Programming a computer for evaluating Insurance risk:
Is it a good idea?
As the “Insurtech revolution” advances in the insurance industry, executives and industry professionals are starting to understand and embrace that there are many advantages this technology can bring to the insurance industry.
Outsiders from the industry are visualizing and thinking of new ways to disrupt the current infrastructure and system to improve the speed of transaction and lower the costs associated with the system. See conferences like Insurtech Connect, which is growing both in attendance and recognition by traditional insurance companies.
Similar to how the first computer was able to play against and beat various chess masters – the first computer will be used to advance the insurance industry by evaluating risk and pricing it autonomously.
Five potential advantages
- Computers are fast at making calculations
- Computers won’t make errors unless the errors are made by humans entering the codes:programming
- Computers won’t get lazy and fail to fully analyze a risk
- Emotions or other personal biases will be removed
- Overconfidence in assumptions will be removed
The beauty of adding computers to automate risk evaluation and/or insurance pricing is the power and scalability of the technology. This will help lower expenses and hopefully pass savings on to consumers (as costs are eliminated from the system). The current system has too many expenses per insurance dollar (30-40 cents depending on product line) going to the evaluation, processing, etc. of insurance product.
This type of technology should definitely be considered for easy to evaluate, easily modeled and/or low volatility type risks. These types of low volatility risk will be easier for oversight over the algorithm. In the event there is an error or code/algorithm mistake – the low vol. business will allow humans to understand the exposure so a company is not over leveraged or over exposed.
In areas where risk is less easily modeled, high volatility, or requires critical thinking it may be more advantageous to have a human. Over reliance on a computer program may result in failure to recognize an issue or failure in the code. Then it may be too late for management, regulators, analysts and/or investors.
Long tail vs. short tail products
Shorter tailed product lines seem well fit for a computer program or algorithm. At the end of 12 or 18 months an insurer will know win or loss, and decide if the algorithm needs to be revised.
Longer tail products may be less suited for programs making the entire decision – insurers and reinsurers could put on considerable risk positions without the ability to de-risk or recognize the profitability of the position. The evaluation period is also much longer to determine Win or Loss.
This is why human analysis will still be necessary.
Advantages of human analysis
- Flexibility of human mind, the ability to shift gears to solve a problem rather than follow code
- Humans have imagination and creativity
- The ability to reason
- The ability to learn
Insurers who embrace this new technology and implement an algorithmic approach to underwriting and risk will have to think about a few post implementation, such as:
- How will you know if the computer is making the most advanced strategic move or has a glitch? If a human reviewing the computer program can’t understand the decision, you may have scenarios like the flash crashes of the stock market or an overweight risk portfolio as the computer program had a glitch.
- Was it the computer coming up with a new advance output that although seems weird is the best fit for a portfolio based upon the algorithm or is their a bug in the formula? Who will be smart enough to question it? Correct it?
- How frequently will this program have to checked? All hours of the day?
- When will you find out- will it be too late?
- Will you be able to de-risk your portfolio with reinsurance or retro reinsurance easier? Will the computer on the other side not accept certain aspects of the portfolio?
We have to view technology as what it’s always been- advancing tools for the betterment of human life. It should not be idolized or feared but embraced.
Insurance professionals and executives should prepare and align themselves for this “Insurtech Revolution” – all levels of organizations – junior level and senior executives should learn how to evaluate data, algorithms, and machine learning.
- Machine learning algorithm vs Actuarial science
- What insurance industry can learn from baseball?
- Will Reinsurers be a thing of the past?
- Language of risk
- Peer to Peer Insurance