How to disclose AI to customers without killing the conversation

Summary

Customer service teams may avoid disclosing AI because they fear customers will leave, but the post argues that unclear or delayed disclosure creates more frustration than honesty. It recommends a short first-turn greeting that says “AI,” explains what the agent can do, and offers an immediate route to a human, with renewed disclosure when a conversation moves to SMS or voice. Web chat, SMS, and voice examples show how to pair disclosure with an exit, while voice should disclose before the caller explains the issue and leave a pause. The post identifies burying the disclosure, using ambiguous labels, apologizing, over-explaining, and omitting escalation as failure modes. It proposes checking Twilio Conversation Intelligence transcripts alongside containment and escalation rates to assess wording changes, citing research on customer expectations and bypass behavior.

Takeaways

1. 1

   76% of consumers think an AI agent should identify itself at the very start of a conversation, while only 22% say agents do so immediately.

2. 2

   Effective disclosures name the system as AI, describe its capabilities, and provide an exit such as “agent” or “HUMAN”; terms like “virtual assistant” are described as ambiguous.

3. 3

   Twilio Conversation Intelligence can be used to count disclosure events by channel and turn, then compare them with containment and escalation rates before and after wording changes.