AI is everywhere in insurance now, transforming traditional processes and methodologies. If you’re in the industry, you’re no doubt already experiencing its effects in a number of ways. My own work, for example, involves using AI to streamline and automate many aspects of property insurance. We use a combination of aerial, ground-level and satellite imagery analysis for insights into the physical characteristics and risk profiles of buildings, enriched further with additional pertinent insurance data such as tax records and building permits.
Generative AI is currently the most visible aspect of AI to most people. That’s because anybody can make use of it, regardless of technical competence. You just talk to it and (if all goes well) get some astonishingly useful answers. We’ll come back later to what happens when things don’t go quite so well.
Despite the widespread applicability and visibility of generative AI across various sectors, its potential within the insurance industry remains somewhat untapped. However, with the emergence of tools facilitating the parsing, comprehension and transformation of intricate insurance documents, the landscape is shifting. With these developments, generative AI is poised to significantly influence underwriting practices, promising improvements in risk analysis, operational streamlining and overall process efficiency.
Let’s zoom out and think about the likely impacts of generative AI on underwriting in general. It’s already pretty clear that the integration of GenAI in the underwriting process has the potential to change this centuries-old process, improving risk analysis, streamlining operations and enhancing efficiency.
See also: The Promise of Continuous Underwriting
The operational perspective
Generative text solutions based on large language models are already powering new business underwriter-focused chat applications, providing instant access to comprehensive answers from underwriting manuals. These AI-driven assistants appear almost magical in their ability to interpret complex queries, synthesize relevant information and provide tailored responses.
Underwriters used to have to consult various manuals, guidelines and policy documents to do their work. But not anymore. GenAI’s automated data extraction cuts out drudge and minimizes errors. It analyzes vast amounts of data, including applicant information, risk factors and historical data, leaving humans to focus on higher-level decision-making.
For anyone brought up on manual underwriting processes, the time savings on offer are mind-boggling. But that’s a one-dimensional view. As well as saving time, GenAI helps identify potential risks or opportunities underwriters may have missed, leading to more accurate, consistent and objective underwriting decisions.
And beyond the evaluation and pricing of risk, it can potentially automate the generation of policy documents and contracts, multiplying time savings and ensuring consistency across policies.
We’re still not done: AI systems can monitor underwriting workflows, identify bottlenecks, and suggest process improvements. This continuous monitoring and optimization can lead to continuing efficiency gains and cost savings.
What are the challenges?
If that’s the case in favor, there are also challenges. Underwriting involves intricate rules, regulations and industry-specific knowledge that AI models are going to struggle to fully absorb without proper training and oversight. Concerns also remain about potential biases in the data used to train LLMs (large language models), and the potential for unfair practices if not carefully monitored.
These questions about accountability and transparency are not unique to insurance. All industries are wrestling with the legal and ethical implications of GenAI. But the highly regulated nature of our sector means that insurers must carefully navigate these challenges, without losing the power of AI to enhance operational efficiency and decision-making accuracy.
That should be possible, as AI frees people to focus on more complex and high-value tasks. However, successful implementation of AI in underwriting will require careful planning, training and change management.
See also: Underwriters' Productivity Can Double
IT considerations
If AI puts new demands on people, the same is true for tech. Integrating AI into existing underwriting systems can affect infrastructure and support requirements. AI models rely heavily on data for training and inference. While the data may be plentiful, do you have a robust data infrastructure in place and the data governance and quality processes that are also crucial to ensure the accuracy and integrity of the data feeding the AI models?
Integrating with existing underwriting systems, policy administration systems and other core insurance platforms may also require new APIs, data pipelines and middleware for smooth data exchange and communication with legacy systems.
On the security front, AI systems handle sensitive customer and underwriting data. So there is a red flag here. Robust cybersecurity measures and data privacy protocols are essential to protect against breaches or misuse of data and to ensure full compliance with industry regulations and data protection laws.
And with persisting skills shortages, insurers will probably need to invest in upskilling the IT workforce or partnering with external AI experts to ensure effective development, deployment and maintenance.
The underwriter of the future
With AI handling data processing, information retrieval and automated recommendations, underwriters can reinvent their role to focus on more nuanced decision-making, client advice and driving innovation.
Ultimately, successful adoption of AI in underwriting is going to be about a best-of-both worlds approach that combines human expertise and experience with the speed and insights of AI technologies. By getting the balance right, insurers can unlock the full potential of these tools while ensuring fairness, transparency and responsible decision-making.
