Customer service chatbot
A program that converses in natural language with policyholders to answer questions, take a claim notification or follow up a file, whose answers bind the company like those of a human adviser.
A chatbot is an interface that converses with the policyholder in writing or by voice, on a website, an app or a messaging service. Early generations followed rigid decision trees; agents based on language models understand free wording, keep the thread of a conversation and write natural replies. In insurance they answer questions about cover, collect the first details of a claim notification, report on a file's progress and route to an adviser, at any hour and at scale. Their main risk is legal as much as technical. An inaccurate answer about the scope of cover, a deadline or a procedure can be invoked by a policyholder who relied on it, and courts tend to hold that the company is responsible for information given by its agent as for that given by its employees. Added to this are the obligations to disclose the automated nature of the interlocutor under the European regulation on artificial intelligence, and the risk of hijacking through instruction injection. Prudent design limits the agent to what it can state in a verified way, with referral to a human for any position on cover.
In February 2024, a Canadian civil resolution tribunal ordered an airline to reimburse a passenger who had relied on its chatbot, which had wrongly told him a bereavement fare could be claimed after travel. The airline argued the chatbot was a separate entity; the tribunal held it was responsible for all information on its website.
Follow the research
One letter a week. First, the week's underwriting note, when one is published: a dated, argued decision, signed by Paul Bernardski. Then what the observatory published on cyber risk, political risk and insurance. No account, no tracking.
No spam, one-click unsubscribe, no data ever sold.