With the advancement of artificial intelligence (AI), insurers can now use more sophisticated algorithms to analyse vast amounts of data when making underwriting decisions. AI can allow insurers to identify patterns and trends that might not be apparent to human underwriters. Ultimately, AI has the potential to make underwriting faster, more efficient, and more accurate.
However, using AI in underwriting is not simple and opens insurers up to risks that need proper attention and careful management. The amount of work required to get an AI system robust enough to be trusted with underwriting decisions is immense. It brings a cost-benefit analysis that is individual to an insurer. Using AI in underwriting could have severe and costly consequences without proper governance.
The question is, ‘Is using artificial intelligence (AI) in your underwriting model a good idea?’ The answer is complex and depends on several factors. This blog will discuss the potential benefits of incorporating AI into your underwriting process. We'll also discuss the risks of using artificial intelligence in underwriting models and why human input is even more critical when using AI.
Benefits of AI in underwriting
Speed
AI can analyse vast amounts of data in seconds, meaning decisions could be made just as quickly. This is particularly important when speed is crucial to securing and retaining customers.
Accuracy
Traditional underwriting is prone to human error and variable decisions, potentially leading to incorrect conclusions. With AI, underwriting decisions are based on data and algorithms, significantly reducing the risk of error provided the data is accurate and without bias (see risks)
Fraud detection and prevention
Insurance fraud is a significant problem in the industry, and it costs companies billions of dollars annually. AI can help to identify patterns and anomalies that may indicate fraud. This can help insurance companies to take action to prevent fraudulent claims from being paid out.
Improve customer experience
By analysing customer data, AI can identify patterns and trends that can be used to create tailored products and services.
AI can also allow insurers to process large volumes of data in seconds, streamlining the application process and providing some customers instant coverage.
Risks of AI in underwriting
Incomplete information
AI performs based on a specified framework and can only gather information it has access to and has been told to gather. This can mean that AI cannot capture all information about a person and their specific risk profile.
Unreliable results
AI algorithms rely on large volumes of data to identify patterns and make accurate predictions. AI will not perform as expected if the data set is too small or learning does not occur over a long enough period to allow enough variations. Likewise, the AI algorithm may produce incorrect results if the data is inaccurate.
High requirement for human feedback
One of the main feedback loops that enable AI to ‘learn’ is the human feedback of “yes, that was a good decision” or “no, that was a bad decision”. This ongoing ‘learning process for AI models requires labour-intensive human input. If sufficient oversight is not committed, AI can be inaccurate and produce unreliable results.
Potential discriminatory outcomes
In society, there is an expectation of fairness and equality, with strict governance around discrimination. This is a real problem in insurance, as, by nature, underwriting is discriminatory. For example, if you are seeking life insurance and are in your 60s, your ‘risk’ is much higher than someone in their 20s, and therefore, your premiums will be significantly higher – if you can get a policy at all. Generally, excluding or discriminating against an individual based on age would be illegal and unethical.
AI can potentially remove bias, as decisions are always a calculation, but what if AI is exposed to bad data sets? Or makes a risk assessment based on trends within a group of individuals and attributes it to the wrong demographic? Or attributes a risk to the incorrect attribute?
For example, what if there is a trend amongst a group of females in their 20s that they lose an arm, and AI attributes this to their age and gender, but it is related to them being part of a group of people who like to feed Great Whites in shark alley?
The issue of transparency with AI in underwriting
Historically, there has been relative transparency between risks and underwriting in insurance. For example, if you have a family history of heart disease, your life insurance premiums will be higher. AI can analyse large amounts of data to identify patterns and trends. With massive data sets, the correlation between risk and underwriting becomes less clear. There is also the consideration of intellectual property as methods of teaching AI to analyse may be commercially sensitive.
The question of accountability with AI
What if an AI system makes a mistake? Who is responsible?
Is it the developer who created the system, the company that deployed it, or the AI system itself? This can be a challenging question to answer and is an area that is currently untested. This means using AI carries potential legal risks that must be carefully considered.
If AI systems are given biased data, they may make bad decisions, resulting in legal action against insurers. Companies must take a proactive stance to govern their AI to eliminate bias. They should also prepare to defend their decisions regarding data selection, quality, and auditing procedures that ensure bias is not present in machine-driven decisions.
Another legal risk is the potential for errors in AI systems. If an AI system makes an error in assessing risk or pricing policies, it could lead to legal action from customers who have suffered harm. If an AI system decides to deny a claim that a human would have approved, it could result in a negative response against the insurer.
The real benefits of using AI in underwriting lie a few years out
While Artificial Intelligence represents a promising new tool for underwriting, it is essential to recognise that it remains a tool and is only as good as those who create it and use it. Significant benefits of AI implementation in underwriting are expected in the next few years as insurers obtain more data and refine machine learning models. It is crucial to approach the integration of machine learning into underwriting models with thoughtful preparation and planning. However, once the necessary investments have been made, underwriting teams have ample opportunities to leverage the technology.
The future of AI in underwriting
AI is transforming the future of underwriting and will likely become more sophisticated and capable of handling complex tasks. As technology advances, we can expect to see even more innovative applications of AI in underwriting, and the future is promising.
While AI has the potential to be a game-changer for the insurance industry, it's essential to keep in mind that technology alone is not enough. Human judgment is still necessary because there are no perfect solutions or answers to risk assessment. Underwriters need access to data that provides insight into trends and patterns within an organisation - information they can use alongside their expertise in deciding which risks are worth insuring and which aren't worth taking.
The power of the AMS rules-based automated underwriting engine
Whether to use AI in underwriting is an open-ended question; we do not have the answer. However, as technology continues to grow, so does AMS Insurance.
We have been leading the way as a software solutions and services company with a rule-based automated underwriting engine in our product, enabling our customers to underwrite life insurance sales without any human intervention. Our clients configure their underwriting engine to make decisions based on customer answers and decision trees, automating the underwriting process. This delivers consistency of underwriting decisions and transparent traceability of the decisions made.
Based on a set of rules configured by the insurer, insurance can be provided, or if the risk is perceived as high, the engine may refer the application to an underwriter. If the application is referred, the Q&As and perceived risks can be viewed by the underwriter, which will support their processing and efficient decision-making.
Our underwriting engine is a game-changer for insurance companies who need an advanced tool designed to streamline the insurance application process, reduce turnaround time, and improve customer satisfaction without the risks of using AI in underwriting.
Because of the developmental structure, our software is predictable and controllable and offers transparency.
Partner with AMS Insurance
AMS is trusted to manage over 50% of New Zealand’s life insurance policies for some of the industry's biggest (and smallest) names.
We cover everything from the initial policy proposals through all the multiple touchpoints for the life of a policy to claims management. We cover all the latest risk-based modular products in tandem with legacy product offerings. Ours is an end-to-end solution on one platform, delivered as SaaS, that integrates into your existing systems.
AMS Insurance Management has a comprehensive range of services, including running our customers' entire policy management technology stack. We tailor our approach to what you need today while planning with you for tomorrow.
Contact us to see how we can enable you to streamline and automate your underwriting processes with confidence.