Probe Help article

Safe AI answers

Learn how to evaluate AI answers, reduce risky context, and escalate when a response is not enough.

prodigiaisupport
Visibility
public
Reading time
1 min read
Last reviewed
Reviewed 2026-05-13

Short answer

Treat AI answers as working guidance. Verify important details before you publish, pay, delete, invite, or change access.

When to use this

Use this article when Prodigi gives an answer that affects public content, billing, permissions, data, or operational decisions.

Requirements

Keep sensitive details out of prompts unless the surface explicitly needs them and you trust the channel.

How to evaluate an answer

Check whether the answer names the right product surface, uses your current account state, and gives steps you can verify. If the answer seems generic, ask a narrower follow-up.

Common problems

AI can miss hidden account state, old UI state, private permissions, or billing context. Do not rely on it as the only proof for high-impact actions.

Read Agent modes and context for prompt structure. Read Get help when escalation is needed.

Escalation guidance

When escalating an AI issue, include the prompt goal, the relevant article, and the safe part of the answer that looked wrong.

Was this article helpful?

Your feedback helps Probe improve articles, search synonyms, and support triage.