Allstate

Objective

Make claims communications clearer and more empathetic at scale, by removing insurance jargon and freeing claims handlers from writing so they can refocus on checking.

The deployment

Allstate inserted a generative AI layer (OpenAI's GPT models) into the drafting of communications sent to policyholders during a claim. CIO Zulfi Jeevanjee states that nearly all of the communications representatives send to customers are now written by AI. The stated goal was writing quality: the old messages, even when framed by standards, carried a lot of insurance jargon and lacked empathy, which frustrated claims handlers and muddled the communication. The claims handler no longer writes; they review for accuracy before sending. At scale, this represents a workforce of about 23,000 representatives handling on the order of 50,000 customer communications a day. Allstate sums up the intent by saying that generative AI gives the customer the benefit of the doubt, where text full of jargon could come across as cold or suspicious.

Results Proof C

Deployment reported by name in the press (Futurism, Insurance Business) with direct quotes from Allstate's CIO on scope and operation. Facts stated by a brand executive, with no associated audited financial metric.

The internal detail is not public. Here is a proven approach that leads to the same result, to adapt to your stack.

1. 1
   Message generation AI

   From the file context, the GPT model writes the communication in plain language, without jargon.

2. 2
   Accuracy review claims handler

   The claims handler checks that the content is accurate and compliant before any sending.

3. 3
   Sending to the customer claims handler

   The approved message goes to the policyholder; the claims handler keeps control over sensitive cases.