If generative AI is on everyone’s lips a year after the launch of ChatGPT, other artificial intelligence technologies make it possible to respond to existing problems, such as the management of incoming emails in insurance.
The wholesale broker AssurOne has initiated the first stage of a project aimed at speeding up the processing of these messages. In volume, these represent several hundred thousand emails each year.
Little training data needed
This entire flow will not be subject to automation from the perspective of augmented managers. However, AI can already generate gains deemed significant, particularly in terms of customer satisfaction and productivity.
Based on a market study, AssurOne selected the InboxCare solution from Golem.ai. This has at least one particularity compared to traditional AI technologies based on machine learning.
The publisher thus claims “its ability to be deployed quickly.” To meet this commitment, Golem.ai indicates that it requires less data during the configuration phase.
“Only 187 emails were needed to configure the solution compared to several thousand or even hundreds of thousands of emails in the case of machine learning and generative AI solutions,” he specifies.
40,000 emails processed in phase 2
However, this reduced sample is also due to the current scope of the deployment, which concerns only part of the services (“Stellantis” and “Auto Constructors Management”). It will be expanded, in batches, later.
Thus, InboxCare is involved in the analysis of 9,000 emails and attachments each month and concerns a user population of 40 managers. The second stage will involve 40,000 emails and attachments and 135 employees.
In terms of functionalities, the AI tool analyzes four types of messages (request for information statements, request for green cards, various requests, emails with attachments).
Concerning attachments, eight categories are covered (registration card, special conditions, SEPA mandates, driving license, identity document, information statement, RIB, various attachments).
Claimed choice of responsible AI
To speed up message processing, “InboxCare assigns a contract number associated with each email and generates a draft response or an automatic message for each request.”
For AssurOne’s operations director, Corinne Le Lann, these automated tasks facilitate the verification of information contained in files, their validation, such as customer reminders.
“Analyzing the messages sent to us and their attachments using French AI, which is explainable and economical in energy resources, is therefore an additional solution to facilitate the daily lives of our teams,” she declares.