Automation Exception and Retry Planner

Use this planner before enabling automatic retries, resuming stored failures, or rerunning an entire workflow. It evaluates whether the proposed recovery crosses completed side effects and which controls remain missing.

Interactive decision asset

Automation Exception and Retry Planner

Define the recovery boundary before enabling another attempt. Values stay in this browser.

Integrity before uptime

01 Failure context

Workflow name Failed step Error classTemporary transport or 5xxRate limitInvalid dataAuthentication or permissionBusiness conflictUnknownObserved outcomeConfirmed failed before side effectSide effect confirmed completeUnknown - request may have completed

02 Completed side effects

Highest-risk prior effectNoneRead onlyReversible writeExternal record or messagePublication or notificationPayment, invoice, or inventoryStable source event ID?YesUnknownNoDestination idempotency support?ConfirmedUnknownNoDuplicate lookup before write?YesNoCompensation action available?YesNo

03 Proposed recovery

Recovery scopeRetry failed stepResume from checkpointFull workflow rerunManual reconciliationAutomatic attempts Increasing delay / backoff?YesNoManual fallback available?YesNo

04 Ownership and evidence

Recovery owner assigned?YesNoInput ID and prior result retained?YesPartialNoCheckpoint after completed effects?ConfirmedUnknownNo

This planner evaluates control coverage. It does not estimate platform charges or guarantee recovery.

One-page recovery brief

Lead intake to CRM and follow-up

High risk

Recommended pathReconcile first

Automatic retryBlocked

Full rerunBlocked

Release gateHold

Recovery decision

Controls to add or verify

Evidence to retain

Source event ID, failed step, error class, prior destination result, attempt count, timestamps, and final resolution.

Platform questions

A low planner score is not permission to retry money movement, publication, destructive changes, or an unknown outcome without destination-specific verification.

Copy briefPrint / save PDFReset example

How to use the brief

1. Model the highest-risk completed effect, not only the step that displayed the error.
2. Treat a timeout after a state-changing request as an unknown outcome until the destination is checked.
3. Use the same stable event ID for the first attempt and every retry.
4. Verify the exact connector can carry the required idempotency or lookup field.
5. Run one controlled failed-step and duplicate-delivery test with non-production data.

What the planner does not do

It does not connect to a platform, inspect logs, estimate retry charges, confirm API semantics, or guarantee exactly-once processing. A low risk score means the entered controls are more complete; it does not prove that the entries are true.

Read Retry-Safe Workflow Automation for the evidence and worked patterns behind the decision rules. Model usage separately in the Billing Topology Worksheet.