The policy changes by conversation
People resolve the same case differently because the rule itself is unsettled. Automation would make the disagreement faster, not smaller.
DECISION GUIDE — AUTOMATION READINESS
Is this workflow ready for automation?
A workflow is ready for automation when it has a clear trigger, a visible finish, reliable inputs, explicit decision rules, known exceptions, and an accountable owner. If people still disagree about the policy or reconstruct the work from memory, fix the process before automating it.
/READINESS TEST05 GATES
- 01The work has a clear start and finish
- 02The required inputs can be found
- 03Routine rules are explicit
- 04Uncertainty has somewhere to go
- 05Someone owns the result
Clear answers across all five gates support a bounded pilot.
Do not automate a disagreement.
Software can enforce a known process. It cannot resolve unclear ownership, conflicting policy, or evidence that nobody trusts.
Each gate protects the next one.
A clear route needs more than repetition. It needs evidence, decision boundaries, exception handling, and ownership.
- 01 — BOUNDARY
The work has a clear start and finish
READY SIGNALA specific event starts the work, and the completed result can be observed without interpretation.
WAIT SIGNALThe workflow begins whenever someone notices a problem and ends when people stop asking about it.
- 02 — EVIDENCE
The required inputs can be found
READY SIGNALThe operator can name the records, fields, documents, and source systems needed for an ordinary case.
WAIT SIGNALCritical context lives in private messages, personal memory, or documents nobody can locate consistently.
- 03 — DECISIONS
Routine rules are explicit
READY SIGNALThe team can separate deterministic rules from classifications, commercial judgment, and regulated decisions.
WAIT SIGNALDifferent operators apply different policies, but the organization treats every result as equally correct.
- 04 — EXCEPTIONS
Uncertainty has somewhere to go
READY SIGNALKnown exceptions enter a review path with the evidence and owner needed to resolve them.
WAIT SIGNALThe proposed system must guess through missing data, conflicting rules, or unusual cases to keep moving.
- 05 — OWNERSHIP
Someone owns the result
READY SIGNALOne role owns the workflow definition, its success measure, and the decision to expand or stop automation.
WAIT SIGNALThe software team owns implementation, but nobody owns whether the operating result is correct or useful.
Some problems need repair before software.
A pause is useful when it prevents a broken rule or unreliable record from becoming a faster source of errors.
The source record cannot be trusted
Missing identifiers, duplicate records, and conflicting versions prevent the system from knowing which input should control the result.
Success has no observable definition
The team wants less manual work but cannot state the correct outcome, acceptable exceptions, or evidence needed after each action.
Prove one complete path before expanding authority.
The first version should finish one useful route and make every stop, review, action, and result visible.
- 01 — OBSERVE
Collect ordinary and difficult examples
Follow completed work, one exception, and one stalled case through the tools and side channels people actually use.
- 02 — DEFINE
Write the allowed path
Name states, evidence, permissions, review conditions, failure behavior, and the person accountable for the result.
- 03 — PROVE
Run one complete slice
Start in observation or draft mode. Compare the system with the operator before granting wider authority.
Questions that change the decision.
Readiness depends on the operating boundary, not the excitement around a tool or model.
Does a repetitive task automatically qualify for automation?
No. Repetition helps, but the task also needs reliable inputs, a clear outcome, and safe handling for exceptions and failures.
Can we automate a process that still changes often?
Automate only the stable parts. Keep changing policy in a configurable or human controlled layer until the organization understands it.
What if the workflow depends on human judgment?
The system can collect evidence, apply routine checks, prepare a recommendation, and route the final decision to an authorized person.
Do we need perfect data before starting?
No. You need enough reliable data to support one bounded path and a clear response when required evidence is missing or contradictory.
What is the smallest useful first automation?
Choose one frequent workflow with a visible start, a checkable result, a known owner, and enough examples to test ordinary and exceptional cases.
NEXT.01 — ONE BOUNDED ROUTE
Bring the workflow. We will test the boundary.
Share its trigger, owner, evidence, decisions, exceptions, and required result. That is enough to start a useful diagnosis.