How it works
From business friction to a working AI capability.
Seven steps, each with defined inputs, defined work, a defined output, and an explicit decision about whether to continue. Nothing moves forward on enthusiasm alone.
01
Understand the business
Inputs
- Business goals and constraints
- System inventory
- Stakeholder availability
Work
- Interviews with leadership and operators
- Constraint and risk capture
- Scope boundary agreement
Output
A written statement of goals, constraints, and the workflows in scope.
Decision to continue
Continue when the scope and the success definition are agreed in writing.
02
Map workflows and friction
Inputs
- Process documentation
- Volume and cycle-time data
- Operator time
Work
- Current-state mapping with the people doing the work
- Handoff, delay, and exception analysis
- Baseline measurement
Output
Validated current-state maps with a measured baseline and named friction points.
Decision to continue
Continue when your team confirms the map reflects how the work actually happens.
03
Prioritize AI opportunities
Inputs
- Current-state maps
- Data and systems constraints
- Agreed scoring criteria
Work
- Opportunity inventory
- Scoring on value, feasibility, risk, data readiness, and adoption
- Sequencing into waves
Output
A scored opportunity inventory and a sequenced roadmap.
Decision to continue
Continue when leadership approves the first wave and its budget.
04
Design the solution
Inputs
- Approved wave scope
- Data access and permission model
- Quality expectations
Work
- Future-state workflow design
- Control design: review points, thresholds, escalation
- Evaluation criteria definition
Output
A solution design with the future workflow, integrations, controls, and evaluation plan.
Decision to continue
Continue when the design, its controls, and its evaluation criteria are signed off.
05
Integrate and test
Inputs
- Design sign-off
- System credentials and environments
- Representative test data
Work
- Build and integrate against your systems
- Test failure modes and edge cases
- Validate with the people who will use it
Output
A tested solution in a controlled environment with documented results.
Decision to continue
Continue when evaluation results meet the agreed threshold and users accept it.
06
Train the team
Inputs
- Tested solution
- Role definitions
- Usage and review standards
Work
- Role-based enablement on real tasks
- Review and escalation practice
- Champion enablement
Output
Trained users, written usage guidance, and a named internal support route.
Decision to continue
Continue to live use when owners, reviewers, and escalation paths are confirmed.
07
Measure and improve
Inputs
- Live usage data
- Evaluation sets
- Feedback and incident reports
Work
- Scheduled evaluation runs
- Quality, cost, and adoption review
- Backlog prioritisation with your team
Output
Periodic performance reporting and a prioritised improvement backlog.
Decision to continue
Continue, expand, adjust, or retire the workflow based on the evidence.
01
Understand the business
Inputs
- Business goals and constraints
- System inventory
- Stakeholder availability
Work
- Interviews with leadership and operators
- Constraint and risk capture
- Scope boundary agreement
Output
A written statement of goals, constraints, and the workflows in scope.
Decision to continue
Continue when the scope and the success definition are agreed in writing.
02
Map workflows and friction
Inputs
- Process documentation
- Volume and cycle-time data
- Operator time
Work
- Current-state mapping with the people doing the work
- Handoff, delay, and exception analysis
- Baseline measurement
Output
Validated current-state maps with a measured baseline and named friction points.
Decision to continue
Continue when your team confirms the map reflects how the work actually happens.
03
Prioritize AI opportunities
Inputs
- Current-state maps
- Data and systems constraints
- Agreed scoring criteria
Work
- Opportunity inventory
- Scoring on value, feasibility, risk, data readiness, and adoption
- Sequencing into waves
Output
A scored opportunity inventory and a sequenced roadmap.
Decision to continue
Continue when leadership approves the first wave and its budget.
04
Design the solution
Inputs
- Approved wave scope
- Data access and permission model
- Quality expectations
Work
- Future-state workflow design
- Control design: review points, thresholds, escalation
- Evaluation criteria definition
Output
A solution design with the future workflow, integrations, controls, and evaluation plan.
Decision to continue
Continue when the design, its controls, and its evaluation criteria are signed off.
05
Integrate and test
Inputs
- Design sign-off
- System credentials and environments
- Representative test data
Work
- Build and integrate against your systems
- Test failure modes and edge cases
- Validate with the people who will use it
Output
A tested solution in a controlled environment with documented results.
Decision to continue
Continue when evaluation results meet the agreed threshold and users accept it.
06
Train the team
Inputs
- Tested solution
- Role definitions
- Usage and review standards
Work
- Role-based enablement on real tasks
- Review and escalation practice
- Champion enablement
Output
Trained users, written usage guidance, and a named internal support route.
Decision to continue
Continue to live use when owners, reviewers, and escalation paths are confirmed.
07
Measure and improve
Inputs
- Live usage data
- Evaluation sets
- Feedback and incident reports
Work
- Scheduled evaluation runs
- Quality, cost, and adoption review
- Backlog prioritisation with your team
Output
Periodic performance reporting and a prioritised improvement backlog.
Decision to continue
Continue, expand, adjust, or retire the workflow based on the evidence.
Systems of record
- CRM
- ERP / finance
- Service desk
- Document store
- HR / ATS
Integration layer
- Authentication
- Permissions
- Data contracts
- Retries and errors
- Audit log
Workflow and AI-assisted steps
- Intake
- Retrieval
- Drafting
- Routing
- Human decision gates
Measurement layer
- Cycle time
- Quality and rework
- Exception rate
- Adoption
- Cost per run
The grid and the table below carry the same information. No hidden score is calculated.
Matrix read in plain text
| Opportunity | Value | Feasibility | Recommended next step |
|---|---|---|---|
| Inbound document intake | High | High | Controlled pilot design — bounded automation with an exception queue. |
| Cited internal answers | High | Medium | Knowledge-system assessment before any build. |
| Response drafting | Medium | High | Copilot pattern with review before send. |
| Case preparation briefs | High | Medium | Workflow discovery to confirm systems and access. |
| Reporting narrative | Medium | Medium | Queue behind a higher-value workflow. |
| Regulated advice | Low | Low | Not an AI decision. Keep with qualified people. |
Find the AI opportunities worth implementing in your business.
Start with a structured assessment of your workflows, systems, and data, and leave with a prioritized view of where AI can create real value.