Automation prioritisation matrix for repeated admin
A decision matrix that screens for rule clarity, exception handling, information sensitivity, ownership and reversibility before implementation effort.
INPUT ─── DECISION ─── OWNER ─── RECORD
Create one candidate per bounded process outcome
Specification detail
Write candidates as observable jobs, such as creating an approved record from a complete form, rather than broad departments such as ‘automate finance’. Name trigger, finish evidence and owner for each. Split a long process where one portion is deterministic and another depends on judgement. Capture current frequency and handling effort only from available records or a short observation period; do not invent annual savings. The matrix is a comparison aid, not a business case. A candidate with unclear boundaries is returned to mapping because any implementation estimate made before scope is understood will conceal assumptions.
Screen hard constraints before scoring convenience
Specification detail
Check whether the process has an accountable owner, stable rules, permitted system access, understood information categories and a safe manual fallback. Identify decisions affecting rights, safety, employment, finance or regulated obligations for specialist consideration. Credentials, uncontrolled sensitive data and unknown downstream actions are stop signs, not small deductions from a total. Also check whether a simpler form, template, responsibility change or existing application feature removes the friction. Screening prevents an attractive volume score from outweighing a fundamental governance or operational gap. Keep every failed constraint visible with its required decision.
Judge failure visibility, reversibility and acceptance
Specification detail
Ask how the team knows an action succeeded, how quickly a failure becomes visible and whether it can be reversed without harming a customer or corrupting records. A promising first automation has controlled test examples, an idempotent or duplicate-safe design where relevant, clear exception routing and a manual fallback. Define acceptance checks before ranking candidates: required fields mapped, unauthorised fields excluded, acknowledgement sent only after a successful record, and failed writes visible to an owner. If success cannot be observed, the process is not implementation-ready regardless of its apparent simplicity.
Practical prioritisation-matrix
Candidate comparison matrix
Apply constraints first, then compare operational evidence. Never let a total score erase an unresolved stop condition.
Bounded outcomeObservable trigger, finish evidence, process owner and excluded adjacent work.
Simpler remedyForm, template, rule, role change or existing feature considered first.
Rule clarityStable deterministic steps separated from policy, judgement and disputed definitions.
Variation evidenceObserved normal routes, incomplete inputs, corrections and exception categories.
Information constraintPermitted fields, prohibited fields, access owner and any specialist review needed.
Failure controlDetection, alert owner, retry or manual route, duplicate protection and reversal.
Acceptance readinessControlled examples and checks for normal, invalid, duplicate and system-failure paths.
Next classificationMap first, simplify manually, bounded specification, specialist review or hold.