A pioneering approach to combining established improvement discipline with practical AI. Lean decides what should remain. Six Sigma supplies evidence. Agile sprints deliver learning. AI supports or automates the right work.
Discover · Eliminate · Redesign · Augment · Automate · Control
A practical extension of Lean Six Sigma discipline: define the problem, measure performance, challenge the causes, improve the process and sustain the gain. AI is selected where the evidence supports it. Augmentation and automation are options, not mandatory steps.
Understand the process before choosing the technology.
Map the customer need, workflow, demand, data and constraints. Establish the baseline for cost, quality, cycle time and human effort.
Output: current-state map and agreed success measures.
Remove work that should not exist.
Challenge duplicate entry, unnecessary approvals, avoidable hand-offs, rework and reporting. Remove waste before making it faster.
Output: waste-removal priorities and a simpler process.
Build a better future-state process.
Clarify ownership, decision rules, standard work and exceptions. Compare process change, conventional automation and AI against the same business need.
Output: future-state design and a justified intervention choice.
Help people make better decisions.
Use suitable AI to support analysis, knowledge retrieval, drafting or triage. Define what people must check and how output quality will be tested.
Output: an augmentation use case with human review and acceptance criteria.
Pilot appropriate work with clear boundaries.
Automate stable tasks where justified. Define exception handling, escalation and fallback. Coordinate specialist technical implementation where needed.
Output: a contained pilot with a named owner and go/no-go criteria.
Prove the benefit and keep it working.
Measure performance against the baseline. Track adoption, errors, operating costs and risk. Put ownership, training, monitoring and review into standard work.
Output: a benefits review and operational control plan.
Every stage asks the same question: does this improve customer outcomes and business performance?
This is an overlap map, not a one-to-one replacement for DMAIC. Discover includes Define, Measure and Analyse. Eliminate and Redesign draw on analysis and improvement. Augment and Automate are possible Improve interventions. Control sustains the result.
Measurement and risk assessment continue across the framework. Root causes are tested rather than inferred from an AI suggestion. Sprints can operate within several stages, and new evidence can send the team back to discovery or redesign.
| Lean-AI stage | DMAIC overlap | Lean, Six Sigma & Quality tools | AI contribution | Agile delivery | Decision / evidence |
|---|---|---|---|---|---|
| Discover | Define · Measure · Analyse | VOC / CTQs, project charter, SIPOC, Gemba, value-stream mapping, measurement-system analysis, Pareto | Summarise feedback or documents and suggest patterns; check source evidence and the measurement system. | Ordered opportunity backlog; discovery tasks and measurable acceptance criteria. | Agreed customer outcome, process baseline, boundaries and data readiness. |
| Eliminate | Analyse · Improve | TIMWOODS, 5 Whys, fishbone, A3 / PDCA, value-added analysis, demand analysis | Help classify work and identify duplicates; validate waste with the people doing the process. | Prioritise waste-removal changes before technical build tasks. | Unnecessary work removed; causes supported by evidence. |
| Redesign | Improve, informed by Analyse | Future-state VSM, flow, pull / Kanban, standard work, FMEA, mistake proofing, decision rules | Compare process change, conventional automation and AI; design reviews for output failures and exceptions. | User stories, Sprint Goal, pilot acceptance criteria and Definition of Done. | Future-state process, owner, exceptions and justified intervention choice. |
| Augment | Improve: design and test | DOE / experiments where suitable, hypothesis testing, sampling, error analysis, human factors | Support knowledge retrieval, drafting, triage or analysis. Evaluate outputs on representative tasks with human review. | Test a small increment, inspect quality and user feedback, then adapt. | Evidence that assistance improves the task without unacceptable error or review burden. |
| Automate | Improve: pilot and implement | FMEA, poka-yoke, standard work, integration tests, exception handling, change control | Automate suitable tasks only within approved boundaries, with escalation and fallback. | Usable pilot increments; Sprint Review and Retrospective; explicit release criteria. | Pilot meets quality, cost and operational acceptance criteria. |
| Control | Control; return to earlier phases when evidence changes | Control plan, SPC where suitable, run charts, process audit, CAPA, benefits review, visual management | Monitor output quality, changing inputs and drift; retain human accountability and a manual fallback. | Operational ownership, improvement backlog and recurring inspection and adaptation. | Sustained benefit, clear ownership, monitoring and a response plan. |
Order opportunities by value, evidence, feasibility and risk. Waste-removal tasks sit alongside process redesign and AI experiments.
Choose a contained outcome and define what must be tested. Agree quality, human review, exception handling and the Definition of Done.
Build and evaluate a small process change or AI use case. Use representative examples, including failures and exceptions, rather than a polished demo alone.
Inspect the result with stakeholders. Review the way the team worked, adjust the backlog and decide whether to continue, change direction or stop.
Sprint length and delivery support are agreed by scope. Scrum Sprints are one month or less; a full Scrum engagement includes its defined accountabilities, events and artifacts. A short workshop alone is not Scrum.
Assess the opportunity, remove waste and agree the smallest pilot that can prove value.
