Spotlight: How agentic AI is redefining claims transformation

From complex documentation to final decisions, agentic AI is transforming how claims are handled. But instead of replacing people, it works alongside them, argues Grady Behrens at Shift Technology.

For years, claims transformation has had straight-through processing (STP) as a key performance indicator of success. 

This, despite many claims and IT leaders knowing that only a minority of claims can be automated in this way, and for good reason: the specifics of individual claims – combined with document review, liability determination, third-party coordination, and policyholder communications – require advanced intelligence capabilities. 

As if that weren’t enough, an insurer’s commitment to fair determination, the risk of fraud, and other issues demand attention to individual claims.

In response, many insurers have pivoted to making manual claims handling more efficient. Generative AI, in particular, has become a powerful tool, as it can assist claims teams in their work through data extraction, summarisation, question and answers, and generating communication. And while efficiency is gained, it maintains ultimate responsibility for almost all tasks with the claims handler.

Underlying these implementations is the assumption that there is still an either/or proposition of competing responsibility, which limits the possible benefits. 

Either insurers can fully automate a claim, or the manual process must remain with improvements. Thanks to innovations in AI, namely agentic AI, insurers can change the model. Instead of an either/or proposition, a new way of thinking about the claims process is possible, with potential for massive gains in efficiency and customer satisfaction.

AI advancements meet policyholder expectations

The newest AI technologies, referred to as agentic AI, leverage large language models (LLMs) to learn, understand, and then complete even the most complex and challenging claims tasks, with a high degree of accuracy. 

In this way, many claims can undergo “STP” by AI agents across the entire claims process. More importantly, most claims tasks can be fully automated, whether gathering a claims declaration, analysing documents, determining relevant insights, checking coverage, recommending a decision, completing claim documentation, or drafting communication. This creates the possibility for massive gains in efficiency, speed, and accuracy in claims.

Yet, according to a recent report from Guidewire, UK customers still report a supermajority (66%) prefer phone contact to report a claim, and nearly half (49%) expect human intervention in any final claim decision. 

This comes at a time when insurers are increasingly grappling with recruiting and onboarding talent, leaving a shrinking but highly skilled force of claims handlers and adjusters strapped for time. In other words, personal interaction and supervision from claims handlers are prioritised by many policyholders, even as insurers struggle to onboard new team members.

With agentic AI taking on many claims tasks completely, human claims handlers can focus on what matters to policyholders: personal interaction and supervision to ensure accuracy and fairness.

Claims processing designed for AI and human collaboration

The combined advances of agentic AI, skills of human claims handlers, and expectations of policyholders can lead to claims transformation designed for agentic AI and human collaboration. With agentic AI serving as a collaborator, it can take full responsibility for tasks in a claim. 

For example, reviewing documents as they arrive, interpreting the relevant information, proposing an outcome, and preparing a response. However, much like in traditional team development, agentic AI can seek the advice of highly skilled claims handlers when needed, learning from human expertise over time to successfully complete more claims autonomously.

With agentic AI taking on many claims tasks completely, human claims handlers can focus on what matters to policyholders: personal interaction and supervision to ensure accuracy and fairness. 

This can be achieved by intentionally returning claim responsibility to handlers for the tasks of decision review and communication, if needed, based on the claim itself and customer preference. In this collaborative environment, future claims handlers can then leverage agentic AI’s learned expertise, with dynamic communication scripts and guidance, decision context, and details on alternatives, in order to more rapidly improve their own handling knowledge and maturity.

Measuring agentic claims transformation

In a process where claims are deliberately worked by both AI and human handlers, the KPI of STP is no longer the benchmark. 

Instead, insurers should consider the metrics which capture the value of all of the tasks automated in the claims process, whether straight-through, automated by AI with humans taking supervision and personal communication responsibility, or partially automated by agentic AI but transferred to expert human handlers to review and complete. KPIs used by Shift customers include percentage of total tasks automated, and can also include percentage of handler time spent with policyholders, or percentage of handler time reduced on tasks, among others.

Beyond automation, agentic AI and human collaboration can more directly connect to customer satisfaction metrics, since they reorient handlers and claims representatives to spend time with policyholders. Finally, the automation of tasks increases the speed of claims processing, improving both team efficiency and customer satisfaction.

Sponsored content