Hi all,
We're integrating the Agentless Email API with an ElevenLabs voice AI agent.
The goal: during a live call, the AI agent sends a transactional email (a link) to the customer and ideally confirms the outcome before the call ends. I'd like to sanity-check our approach for detecting the send outcome and ask whether there's a better pattern than what we've found.
Workflow I'm using (two calls):
- POST /api/v2/conversations/emails/agentless to send (returns a conversation ID)
- GET /api/v2/conversations/emails/{conversationId} to retrieve the conversation and check the outcome
What we've observed in testing:
- The only way we've found to distinguish success from failure is the presence/absence of an "errorInfo" field in the participants array.
- Success: no "
errorInfo" field present.
- Failure (e.g. invalid address / bounce): an "
errorInfo" field appears but only after ~20–25 seconds (consistent across 3 tests with an invalid address). I confirmed the timing by polling the GET endpoint every second until the field appeared.
- In the first few seconds, a successful send and a send that will later bounce look identical in the API response, because "errorInfo" isn't populated yet. The failure only becomes visible once bounce/error processing completes (~20–25s).
The challenge: That ~20–25s delay is longer, so the AI agent can't reliably confirm delivery in-call and act accordingly, if error occurs, It needs to hand off to a human, else end the call.
Need help understanding:
1. Is checking for "errorInfo" the recommended/reliable way to determine the outcome of an agentless email, or is there a better field/approach?
2. Is there an event-driven option (e.g. Notifications API / webhook / conversation event) we could subscribe to instead of polling, so we're notified when the bounce/delivery status is finalised?
3. Is the ~20–25s delay before a bounce surfaces expected behaviour, and is there any way to get a faster delivery signal?
Any guidance or better workflow would be much appreciated. Thanks!
#API/Integrations#ArchitectandDesign------------------------------
Ramsha Shaikh
Engineer, AI
------------------------------