Smarter claims management, smoother settlements

The project began as an experiment in late 2023, leveraging the advanced capabilities of large language models (LLMs) to simplify claims processes. From a proof of concept to a functional product, Copilot's journey involved continuous feedback from users and integration into real-world claims processes.

"Originally we designed Insurance Copilot for complex claims, but users began using it for simple tasks like drafting emails, highlighting its versatility and time-saving potential," says Saral.

Unlike traditional systems with hardcoded processes, the Insurance Copilot operates flexibly, autonomously selecting the right tools to address user queries. Here's how it works:

1. Data gathering: Insurance Copilot collects data from claims and contracts, providing a summarized overview to help claims handlers quickly grasp the essentials. It also identifies the responsible claims expert.
2. Document analysis: It also searches metadata, analyses images and interprets individual agreements, comparing claims against policy details and ensuring the customer gets the full value of what they have paid for.
3. Decision support: The tool highlights discrepancies, compares claim information with invoices and incident descriptions and suggests next steps.
4. Streamlined communication: It can also draft contextually accurate emails to experts or involved parties, ensuring clarity, and saving valuable time.
   

By consolidating multiple data sources and automating repetitive tasks, claims handlers can process cases faster, leading to quicker, smoother settlements and improved customer satisfaction. It also eliminates many of the aggravations that previously confronted claims experts, such as time-consuming searches, data overload and repetitive administrative tasks.

Insurance Copilot helps to avoid unnecessary payouts by flagging critical factors that claims adjusters might otherwise miss. The tool enhances decision-making accuracy by providing data-driven insights, reducing the risk of overpayment and unnecessary costs, leading to better pricing for all customers.

While Insurance Copilot automates many processes, it adheres to a 'human-in-the-loop' approach. This ensures that critical decisions, such as coverage and payouts, must be made by skilled professionals. AI provides recommendations and insights, but the claims handler verifies the information and makes the final call, maintaining accountability and compliance.

Initially implemented for automotive claims, Insurance Copilot has since expanded to property claims. A specialized cross-functional team is now tasked with scaling the solution across other Allianz operating entities (OEs), including regions like Asia.

"In motor and property claims, the copilot has proven highly effective. We're still addressing challenges like working with scanned documents for health claims," comments Saral. "Claims differ vastly – motor, property, or health – each has unique nuances. We learned early on that one-size-fits-all doesn't work, so we fine-tune the tool through a reiterative process for each domain we expand into."

Preliminary assessments estimate that the Insurance Copilot could deliver a significant business impact and a very profitable business case, through improved productivity and reduced overpaying.

"Interest from other business units has grown significantly after our demos and pilot programs," Saral notes. "The demand indicates a clear need for such tools across Allianz operations."

With plans to expand its application beyond automotive and property claims handling, Insurance Copilot represents a significant step forward in leveraging AI to tackle the complexities of insurance processes. By combining automation with human expertise, Allianz sets a new standard for efficiency, accuracy, and customer service in the insurance industry.