> ## Documentation Index
> Fetch the complete documentation index at: https://www.voiceflow.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Return conversations to AI

> The AI resumes on a configured follow-up path after the human conversation finishes. Start with a working two-way handoff.

## What you will accomplish

The AI resumes on a configured follow-up path after the human conversation finishes.

## Before you begin

Start with a working two-way handoff. Use the Voiceflow web chat or a channel that implements the return-to-Voiceflow request. A custom interface must handle the handoff lifecycle; receiving a closure event alone is not a universal promise of automatic resumption.

## Configure the return

<Steps>
<Step title="Create the follow-up branch">
Connect the handoff return path to a short follow-up such as “Is there anything else I can help with?” Keep it separate from the failed-handoff branch.
</Step>
<Step title="Deliver the end event">
For custom connections, translate actual provider closure into `conversation.ended`. Send it only for the active mapped conversation. An assignment change, internal note, or temporary disconnect is not closure.
</Step>
<Step title="Resume through the channel">
Test the channel’s end-of-handoff behavior. The runtime handles a return-to-Voiceflow request and can add a summary of the recorded human exchange to AI context. SMS has a server-side return flow; custom interfaces must implement the appropriate request.
</Step>
<Step title="Test a fresh and a resumed turn">
Have the human resolve the test issue, end the conversation, and continue as the customer. Confirm the AI follows the return branch and does not claim it cannot see the human exchange.
</Step>
</Steps>

## Verify the result

Human replies stop after closure and the AI continues only on the intended path. Replayed end events must not restart the flow repeatedly. If you want the session to end entirely, leave the return disabled and verify that behavior too.