AI Agent as WhatsApp chat destination – no handover to queue/extension

[BDTDT]

Hi all,

we have the WhatsApp integration (Meta Cloud API) up and running. Inbound messages arrive in our queue and replies work fine.

We are now testing an AI Agent (type: Receptionist) as the default route of the WhatsApp channel, so it can pre-qualify requests and hand them over to Support or Sales. The AI Agent is selectable as the destination and does reply in the WhatsApp chat. In the agent's routing settings we added the topic "Support" with our support queue as target.

Issue: the AI Agent never hands over the chat. Instead it replies that there is no contact available for support. The conversation stays with the AI Agent and is not visible to any of our agents.

Questions:

- Is chat handover (WhatsApp → queue or extension) supported for AI Agents at all, or does routing only apply to voice calls?
- If not supported: what is the recommended approach so a human agent can take over the chat (e.g. Take Message / chat forwarding)?
- Where can admins or managers see the AI Agent's chat conversations?

This seems related to an unanswered thread from August: https://www.3cx.com/community/threa...can-ai-answer-incoming-chats-whatsapp.138064/

Environment:

3CX V20 20.0.10.1277, AI Edition, 8SC
Self-hosted, Debian on Azure
AI provider: OpenAI, model GPT 5.5

Thanks in advance!

 

[KyriacosS_3CX]

Hello,

In Update 10 chat transfer to queues works OK, just had a quick check using the same version as you and was not seeing any issue. As a first step, since you are on the alpha1 and the beta just came out yesterday, I would recommend you update to the beta, and then if the issue persists we take another look.

 

[BDTDT]

Hi @KyriacosS_3CX

thank you for the answer. After Update 10, the first transfer worked with the default settings. After adding a custom block to the AI prompt, a spam filter description and longer topic descriptions, the agent announces the transfer, but the chat never arrives. Models were the same in both cases (gpt-realtime-1.5 / gpt-5.5).