What This Looks Like
A small error appears early in an AI output, tool result, field, classification, route, or workflow step. Later steps reuse or build on that error until it becomes a larger failure, such as a wrong decision, broken automation, bad review result, or corrupted downstream artifact.
Why It Matters
Early AI errors can become more damaging when later steps treat them as stable facts. A small mistake may be easy to fix at the source, but much harder to unwind after it has propagated through tools, summaries, approvals, or generated records.
Structural Signal
An error propagates and amplifies across downstream structure. The issue is not only the original mistake; it is that the workflow allows the mistake to spread into larger effects.
Common Triggers
- Early outputs are reused without validation
- Downstream steps treat AI text as authoritative state
- A wrong field value is copied into multiple records
- Review focuses on final output but not intermediate assumptions
- Tool calls or automations consume unverified model output
- The workflow lacks checkpoints that catch small errors before propagation
When to Use This Issue
Use this Issue when a small AI or workflow error spreads through later steps and becomes a larger failure.
When Not to Use This Issue
Do not use this Issue when the error stays local and does not affect downstream work. Use it when propagation is central to the failure.