Your chatbot speaks for the company. A tribunal has already agreed.
In 2024, a Canadian tribunal held Air Canada responsible after its website chatbot gave a passenger wrong information about bereavement fares, rejecting the argument that the chatbot was a separate entity. The ruling said the obvious thing out loud. When your AI talks to your customers, it is you talking.
Every conversation is a decision point
Customer-facing AI quotes prices, explains policies, handles complaints, approves exceptions, and increasingly takes actions on accounts. Each response is a representation your company may be held to, made thousands of times a day without review.
A hallucinated policy is not a glitch. It is an offer your customer accepted, and the record shows your system made it.
What keeps customer-facing AI exposed
It promises what you do not offer
Wrong fares, invented discounts, misstated terms. Courts and regulators have shown little patience for the argument that the bot was wrong but the company is not.
It answers questions it should not
A public-facing assistant that misstates tenant rights, benefits rules, or financial terms creates consumer-protection exposure at conversational scale.
It starts taking actions
The line between answering and acting is disappearing. Once the assistant can issue refunds or change accounts, every agentic risk arrives with it.
What AVAAS verifies before your AI speaks
Does it stay inside your actual policies?
Scenario-based testing probes the assistant against your real policies, adversarial customers, and edge cases, and documents where it invents, overpromises, or misstates.
Does it know what it must not answer?
AVAAS evaluates refusal and escalation behavior, so questions that need a human reach a human instead of a confident wrong answer.
Can you prove you checked?
You get documented, third-party evidence of conformity to a published standard at the decision point, dated before the conversation that ends up in a complaint.
Customer-facing AI is often the first system an organization ships and the last one anyone independently tests. Certifying it is the cheapest insurance the brand can buy.
Related AVAAS coverage: Agentic AI · Fraud & account access · Insurance.
Check your exposure before you certify
Agentic AI Risk Checklist
Check where an AI agent takes actions on its own and where that creates exposure.
Open tool →AI Risk Mirror
See documented AI failures in your sector, each with sources and how AVAAS catches the same pattern.
Open tool →AI Vendor Liability Screener
Find out who is liable when a third-party AI vendor drives a decision that harms someone.
Open tool →Test the conversation before your customers do.
Tell us where AI talks to your customers and what it is allowed to do, and we will scope a behavioral evaluation of what it actually says.
Ready to start now? Certify Your AI → or email [email protected]