AI cost management has become a primary concern for executives. Some companies, such as Travelers Insurance, are adding flexibility to their model selection — while also building their own to mitigate costs.
In June, the insurer, which generated $49 billion in revenues in 2025 and employs 30,000, unveiled a proprietary large language model called TravelersLLM. The internal model delivers better results than commercially available AI models when it comes to insurance-related questions and is cheaper to run than frontier models, according to Mojgan Lefebvre, EVP, chief technology and operations officer at Travelers.
Although the company worked to build and incorporate a lower-cost internal model into its ecosystem, Lefebvre said the model works alongside frontier models and is not a replacement for the innovation and advancements offered by frontier developers. When Travelers’ applications handle a query, it’s either directed to TravelersLLM or a frontier model depending on the task, Lefebvre said.
“In general, whether AI or any technology investment, we have a very keen focus on ROI,” she said. “The approach to the TravelersLLM is, if you don’t need to be using the most expensive frontier model to answer the same question, we shouldn’t be. The TravelersLLM directly contributes to this ROI-mindedness and saying, ‘Use the right model for the right problem.’”
Along with reduced costs to run TravelersLLM, using a variety of models within the insurer’s ecosystem means that Travelers doesn’t depend on any single external model or vendor, she added.
“It absolutely improves the economics of using AI at scale for us across the enterprise, and it gives us strategic flexibility,” Lefebvre said.
Building TravelersLLM
The insurance industry is a specialized domain, and TravelersLLM filled a gap that general purpose models couldn’t, Lefebvre said.
TravelersLLM was trained on millions of company documents and evaluated against tens of thousands of insurance domain questions. It was also built together with subject matter experts in underwriting, claims, operations and service management, she said.
The model makes the institutional knowledge of its domain experts available to employees across the company. Applications access models, including TravelersLLM, through APIs.
“For most employees, the underlying models should largely be invisible, and they shouldn’t have to think about it,” Lefebvre said. “They should simply have the best intelligence available at the point of decision.”
Mapping tasks to an AI model is a trend happening across industries to address the increasingly high cost of always running queries through the most advanced AI models. To address customer’s cost concerns, Snowflake this week debuted an AI cost management feature that dynamically routes tasks based on cost and quality to an appropriate AI model. AWS and Oracle also implemented AI cost management features and tools this year.
AI’s future hinges on data
Travelers began modernizing its technology foundation more than 10 years ago, with data being a core part of the company’s strategy, Lefebvre said.
Even so, Travelers still operates some legacy systems, as the company’s goal wasn’t to modernize everything, but to modernize intentionally, focusing on where doing so would provide the greatest value to customers and employees, she said.
Data accessibility and usability continues to be a central focus area, along with the semantic layer, ontology and knowledge graph, which are critical to how enterprises connect data to AI systems, she said. Travelers currently runs 70% of its compute in the cloud, she added.
“There’s no way on Earth that we could be doing what we’re doing with AI today if we hadn’t been making the investments in modernization and focusing on data and connectivity of our data — its accessibility and quality,” Lefebvre said.
Lefebvre said as AI has evolved from a tool driving individual productivity to being embedded across company workflows, there remains a vast untapped future for the technology within the enterprise. TravelersLLM itself will serve as the foundational capability for agentic AI, according to Travelers.
“The next frontier is that true transformation where, with AI, you absolutely can start thinking about new business models, products and services,” she said.
