Answer support tickets with AI

What it is

Most support queues repeat the same questions. A support AI agent answers the ones your help content already covers, drafts replies for the rest, and hands sensitive conversations to a person.

How it works

1. Audit the top ticket reasons and the articles that answer them. Write or fix the missing articles first (see a help center AI can use).
2. Connect a support AI to that content only. Anthropic’s guidance on reducing hallucinations recommends the same discipline: tell the model to use only the documents provided, give it permission to say it doesn’t know, and have it cite a source for each claim.
3. Set hand-off rules. Billing disputes, complaints, and legal or safety topics go to a person.
4. Review weekly. Read a sample of AI conversations and fix the content gaps they reveal.

Benefits

• Faster first responses, at any hour.
• Agents spend their time on complex cases.
• Content gaps become visible.

Limits and risks

• Wrong answers cost trust. Restrict the AI to sourced content and route uncertain cases to people.
• Tell customers they’re talking to AI, and how to reach a person. The European Commission says the EU AI Act’s transparency rules, in effect from August 2026, mean people using a chatbot should be made aware they are interacting with a machine.
• Watch outcome-based billing. On 2026-10-16, Intercom priced Fin at $0.99 per outcome (for example, a resolution or a configured Procedure hand-off) and $9.99 per sales qualification, with at most one charge per conversation and minimum commitments. Intercom counts a resolution when the customer confirms the answer helped or leaves without asking for more help, so set usage alerts and check what you are billed for.
• Region matters for agents. Intercom’s MCP server, which lets AI tools read conversations and contacts, supports US- and EU-hosted workspaces but not AU-hosted ones yet.