What are the benefits and challenges of AI?

It’s been nine months since ChatGPT has shaken our world. I am a data scientist, applying artificial intelligence (AI) in insurance for over a decade. I have never received more AI-related questions from people around me than in recent months. They ask: What is AI exactly? How will it change our industry? How should we use AI? Will AI steal our jobs?

The interest is high, but so are confusion and fears. Those mixed sentiments prevalent in the current insurance industry are understandable. The speed of change is so fast it is hard to keep up.

Here is the good news: Over the past years, our industry has already successfully deployed many transformational tools based on AI. Visible or not, AI has already been affecting various touchpoints of our business. Those may come in other names than AI, such as machine learning, deep learning, natural language processing (NLP), large language model (LLM), generative AI and GPT (generative pre-trained transformation), but they all belong to the AI category.

My view on AI and insurance is generally optimistic: With the mature understanding and right skill sets, combined with strategic vision and ethical principles, AI will be a great catalyst, enabling 100 years’ worth of insurance business transformation in just a decade. Let me suggest how such transformation can be accomplished from consumers and industry perspectives.

AI can enrich the life insurance customer journey

Let’s start with how AI can benefit consumers’ life insurance journey. Legacy practices have long constrained the underwriting and claim process. Customers’ information, often sent via scanned or faxed documents, contains unstructured, i.e., non-organized free-form texts. AI can transform these texts into structured data with NLP, an AI technology that enables computers to interpret and manipulate human language. NLP can automatically detect key information and directly map it to the underwriting or claim process.

AI also helps assess applicants’ risk by recoupling and crossing large amounts of data from different sources, allowing insurers to conduct a systematic analysis. As a result, underwriting and claim processes become faster, more accurate, and affordable.

No need for underwriters and claim professionals to panic here – this does not mean we will not need human specialists anymore. The correct implication instead is that they can now outsource simple or repetitive tasks to AI and concentrate on more complicated cases.

AI can also play a crucial role in strengthening insurance customer relationships. For example, insurers can build LLM-powered innovative customer engagement programs that assist and enhance existing human-based interactive services such as personalized financial advice, insurance term education, claim filing advice, automatic updates, etc.

AI’s benefits and opportunities for insurers

AI can also add measurable value to insurers’ business process. For example, in actuarial assumptions, machine learning technology helps insurers accurately estimate the mortality rates of specific insured groups.

Some life insurers and reinsurers are also utilizing AI for building innovative solutions. AI-based lifestyle monitor programs, combined with biometric risk factors and mortality assumptions, offer preventive advice to optimize users’ health and mitigate future claim risks.

Other AI usage examples include generating recommendations for customized claim letters, querying contracts with third parties such as reinsurers and distributors, summarizing data for portfolio performance management, risk modeling improvement, and task automation.

Risks and challenges of using AI in insurance

As insurers, we all know that everything has risks. After all, assessing and providing solutions to mitigate risks is our core business. We also know that the risk we recognize often comes from our false perception, not based on facts. As risk experts, we need to carry our ability to distinguish true risks from false ones.

See also Be Aware of Amputation Hazards

One of the true AI risks are related to ethics and bias. Today we have very few clear regulations or processes for mitigating ethical issues in using AI, creating confusion and concerns.

The machine learning model uses applicants’ personal data and makes statistically based actuarial decisions. Some of these decisions may be perceived as discriminatory, despite being evidence-based. This argument is nothing new, as the life insurance industry has been using similar statistical models, utilizing mortality and morbidity data to categorize insureds’ risks for years.

The main risk does not come from AI itself but from the challenge of setting correct definitions and governance of bias in insurance. Fairness and transparency are core principles for our business, but an updated definition of these terms is needed to accommodate the new reality.

Another major risk is data quality. AI can wrongly learn from human mistakes, taking bad or wrong data without knowing it. Unlike humans, AI cannot exercise proper judgment when making decisions based on bad data. As today’s AI tools look so impressive, we tend to trust and delegate them too much, believing they are as good as humans or even smarter. But let’s not forget: AI is still just an algorithm that learns based on human tasks, and humans make mistakes. We must keep this in mind and set a proper standard to mitigate bad data risk.

“Technology is neither good nor bad; nor is it neutral” is the first of the six “Kranzberg’s Laws” about the role of technology in society defined by historian Melvin Kranzberg.