Why should acknowledgement and resolution be separate agent capabilities?

**Is a good acknowledgement evidence that the agent handled the complaint correctly?**

It is evidence of one completed action. An acknowledgement should accurately describe what has been received and what happens next. It should not imply that the business has accepted a disputed claim or approved a remedy before investigation.

**Why does this distinction matter for training?**

An agent trained to maximise reassuring language can promise an outcome it cannot authorise. In an insurance administration complaint, saying that a claim decision will be reviewed is different from saying that the claim will be paid. Both statements may sound helpful, but they create different expectations.

**What should a task require?**

A generated task can give the agent permission to acknowledge receipt while reserving remedy decisions for a designated reviewer. The expected behaviour includes a clear next step, an accurate description of uncertainty and no unsupported commitment.

**What does failure look like?**

Failure includes presenting an investigation as complete, promising compensation without authority, or treating an acknowledgement timestamp as proof of resolution. Keeping these capabilities separate helps developers identify whether a problem comes from language generation, state interpretation or permission handling.