Stabilise an AI Content-Review Flow
Define the service and workflow, establish WIP and expedite governance, use flow evidence correctly and design a safe improvement experiment.
27 minutes
Scenario
An AI content-review service has 31 items in progress, three reviewers and daily urgent requests. Active review is short, but waiting, rework and lost evidence make completion unpredictable.
- Your role
- Service owner redesigning the flow system
- Method
- Kanban
Evidence pack
Demand
The service receives 18 standard and 3 claimed-urgent items per week.
WIP
Thirty-one items are marked In progress across drafting and review.
Time
Median active review is 96 minutes; median end-to-end cycle time is 9 working days.
Age
Seven unfinished items are older than 12 working days.
Quality
First-pass acceptance is 61%; rejected items receive new cards and lose earlier evidence.
Urgency
Any manager can apply the urgent label and no displaced work is recorded.
Constraints
- Do not rank individual reviewers.
- Preserve legal and owner approval gates.
- Do not expose confidential content on the shared board.
Case steps
Work through each prompt using the evidence pack. These guided cases support self-directed practice; server-scored attempts are not available yet.
Define the customer, work item, start/finish, demand types and sensitive-data boundary.
Design actual workflow states, pull/entry/exit policies and rework handling.
Set an initial WIP limit and governed expedite policy with rationale.
Define flow measures, an SLE and one improvement experiment with quality guardrails.