EXL has launched what it says is the first specialised large language model for insurance to support claims reconciliation, data interpretation, question-answering, anomaly detection and chronology summarisation.

The model “outperforms leading pre-trained models in accuracy across a wide range of claims and underwriting-related tasks”, EXL says. “The EXL insurance LLM has been developed to address the highly specialised needs of the insurance industry, which has struggled to leverage off-the-shelf, general LLMs that lack fine-tuning with private insurance data and domain-specific understanding of business processes.”

Generic LLMs fail to address the nuanced challenges facing insurance companies during claims handling, leading to inefficiencies, high indemnity costs, claims leakage, longer settlement timelines and compliance risks, it says. 

“By focusing exclusively on insurance-related tasks, EXL has incorporated its deep knowledge of the insurance sector and highly tailored proprietary data to create the most accurate LLM in the industry.”

The model was built by EXL AI Labs using the Nvidia AI platform. Testing found it achieved a 30% improvement in accuracy on insurance tasks compared with generic LLMs, while ensuring “full regulatory compliance to the new AI laws applicable to Australia”, EXL head of insurance Australia and New Zealand Vilas Madan says.

“LLMs have the potential to transform insurance ... EXL developed curated data sets with insurance-specific tagging, labelling and question-and-answer pair creation to fine-tune an LLM using our Nvidia-powered [generative AI] platform,” he said.

The model can aggregate and reconcile hundreds of thousands of medical records, claims histories, handwritten notes, call logs and other claims and underwriting-related information.

Data and insights are automatically categorised and fed into a range of core functions, from claims adjudication and provider engagement to payment integrity and customer service.

This enables faster, more accurate negotiations with providers, assessments of anomalies and inaccurate payments, and more personalised, real-time conversations with customers, EXL says.