New reusable worksheet draft • Birkan episode planning pack
Episode segment: 07:00–11:30
Purpose: Check whether an apparent mismatch is a real discrepancy.
This worksheet starts empty. It contains no Birkan records or verified outcomes. Use actual approved records or clearly labelled demonstration data. It is not claimed to be an installed OpenAnalyst product template.
Fill in
- Business question: fill in; keep unknown when unavailable
- As-of time and source coverage: fill in; keep unknown when unavailable
- Stable order / item key and record granularity: fill in; keep unknown when unavailable
- Source A, refresh time, event time, status definition, unit: fill in; keep unknown when unavailable
- Source B, refresh time, event time, status definition, unit: fill in; keep unknown when unavailable
- Match rule and excluded records: fill in; keep unknown when unavailable
- Expected relationship between records: fill in; keep unknown when unavailable
- Observed discrepancy: fill in; keep unknown when unavailable
- Alternative explanations: fill in; keep unknown when unavailable
- Check performed and result: fill in; keep unknown when unavailable
- Confirmed finding or unresolved hypothesis: fill in; keep unknown when unavailable
- Person who verified and verification date: fill in; keep unknown when unavailable
Fit into the existing workflow
- Existing routine and tool: describe where people already do this work
- Trigger to investigate: An analyst starts a check, a supported refresh event occurs, or a configured schedule runs. Record the actual demonstrated mode.
- Repeated manual effort before: lookups, copying, checking, chasing, or preparation
- AI task to investigate: Repeat the defined source comparison and surface exceptions. Confirm the actual supported behavior.
- Person and decision: who reviews, decides, or continues the task
- Working destination: The analyst’s actual reconciliation or exception-review location. Record the actual destination.
- What changes for the team: new habit, training, ownership, or data-entry requirement
- Completion evidence: source-backed result, accepted record, or execution receipt
- Next run and exception handling: how it repeats; who responds when it fails
State whether the captured example is manually started assistance, a saved repeatable procedure, scheduled automation, or event-triggered automation. Show its actual behavior and do not infer background execution from a generated answer.
Question to adapt
Reconcile the selected records for order as of time. Match stable keys, record granularity, units, and status definitions before comparing. Show supporting rows and source freshness. Separate possible software or data defects from missing coverage, stale information, and legitimate timing differences. Label explanations as hypotheses until checked.
Check the result
A mismatch is not a software bug until its cause has been checked; show supporting and conflicting evidence.
Show on camera
Follow one task from its existing input through the AI step, the person’s decision, and the working destination. Explain what changed from the previous routine and show what proves completion.