# Automating approve, warn, and deny | ReturnSift

Source: https://returnsift.com/blog/automating-approve-warn-deny.html

The three decisions in a modern returns flow have very different automation profiles. Approve should be almost fully automatic, because most customers are honest and speed is the experience.

Warn is semi-automatic. The system detects the borderline pattern and issues the policy notice, but the threshold and the wording deserve human review until you trust the calibration.

Deny should stay human-led for the foreseeable future. The system assembles the evidence and recommends; a person decides. That keeps the strongest action defensible and gives edge cases the judgment they need.

Getting the mix right is what makes automation pay. Approvals get faster for the honest majority, warnings go out consistently instead of when an agent has time, and denials arrive with a complete evidence file every time.

### Keep reading

[Return scoring: triage every request in seconds](return-scoring-triage.md)

[The signals that separate honest returns from abuse](signals-honest-vs-abuse.md)

[Returns fraud glossary](https://returnsift.com/glossary.md) · [ReturnSift vs Loop](https://returnsift.com/alternatives/loop.md) · [ReturnSift vs Happy Returns](https://returnsift.com/alternatives/happy-returns.md)

[Get a free returns fraud scan](mailto:hello@returnsift.com?subject=Free%20returns%20fraud%20scan)