AI Customer Service Agent Development

Description

We sell a subscription product with about 12,000 active customers and a two-person support team. Nights and weekends currently mean an autoresponder that says we will reply on Monday. The goal is an agent that handles the out-of-hours load: subscription status, invoice copies, plan changes and the password-reset dance. Anything about cancellation or a refund goes into a queue for Monday with a straight answer about when a human will reply — we would rather set an honest expectation than have an agent negotiate a retention offer. Our Intercom history is 40,000 conversations. Use it to find the actual top intents rather than guessing from our website FAQ, which we know is out of date.

• Mine the Intercom history to establish real intent distribution before building anything
• Implement the self-serve intents against our billing and auth APIs
• Route cancellation, refund and complaint intents to a queue with an honest response-time message
• Hand over full context so Monday's agent does not restart the conversation
• Report weekly on deflection, escalation and any answer a customer rated wrong

• Agent live in Intercom for out-of-hours coverage
• Intent analysis report from the conversation history
• Escalation queue with context handover
• Weekly reporting dashboard