The Ethics of AI in Commercial Insurance: How to Approach this Revolutionary Technology Responsibly

From confronting biases inherent in data to ensuring transparency and regulatory compliance, stakeholders must proactively address these challenges to foster responsible AI development and deployment within the commercial insurance sector.

(Image credit: Dylan Gillis/Unsplash.)

In the landscape of the insurance industry, the integration of artificial intelligence (AI) and machine learning technologies has become instrumental in enhancing operational efficiency and decision-making processes, from underwriting to claims and anywhere in between. According to a recent Conning survey, 77 percent of C-Suite insurance decision makers noted they are adopting the technology to some degree. However, the profound impact of AI adoption is accompanied by a host of ethical considerations that necessitate thorough examination and strategic navigation. From confronting biases inherent in data to ensuring transparency and regulatory compliance, stakeholders must proactively address these challenges to foster responsible AI development and deployment within the commercial insurance sector.

Challenges of AI adoption

At the forefront of ethical concerns surrounding AI in commercial insurance lies the issue of bias and discrimination. The reliance on AI models trained on historical data introduces the risk of perpetuating and amplifying existing biases, thereby undermining fairness and equity in insurance underwriting processes.

Imagine a small business—let’s call it “Cafe Cozy”— seeking insurance to cover potential damages or losses. The owner applies for a policy through an insurance company that solely uses an AI system for underwriting. This system evaluates applications based on historical data, including business type, location, revenue, and past claims.

Now, let’s say the AI system has been naively trained on data showing that businesses in the hospitality sector, like cafes and restaurants, have a higher frequency of claims related to water damage (due to plumbing issues, for example) or customer injuries (like slips and falls). If the system has not been adequately adjusted to account for mitigating factors—such as Cafe Cozy’s investment in top-notch plumbing and a well-maintained, safe premises—it might still categorize Cafe Cozy as high-risk based solely on its industry and past data trends.

Consequently, Cafe Cozy could face higher insurance premiums or more restrictive coverage, despite the owner’s proactive measures to reduce risks. This scenario illustrates a form of industry bias, where the AI’s decision-making process could unfairly penalize certain types of businesses based on broad, industry-wide trends, without considering individual business practices that mitigate those risks.

Furthermore, if Cafe Cozy is located in an area historically associated with higher claims—perhaps due to environmental factors or higher crime rates—the AI might apply additional premiums based on geographical bias, compounding the issue.

While AI can enhance efficiency and objectivity in evaluating risks, it’s crucial to ensure these systems are sophisticated enough to recognize and weigh the specific actions businesses take to manage risks. Moreover, the opacity of complex AI algorithms exacerbates this challenge by hindering the identification and mitigation of bias, necessitating robust measures to promote transparency and accountability throughout the AI lifecycle. GPT-3 uses more than 150 billion parameters and although not publicly disclosed, it is estimated that GPT-4 surpasses one trillion. Modern AI complexity is unimaginable.

While bias remains a primary focus, the ethical landscape of AI in commercial insurance extends beyond this singular concern. Transparency emerges as a critical consideration, particularly concerning the increasingly complex nature of AI models. The opacity of these algorithms poses challenges for regulatory oversight and accountability, raising questions regarding data security, job displacement, and the responsible use of AI-generated imagery. Furthermore, the intentional introduction of bias by developers underscores the imperative for comprehensive ethical frameworks to govern AI development and usage effectively, as the recent events surrounding Google’s Gemini image generator have shown us.

Regulators face significant challenges in addressing the ethical implications of AI in insurance, particularly concerning transparency, data security, and job displacement. Recent initiatives, such as the New York state circular letter release in January, provide essential guidance for insurers navigating the ethical dimensions of AI adoption. However, the practical implementation of regulatory directives remains the real challenge, requiring the development of effective enforcement mechanisms and metrics for measuring compliance to ensure alignment with ethical standards while fostering innovation and competitiveness.

How insurers can get and stay ahead

A robust code of conduct tailored to the insurance industry plays a pivotal role in mitigating bias and promoting ethical AI practices. Such a code should encompass guidelines for both model creation and usage, addressing concerns related to data confidentiality, fairness, and transparency. By fostering collaboration between industry stakeholders, including insurers, regulators, and AI developers, a comprehensive code of conduct can facilitate the responsible development and deployment of AI technologies, safeguarding against unintentional biases and ethical lapses.

Central to the ethical adoption of AI in insurance is the concept of human oversight, which serves as a critical safeguard against algorithmic biases and errors. Adopting a “human-in-the-loop” approach could ensure, for now, that AI operates under human supervision and direction, enabling insurers to leverage the transformative potential of AI while upholding ethical standards and regulatory compliance. By empowering human experts to provide guidance and oversight, insurers can navigate the complex ethical terrain of AI adoption with confidence and integrity.

However, crafting and enforcing AI regulations present formidable challenges, necessitating collaborative efforts to establish clear guidelines and robust enforcement mechanisms. The dynamic nature of AI technologies, coupled with evolving regulatory landscapes, requires ongoing dialogue and adaptation to address emerging ethical concerns effectively. Furthermore, the use of AI to check AI poses unique challenges, requiring careful navigation of technical standards and client relationships to ensure transparency, accountability, and trust. I work with AI for a living and struggle to stay up to date on the latest developments, imagine regulators and other people whose job is not directly related to AI.

For any and every insurer, as they may consider leveraging AI partners, it is mission critical to evaluate SaaS providers to determine if they have a proven, well established track record of ethical AI practices. In order to stay ahead of ever evolving global regulations, a provider must dedicate ongoing resources to maintaining a healthy evaluation process of how to not only best use AI, but ultimately ensure that this is done so from a responsible lens.

As the insurance industry continues to embrace AI technologies to drive innovation and efficiency, ethical considerations remain paramount. By addressing concerns related to bias, transparency, and regulatory compliance, stakeholders can navigate the ethical waters of AI adoption while fostering trust and accountability within the industry. Through collaborative efforts, robust ethical frameworks, and human-centered oversight, insurers can harness the transformative power of AI to shape a more equitable, resilient, and ethically responsible future for the commercial insurance sector.

The Future of Customer Service is in the Hands of the Bionic Workforce


Leandro DalleMule //

Leandro DalleMule has nearly 30 years of experience helping several businesses in retail, energy, manufacturing, telecom, and especially financial services to achieve sustained increased profitability through data, AI, and analytics. DalleMule is the Global Head of Insurance and General Manager for Planck, which bills itself as the most sophisticated Generative Artificial Intelligence platform for commercial insurers, working with clients in the U.S., Australia, Germany and Japan, with plans to further expand. Previously, he spent five and half years as AIG’s Chief Data Officer, building and leading the Data Management function across the company. While at AIG, he developed and executed several data-related initiatives including governance, standardization, quality measurement & improvement, creation of a single-source of truth, architecture, actuarial, BI and analytics delivery. DalleMule holds a B.Sc. in mechanical engineering from University of Sao Paulo, Brazil, a magna cum laude MBA from the Kellogg School of Management at Northwestern University and a graduate certificate in applied mathematics from Columbia University.

Leave a Comment