If the AI service behind Aide's replies is down or overloaded, Aide now notices within a few failed requests and hands new conversations to your team at once, instead of making each customer wait minutes while it retries. When the service recovers, AI replies resume on their own.
How it behaves
- Chat, email and messaging channels: while the AI service is failing, a customer's message goes to your team within seconds, the same way as when Aide can't produce a reliable answer: the customer sees the handoff message from your Handoff settings, and your team gets an internal note, "Aide handed this chat off: Aide could not produce a reliable answer."
- The first messages of an outage: Aide gives up on a slow answer after 30 seconds and one retry (it used to wait up to 90 seconds, three times), so the few conversations that arrive as an outage starts are handed off within a few minutes at most, and the ones after them at once.
- Recovery: about every 30 seconds Aide tries one request; as soon as one succeeds, AI replies are back for the next customer message. Conversations handed off during the outage stay with your team.
- If you turned off handing off when Aide isn't confident (AI Agent > Handoff): the customer gets the usual request for more detail instead, and the conversation stays with Aide.
- Helpdesks where Aide always drafts: no draft is posted for a message that arrives during the outage, as when a draft can't be written for any other reason.
- Work nobody is waiting on (tagging, titles, Copilot drafts, knowledge gap checks) keeps its longer retries, and doesn't call the AI service while it's down.
Also in this change
- Aide's team is alerted when the AI service starts failing and when it recovers.