But as technology advances, a key question emerges: Can artificial intelligence (AI) enhance professional judgment, reduce errors, and improve efficiency in insurance decisioning?

The rise of large language models (LLMs) such as OpenAI’s ChatGPT has fueled excitement—and skepticism—about AI’s role in professional services. While these models have demonstrated an impressive ability to pass professional exams, draft legal arguments, and even assist in underwriting, their use in decision-making has been met with caution. AI systems, after all, are prone to inconsistency, bias, and even hallucination—issues that make them unreliable when it comes to high-stakes professional judgment.

So how can insurers harness AI’s power while maintaining trust, accuracy, and explainability? The answer lies in a hybrid approach—one that combines AI’s strengths in data processing with structured decision models designed to reflect expert reasoning.

The Challenge: Inconsistency in Human Judgment

Insurance decision-making involves a delicate balance between policy language, claims evidence, and risk assessment. Yet studies show that even experienced professionals can be inconsistent in their assessments.

A study of claims handling decisions across multiple insurance organizations found that even simple, binary decisions—such as whether to accept or reject a claim—often resulted in varying outcomes. Research into professional judgment across industries, from medicine to law, has shown similar inconsistencies. Even when presented with identical information, professionals frequently arrive at different conclusions—a phenomenon known as “noise” in decision-making.

In an industry where accuracy and fairness are paramount, this inconsistency can lead to increased indemnity costs, disputes, and inefficiencies. AI presents an opportunity to address this challenge—not by replacing professionals, but by augmenting their expertise with data-driven insights.

Beyond Large Language Models: A Smarter Approach

The rapid adoption of generative AI has demonstrated its usefulness in automating tasks like document drafting, summarisation, and legal research. However, LLMs alone are not well-suited for professional judgment in complex areas such as claims evaluation and underwriting. LLMs are probabilistic;  they generate responses based on their training data and when judgement under uncertainty is required, this can lead to suboptimal or biased outcomes unless guided to do otherwise.

Recognizing these limitations, a new approach has emerged: hybrid neuro-symbolic AI. This method combines the text-processing capabilities of LLMs with structured decision models that reflect expert reasoning. Instead of relying on AI to “make decisions” in a black-box manner, hybrid models use AI to extract relevant data from documents and evidence, while decision-making is guided by established professional criteria.

One such example is Evidential Reasoning and Belief Rule Base (BRB), a methodology developed by the University of Manchester that enables AI systems to handle complex, uncertain, and multi-faceted decision-making—much like a human expert only optimally and consistently. Unlike pure machine learning models, BRB-based systems are explainable, ensuring transparency in how decisions are reached.

A Practical Application: AI-Powered Risk Decisioning

Consider the example of Directors & Officers (D&O) insurance, where policy wording is highly complex and risk factors vary significantly. AI-driven risk identification tools can analyse policy language, endorsements, and claim files to highlight potential exposures in a matter of minutes—reducing the need for exhaustive manual review while maintaining expert oversight.

Similarly, in medical malpractice claims, AI can assist in reviewing extensive medical records and expert reports to identify key liability factors. By automating attribute extraction—such as determining whether a delayed MRI scan contributed to a misdiagnosis—AI can help claims handlers focus their attention on high-risk cases, improving efficiency without sacrificing accuracy.

Crucially, AI-powered decisioning tools must be designed with regulatory and ethical considerations in mind. This includes:

• Bias control: Ensuring AI models do not replicate or amplify discriminatory patterns.
• Transparency: Providing clear explanations for how risk assessments are made.
• Data security: Protecting sensitive personal information through anonymization techniques.

The Future of AI in Insurance

AI is not here to replace underwriters or claims professionals—but it is here to transform how they work. By augmenting human expertise with structured, AI-driven insights, insurers can improve consistency, reduce cost leakage, and accelerate decision-making.

The future of insurance decisioning will not be about human vs. machine, but about human + machine—where AI enhances, rather than replaces, professional judgment.To explore these advancements further, industry leaders will gather at the upcoming SmartRisk launch event, where Kennedys IQ will showcase how AI-powered decisioning is shaping the next era of insurance risk management. For more details click here.

About the Author: Karim serves as Chief Products Officer at Kennedys IQ, the client facing technology arm of Kennedys LLP, creating ‘baked in legal’ technology products for use by financial services clients.

With a technology background spanning 20 years, Karim has journeyed from his roots as a product manager to leading global development teams, and steering venture-funded tech enterprises. Karim is passionate about working with both academia and Professional Services to combine their expertise with advanced machine intelligence, working in the education, legal and insurance spheres to optimise decisioning, always with an eye on reliability and consistency. His career has included multiple award-winning enterprise software applications and platforms, including the coveted ‘Financial Times Innovative Lawyers’ award in 2019.