A strong first workflow is narrow enough to prove, important enough to matter and bounded enough to govern.
Five tests for a useful starting point
Choose a process with five practical characteristics.
- It repeats. The work returns often enough for a prepared starting point to save reconstruction.
- It crosses systems. Requests, documents, records, messages and tasks hold different pieces of the thread.
- It ends in an artifact. A person can inspect, approve, use or send the output onward.
- Authority matters. A named owner decides what is accepted, released, escalated or held.
- Memory changes next time. A decision, correction or unresolved issue affects the next run.
The strongest candidate is usually a modest process with a useful artifact that can reach someone quickly. Its value comes from making the next run sharper and easier to review.
How the workflow behaves
A workflow is one governed, repeated piece of work. A person can run it, an event can fire it, a schedule can time it or a webhook can call it from outside. The trigger begins the work; the workflow gathers relevant evidence and prior context, prepares the expected artifact, stops for its named approver and records the result.
New workflows rehearse first. A dry run lets the team inspect evidence, assumptions, missing fields and proposed output before external action. Changes ship as versions, so the team can compare a new version with its predecessor and reverse a change when needed. A critical missing input parks the run and asks for the required information.
Owners, approvers and collaborators belong in the workflow itself. Their roles make the stop condition visible and leave a useful record for the next person who needs to understand what happened.
What proof looks like
Proof is an operating fact with a status, a baseline and a boundary. A delivered client workflow can be described with the client’s identity protected. Customer zero shows a workflow running inside the business that built it, while independent customer validation remains a separate status. Architecture proof shows that the design and controls exist; a measured business outcome requires its own evidence.
For a first workflow, show the trigger, assembled evidence, prepared artifact, named approval gate, missing-input behaviour and state retained for the next run. Then show what the next run did differently. Fewer repeated explanations, earlier visibility of missing evidence or a more complete review can be meaningful operating changes when described with their actual scope.
Recognise the lane
Industries carry different nouns, while the workflow shape repeats. Energy operations connect asset events, warranty evidence, maintenance records and commercial obligations. Risk and controls connect regulatory watch, obligations, testing, assurance and board reporting. Construction and industrial work connect pursuits, issues and follow-through. Property and investment work connect briefs, evidence, advice and objections. Infrastructure work connects qualification, diligence and portfolio read-through. Professional services work connect client memory and obligations.
A lane is a family of workflows. Choose one bounded member first: regulatory watch or a control test in a bank, an event evidence room for an energy operator, or intake to reviewed draft for a research team. Each gives the organisation a concrete place to learn what should be retained and who should decide.
Map a useful starting point
Follow the context through the process: the request that arrives in one place, evidence checked somewhere else, the decision made in a meeting and follow-through held in one person’s head. Name the artifact that should come out, the person who owns the decision and the memory that should remain.
Then bring one workflow. Ask which repeated work should begin sharper next time. The workflow brief is a useful place to describe it.