Your AI Has Two Answers, It Gave The Customer The Wrong One
Reported by Forbes1 source ↓
A KMS Lighthouse CEO argues that customer-facing AI can give wrong answers when company information conflicts, and that businesses should resolve those inconsistencies before deploying it.
Key points
- Forbes contributor Sagi Eliyahu says retrieval tools may find relevant documents but cannot determine which conflicting one is authoritative
- The article cites a 2024 Canadian tribunal ruling that held Air Canada responsible for its chatbot’s incorrect bereavement-fare guidance
- Eliyahu recommends assigning owners and review dates to customer-facing information and using the same verified answers for agents and chatbots
Details
Eliyahu, CEO of KMS Lighthouse, says AI systems can select and confidently present one of several conflicting company policies because they lack the informal judgment human agents use to resolve discrepancies. He argues the underlying knowledge, rather than the model alone, can determine whether customer-facing answers are accurate.
The article points to Air Canada, which a Canadian tribunal held responsible in 2024 after its chatbot gave a passenger incorrect information about applying for a bereavement fare refund. Air Canada argued the correct policy appeared elsewhere on its website, but the tribunal did not accept that defense. Eliyahu recommends consolidating duplicate information, assigning owners and review dates, and auditing knowledge before expanding AI to new topics or channels.
Written from coverage as of
Images
- Forbes