Public case / February 14, 2024

Moffatt v. Air Canada: when a chatbot's answer becomes your answer

Public case, used for illustration under our confidentiality policy. The organization named is not an Endorphin Global client. Facts come from the source cited below.

01 / What happened

What happened

02 / The control gap

The control gap

No control over what a customer-facing AI says. The chatbot's answers were not tied to the approved policy page, and nothing stopped it giving an answer the company had not approved.

03 / Our approach

How Endorphin Global would approach it.

01

Map

List every question customers ask the AI and where the correct answer lives today.

02

Redesign

Build an approved-answer knowledge base and a procedure to update the AI whenever a policy page changes.

03

Implement

Test answers before go-live, add a human escalation path, and show a clear AI disclosure.

04

Measure

Sample conversations each month and track answers that did not match the approved source.

04 / This week

Questions to ask your team this week.

  1. Which customer-facing tools give answers we have not approved word for word?

  2. When a policy changes, who updates the AI, and how fast?

  3. Can a customer reach a person when the AI is unsure?

05 / Related

Related services

Policy, procedure, and workflow redesign ↗

06 / Sources

Sources

  1. ABA Business Law Today ↗ (facts dated February 14, 2024)

Written by the Endorphin Global team. Last reviewed November 15, 2026. We re-check every source each quarter and update the page if the case changes.

Your next move

Facing a similar problem?

A short, free first call.
A clear answer on whether we are the right fit.