Putting AI at the Core of Your Claims Ops

Artificial intelligence (AI) can provide limitless opportunities for an organization looking to improve its processes, but how can its core functions assist everyday claims operations? In this session at RIMS 2021, Sandip Chatterjee, SVP of Analytics at Marsh Advisory, discussed the current state of AI in claims, including the lack of widespread adoption and its continuous evolution.

AI in the World of Claims Management

There are a number of applications for AI in the claims process, but use varies based on a company’s processes. It can be used in the initial claims assignment process to drive the best outcome possible, finding the best provider for the claimant. AI’s use also extends to the direction of care for an injured worker, creating an optimal path for medical management.

In the life cycle of a workers’ compensation claim, AI can be used in triage and assignment, investigation and compensability, reserving, medical management, disability management, settlements and cost recovery. Additionally, AI can help with litigation and fraud management.

AI’s Limited Success

There has been a major push to adopt AI models given the time and effort put into its data models, but there is still some hesitancy. Less than widespread adoption can be boiled down to:

• Accuracy and reliability/consistency of model recommendations
• Availability and access to model recommendations
• Follow through and ability to learn/improve model recommendations

The New World of AI

In the last few years, AI has seen remarkable advances that necessitate a reset in its uses for claims management. Reassessing its applications could prove beneficial for any company looking to improve their claims handling process. A few of the critical changes to consider include:

• More data – Not just data within the claims management system, but data from video recordings and IoT devices that can help with claims processing.
• GPTs and deep learning algorithms – Last year, GPT 3.0 trained 175 billion parameters compared to GPT 2.0 which trained 1.5 billion.
• Modeling infrastructure – Amazon Web Services, Azure, and Google Cloud platforms offer solutions to build, deploy and retrain these models.
• GPUs – More processing power and on-demand availability mean a single GPU today can be included in the top 500 supercomputer list in 2008.

Making Next Generation Claims Management Powered by AI a Reality

When considering how to apply AI in your claims management solutions, think through your data strategy. Are you collecting your data from claims notes, IoT devices, emails, phone conversations or external sources? The ability to collect all the data and put it into a form to train the models is critical to improving processes. Investing in AI modeling and keeping up with changing trends will also be essential to continuous improvement in data processes. Lastly, invest in driving change management across the organization. Think through how AI can bring your company into the future and what the impact will be for all stakeholders.