MassMutual: How a Flexible AI Strategy Delivered a 30% Productivity Boost

Insurance company MassMutual revamped its artificial intelligence adoption approach, prioritizing flexibility and short-term contracts. Instead of long-term commitments to vendors, the company built an infrastructure enabling rapid AI model switching in response to market trends. According to Sears Merritt, MassMutual’s CTO, the dynamic AI market demands readiness for swift changes.

The results of the new strategy are impressive: developer productivity surged by 30%, while contact center query processing time dropped from 10 minutes to one. Meanwhile, the cost per query fell from dollars to cents. The company also actively tests open-source solutions to avoid dependency on proprietary models and maintain cost control.

MassMutual categorizes AI adoption projects into two groups: tools to enhance employee productivity (e.g., Copilot*) and targeted initiatives to optimize business processes. All projects are initially evaluated against predefined success criteria, not by the number of implementations. The company also introduced a 'trust score' system, assessing model response quality based on employee feedback.

An interesting example is the model selection for the contact center. Employees were offered two versions: a fast but less accurate one and a more expensive but higher-quality option. Most preferred waiting a few seconds for a more accurate response, confirming that users are willing to sacrifice speed for quality if the difference is noticeable.

Currently, MassMutual is gathering detailed analytics on model usage to automatically distribute tasks among them in the future based on cost, quality, and user needs. This will optimize expenses and enhance AI infrastructure efficiency.