Skip to main content
System actions are built-in tools you add with @ in instructions. They change how Typewise handles the conversation. They do not call an external integration and do not need setup in the Actions hub. You can add them in: Type @ in the editor, then pick a tool from the System group. Selected tools appear as visual pills. Handoff conditions fields (Specialist, company, or channel) are prose rules only. They do not use the @ tool picker. Use system actions in the instructions or guidelines fields instead.

​
Close conversation

Use Close conversation when the issue is resolved and the AI should end the ticket. Add @Close conversation for cases such as:
  • The end customer confirms their question was answered
  • A routine request was completed and no follow-up is needed
When the AI uses this action, Typewise closes the ticket. This is separate from the timer-based Automatically close tickets after setting on Chat and WhatsApp guidance, which closes tickets after inactivity.

​
Handoff to human agent

Use Handoff to human agent when the AI should transfer the live conversation to a Human Agent for further help. Add @Handoff to human agent when you want an explicit handoff step in the flow, for example after collecting details that a human must review. When the AI hands off:
  • The ticket follows your configured ticket handoff destination (Typewise, Zendesk, Salesforce, or Email)
  • On Chat or WhatsApp, the customer may see a paused or waiting experience while a human takes over
  • Human Agents see the handoff in the conversation timeline; see AI handoff requests
Prefer Switch to email & handoff instead when the follow-up will take hours or days and keeping a live Chat or WhatsApp thread open is a poor fit.

​
Switch to email & handoff

Use Switch to email & handoff when the AI should finish collecting details on Chat or WhatsApp, then continue the request over email with a Human Agent. This fits async work such as warranty checks that can take hours or days. A normal handoff keeps the live channel open and can show a paused-chat experience while the customer waits. Switching to email first is a better fit when the reply will not be immediate. Typical setup:
  1. Tell the AI which customer details to collect, including the customer’s email.
  2. When those details are ready, instruct it to use @Switch to email & handoff.
  3. On Chat or WhatsApp, the AI sends a short confirmation that the request will continue by email, switches the ticket to Email, and hands it off to a human. The customer does not get a “chat paused” hold experience.
  4. A Human Agent picks up the Email ticket, with collected details available in the customer details sidebar, and replies by email.
Behavior to know:
  • Works on Chat and WhatsApp. On Email tickets the action does nothing.
  • If the customer’s email is missing, the channel does not switch and this action does not create a handoff.
  • Switching to email is one-way. The ticket cannot switch back to Chat or WhatsApp.
  • This is separate from the timer-based Automatically switch to email channel after setting on Chat and WhatsApp guidance, and from a Human Agent switching channels manually in the composer.

​
Verify email address

Use Verify email address when the AI should confirm who it’s talking to. It triggers the same email verification as a Trusted only input, without waiting for one. Add @Verify email address for cases such as:
  • Verify every customer before discussing anything account-related
  • Verify at the start of the conversation (put the mention in your Chat channel guidelines)
What the customer sees on Chat: a verification prompt replaces the message box. The customer enters their email address, receives a one-time code, and types it in. On success the message box returns and the AI continues with the original request. Behavior to know:
  • Already-verified customers are never asked again. A customer who entered a code earlier, or whose identity your backend signed, passes silently.
  • On Email tickets there is nothing to prompt: a sender who replied has already proven they control the address, so the action confirms without sending anything.
  • After verification, the address is the conversation’s verified user_email. Every Trusted only input mapped to it runs without another prompt; see Security & verification.
  • The customer can correct their email address or request a new code, but can’t skip the prompt or keep chatting. Abandoning the chat leaves them unverified, and actions with Trusted only inputs never run for them.
  • Whether verification runs is up to the AI following your instructions; Typewise doesn’t force it. To guarantee identity on every chat, sign it with Chat identity verification.

​
See also