エラーと回復

Catch failure at the task and recover (error boundary)

Use when

• A task may fail and needs manual handling at the same point.
• Retry is not desired; recovery replaces reattempt.

Pattern with simple words

1. Start.
2. Service task: Process transaction, with an interrupting error boundary.
3. On error: User task Handle manually. End Resolved.
4. Normal path: Process transaction completes. End Completed.

Note: Recovery happens at the point of failure.

Other scenarios

• Finance: Invoice posting failure routes to manual entry.
• Healthcare: Insurance verification failure triggers manual follow‑up.
• Manufacturing: Automated QC failure triggers manual inspection.
• Retail: Online order submission failure routes to an agent.
• Telecom: Activation failure dispatches a field technician.

Retry a failed step before escalating

Use when

• Temporary issues often resolve on retry.
• You want a fixed retry count before escalation.

Pattern in simple words

1. Start.
2. Service task: Sync data, with an interrupting error boundary.
3. On error: Loop back to retry up to a defined maximum. If still failing, User task Escalate issue. End Escalated.
4. Normal path: Sync data completes. End Synced.

Note: Limit retries and log each attempt.

Other scenarios

• Finance: Payment gateway retry before manual billing.
• Healthcare: Lab data transfer retry before escalation.
• Manufacturing: Calibration retry before technician review.
• Retail: Supplier API retry before manual purchase order.
• Telecom: Provisioning retry before technician dispatch.

Graceful degradation with optional system fallback

Use when

• An optional automated step might fail.
• The process can continue manually without blocking completion.

Pattern in simple words

1. Start.

2. Service task: Automated quality check, with an interrupting error boundary.

3. On error: User task Manual inspection. End Checked.

4. Normal path: Automated quality check completes. End Checked.

Note: Manual fallback preserves completion.

Other scenarios

• Finance: Automatic report generation fails; produce a manual report.
• Healthcare: Insurance lookup fails; run a manual query.
• Retail: Product sync fails; update the catalog manually.
• Manufacturing: Scheduler fails; create jobs manually.
• Public sector: Automatic verification fails; perform manual review.