#In conversational AI, is it possible to detect the caller’s phone number and customize the welcome

1 messages · Page 1 of 1 (latest)

sly sableBOT
#
AI-generated Answer for swarek.eth

Yes. For inbound phone calls you can read the caller’s number (E.164) and set the agent’s language and greeting before the call starts:

  • Twilio inbound calls: ElevenLabs makes a pre-call webhook that includes caller_id. Your webhook can return conversation_config_override to set agent.language and agent.first_message (e.g., map +1 -> en, +33 -> fr). Enable overrides for Language/First message in the agent’s Security tab. (elevenlabs.io)
  • SIP trunking: caller_id is also exposed (via headers) and becomes available in the pre-call webhook and as a system dynamic variable. (elevenlabs.io)
  • System variables: system__caller_id can be used inside prompts/tools if you prefer dynamic variables over explicit overrides. (elevenlabs.io)

Alternative to country-code routing: you can configure multiple languages on the agent and enable the Language Detection system tool so the agent auto-switches to the caller’s spoken language. (elevenlabs.io)

In short: parse the country code from caller_id in your webhook and return language/first_message overrides, or enable language detection for automatic switching. (elevenlabs.io)

Sources:
Twilio personalization | ElevenLabs Documentation
SIP trunking | ElevenLabs Documentation
Dynamic variables | ElevenLabs Documentation
Language | ElevenLabs Documentation
Twilio personalization | ElevenLabs Documentation

glass mulch
#

And is it possible to, between the pre-call webhook and the call, we do an API call for like a LLM to totally personalize the welcome message based on the datas we have on the user ?

sly sableBOT
#
AI-generated Answer for swarek.eth

Yes. ElevenLabs’ inbound Twilio “conversation initiation” webhook is designed for this. When a call comes in, ElevenLabs calls your webhook with caller details (including caller_id). Your endpoint can call any external services (e.g., your own LLM or database), then return conversation_initiation_client_data that overrides the agent’s first_message, prompt, language, voice, and/or supplies dynamic_variables. The platform fetches this during Twilio’s dialing phase, so the personalized welcome is ready when the audio connects. Make sure you enable which fields may be overridden in the agent’s Security tab and respond within a reasonable timeout.

Key details from the docs:

  • Webhook inputs: caller_id, agent_id, called_number, call_sid.
  • What you can return: conversation_config_override.agent.first_message, .prompt, .language, tts.voice_id, plus dynamic_variables (all variables defined on the agent must be provided). Optional custom_llm_extra_body is also supported.
  • Timing/UX: the data is fetched in parallel during call setup; ensure your webhook responds promptly to avoid delays.

So, yes—you can call your own LLM in the webhook and fully personalize the welcome message before the call connects, as long as you return the supported initiation payload within the allowed time.

glass mulch
#

Quand on fait un systeme multi agent en conversational ai (un agent router puis envoyer au bon agent), est ce que quand le routeur envoie au nouvel agent, le nouvel agent a le contexte complet de la discussion ?