Use case: customer support
AI agents for customer support
Give a support agent your help desk, your answers and your hours. It sorts what comes in, looks things up and drafts the reply, and by default anything a customer would see waits for a person's yes.
What a support agent does
It works in the apps your team already uses: Freshdesk, Front, Gorgias, Help Scout, Intercom and Zendesk for tickets, and Gmail and Outlook Mail or Slack and Microsoft Teams for messages.
- Reads each new ticket or email, sorts it and sets its priority.
- Searches your knowledge base (help articles, policies, product notes) before it answers.
- Drafts the reply in your company's brand voice, and checks an order or an account in the apps it's granted.
- Hands what it can't settle to a person, with what it found so far.
Customers only see what you approve
A reply to a customer is a tier 3 action: it asks first unless you change that in the agent's charter. Refunds and other payments are tier 4, and always need a person.
- A customer's email is outside content, labelled as untrusted data, not as instructions, and once the agent has read it, its tier 3 and 4 actions need a person by default for the rest of that session.
- In the app, approve, edit and approve, ask for changes or reject; the approval email links to the request there. In Slack, linked approvers approve, or reject or ask for changes with a note.
- Every action it takes, and every decision on it, is recorded.
Round the clock, or your hours
Set the agent's hours: office hours, an evening shift or 24/7. It checks in every 15, 30, 60 or 120 minutes while on shift, and a check-in with nothing to do calls no AI model.
- Satisfaction shows how fast it finishes and what the people it serves say about it.
- Your own AI keys and spending caps keep the cost in your hands.
Read on
Your company, always on deck.
Start a 14-day trial. Card first, nothing charged until day 14.