← Message journalAI in support

An AI support policy your team can use

Define sources, private-data checks, handoffs and how corrections become guidance.

Message.com · September 23, 2026 · 2 min read

Write down the allowed work

Template: Message AI may answer questions about [products and services] using [approved knowledge sources]. It should ask a clarifying question when the request is incomplete and avoid making unsupported promises about [restricted topics]. Replace the brackets with your actual business rules.

Separate public answers from customer records

Template: General policies can be answered from published knowledge. Before disclosing [order or account details], the customer must complete [the supported verification process]. An email address supplied in chat does not establish account ownership.

Define handoff conditions

Template: Bring in [team or role] for [sensitive requests, exceptions or explicit requests for a person]. Give the customer a short explanation and the receiving employee a summary. If the team is unavailable, explain the approved follow-up route.

Review corrections

Template: [Owner] reviews proposed knowledge and guideline changes. Record the original question, corrected answer and source. Test the correction against the original request and related cases before treating it as ready for customers.

Keep the policy usable

Use specific examples from your business, review it when services change and remove contradictions. Tone guidance can help AI sound natural, but it must not override identity checks or make an unavailable action sound completed.

Keep going

Put the next step into practice, or ask us about your setup.