Intercom Support Agent

Half of live chat is the same handful of questions, and the other half is the reason people still need to be reachable. This agent takes the first half instantly, in a sentence or two, using only the playbook you wrote, and refuses to guess at the rest. Anything about money, access, or a customer losing patience gets an acknowledgement, an internal note with the context, and a Slack ping, so a human picks it up already knowing what happened.

What people use it for

• Instant answers to repeat questions - A small set of questions makes up most of a chat inbox, and each one still costs an agent a context switch. Those get answered in seconds without anyone opening Intercom.
• Cover the hours you are asleep - Customers in other timezones get a real answer instead of a form and a wait. Anything the playbook does not cover is handed over with a note, ready for the morning.
• Handovers that carry context - When it escalates, the note already says what the customer wants, what was checked, and what is blocking. That is the difference between picking up a conversation and restarting it.
• One playbook, one answer - Chat is where improvised answers become promises. Grounding every reply in a written playbook keeps limits and pricing consistent, and keeps unwritten promises out of the thread.

Before you fork

What do I need to connect before it runs?

An Intercom connection that can read conversations and contacts and write replies and notes, plus the Slack channel where handovers should land. You then write two things: the playbook of answers you are happy for it to give, and the rules for what a person must own. Start the playbook narrow, since it only replies to what the playbook covers, and widen it once you have read a week of transcripts.

How do I stop it answering things it should not?

Anything outside the playbook is a handover by design, so the playbook is the whole control surface. Money, refunds, account access, cancellations, and angry customers escalate regardless of what the playbook says, and it is barred from promising a refund, a discount, a callback time, or a fix date.

What does it cost to run?

One agent run per conversation event it receives, so cost follows chat volume. Escalations are cheaper than answers because it stops after a note and a Slack line rather than composing a reply. If your Intercom fires on more event types than you want triaged, narrow the trigger before going live.