Contract
- 3
Governed Experiment
- 4
Evidence & Scale Decision
And it keeps going after the pilot.
The same six movements repeat as the operating model evolves — observation feeding models, models feeding experiments, evidence deciding what survives.
- 01ObserveSee how work really happens
- 02ModelBuild the work and value model
- 03RedesignDefine better ways of working
- 04ExperimentRun safe, governed experiments
- 05EvidenceCapture results with proof
- 06EvolveDecide, adapt and scale
What gets measured, and against what.
Four views, each with a baseline to compare against. Without the baseline a number is a claim; with it, the number is evidence.
AI Consumption & Cost
Tracked- Tokens
- Model usage
- Agent calls
- Cost per outcome
- Baseline
- Observed
- Budget threshold
Operational Efficiency
Tracked- Cycle time
- Throughput
- Wait time
- Capacity
- Baseline
- Observed
- Target
Outcome Quality
Tracked- Accuracy
- Consistency
- Rework
- Quality risk
- Baseline
- Observed
- Target
Return on AI Investment
Tracked- Full delivery cost
- Attributable outcome value
- ROI development over time
- Baseline
- Observed value
- Target
Nothing here is a black box.
The method is meant to be inspected, argued with and adapted. That only works if the logic, the evidence rules and the governance are all visible.
It plugs into the systems you already run.
No rip and replace. The work graph reads from the tools your teams are already in.
- Work Analytics
- Business Systems
- Collaboration
- AI Platforms
- Governance Systems
And it is not tied to one function.
The same contract and evidence discipline applies wherever humans and agents share a workflow.
- Product & Innovation
- Technology
- Operations
- Customer & Sales
- Finance
- People & Organization
- Knowledge & Decisions
- Governance & Risk
Start with one workflow you already run on AI.
Not a transformation programme. One existing workflow, measured honestly, improved under guardrails — and a decision at the end that is backed by something other than enthusiasm.
- 01
Existing AI Workflow
Select one existing AI-enabled workflow.
- 02
Baseline Performance & Cost
Measure current performance, quality and cost.
- 03
Focused Optimization Pilot
Test changes with clear hypotheses and guardrails.
- 04
Evidence & Scale Decision
Decide to scale, revise or stop based on results.
What would you want the pilot to prove?
- Improve efficiency
- Improve quality
- Reduce token/model cost
- Demonstrate ROI
- Improve governance
A short conversation about the workflow, the baseline and what evidence would change your mind.