Learning objective
Distinguish clear, complicated, complex, chaotic and confused components, then choose a response pattern that fits the causal conditions.
Why it matters
The same programme can contain routine, expert, emergent and crisis work. One management style across all four either over-controls learning or under-controls harm.
Core concept
Classification is provisional. The practical skill is to decompose mixed work and notice when the context changes.
Scenario
An insurer is introducing an AI claim assistant. Access control follows a tested procedure. Model integration needs specialist analysis. Customer response to explanations is uncertain. A live defect is sending some payments twice.
Worked decision
Use the procedure for access, experts for integration, safeguarded probes for explanations, and immediate containment for duplicate payments. Do not label the whole programme Complex.
Decision rule
Choose the least speculative response consistent with the evidence and harm boundary. If the component cannot yet be placed, keep it Confused and collect or decompose rather than forcing consensus.
Ethical boundary
“Probe” is not permission to experiment on people without consent, fairness review, safe fallback and stopping rules. “Chaotic” is not permission for indefinite emergency authority.
AI lens
AI may organise evidence and alternative classifications. Humans own the domain judgement, acceptable harm, decision authority and transition between modes.
Quick check
A pattern can only be understood after several bounded interactions, and no single expert can predict it reliably. What response fits?
A. Apply one best practice.
B. Commission a longer forecast only.
C. Run safe-to-learn probes, sense patterns and adapt.
D. Declare a permanent crisis.
Answer: C. The situation is Complex if patterns emerge through constrained action.
Next step
Use the Cynefin workspace on the incident-and-innovation portfolio case and record how each component will be reclassified.