Message AI overview
Connect business knowledge, customer context and team handoffs across chat, tickets and phones.
Reviewed against the implementation · September 25, 2026
Start with approved knowledge
Create knowledge from public website content, selected connector sources or a business interview. Review generated drafts and publish the answers you want the customer-facing assistant to use. A connected account is not permission to publish every document in it.
Keep private context separate
Customer history and private order details have different access rules from public FAQs. Use the supported verification flow before retrieving private order information. An email match in an agent-only CRM panel is context for the employee, not proof that a visitor owns the record.
Configure the channel
Chat, ticket and phone AI have their own routing and activation settings. Configure the applicable site, allowed knowledge and fallback destination. A requested human handoff should retain the conversation and relevant context. Check the channel configuration before assuming every AI product is enabled.
Teach the company’s approach
Use internal coaching to propose a company guideline, then review and approve it. Guidance belongs to the workspace. An employee’s one-off instruction does not automatically become permanent policy for all customers.
Understand the price
Message AI is $5 per workspace per month and included in Business. It covers unlimited product AI use across chat, tickets and phones. Phone minutes and channel delivery fees remain separate. Task agents are distinct: one is included and additional agents are $5/month.
Check before rollout
Test a grounded question, a missing answer, an off-topic prompt, a request for a human and a private-data request. Evaluate the actual response and handoff, not just whether the API returned success. Customer-facing inference uses Message-managed infrastructure; consult the sub-processor list for other services involved in delivery and voice.