Task

Draft a reply from your own policy with AI

Answer a customer using the wording of a policy you paste in, so every promise can be traced to a line you already stand behind.

Work · Customer support L3 · AI drafts, you review

The shortest safe flow

  1. 1Remove sensitive details and decide whether this task is suitable for AI.
  2. 2Give the tool context, constraints and the shape of the result you need.
  3. 3Read, edit and check the result before it reaches another person.

A starting prompt

Answer using only the policy text provided. Quote the clause behind each statement. Where the policy does not cover the question, say that plainly rather than inferring.

A prompt is a starting point, not Evidence. Adjust it to your context and keep the original input nearby.

Human check before use

  • Point at the sentence in your policy behind every promise — if you cannot, delete it
  • No amounts or deadlines the policy does not state
  • Where the policy is silent, say so instead of guessing

Where this comes from

This page is a research draft. No real usage, success rate, elapsed time or edit rate has been measured.

Source type
Internal prototype content
Collected on
2026-08-11
Verification status
Awaiting human verification
Sample size
n = 0

Something here is wrong?

Point out a missing step or an unsafe claim. Your note is saved in this local Alpha session for review.