By now we’ve all heard the dire predictions about how artificial intelligence (AI) will eliminate all underwriting jobs in the not-so-distant future. We’ve heard how, as these tools grow more powerful, robot underwriters will be able to more efficiently and cost effectively write all lines of business. But while these new intelligent technologies are powerful, and carriers need to be examining how they can improve and transform their underwriting operations, predictions on the death of underwriting, while provocative, are tremendously premature.
Let’s take a simple example. Near my house is this amazing bakery that has been an institution in the neighborhood for generations. Its primary operation is as an industrial bakery. They bake and deliver rolls and loaves of bread to many of the supermarkets, restaurants and delis in the area. The great thing about this place is they also have a small retail operation. You can go into the bakery and pick rolls right off the baking racks. On holidays such as Thanksgiving the place gets crowded and people grab hot rolls right off the conveyor belts. When I used to go in with my dad, he always knew to get a few extra rolls since not all of them would make it home. A few years ago, one of the iconic delis in our area got in some trouble and the bakery bought it and added a second location right in the bakery. So now you can grab a hot roll right out of the oven and have a fresh sandwich made on it. Is there any wonder why this place is so loved?
Now from an underwriting perspective this is a large small business or small middle market operation, but it is also one that doesn’t fit neatly into any of our traditional models. It isn’t just a bakery, a delivery operation, a retail store, or a deli. Instead it is a combination of all of them. And I doubt there are many other bakeries across the country just like it, which makes it difficult to take a statistical sample or create a model.
AI is very powerful and is getting better all of the time, but to expect it to perform a correct one-off evaluation of a risk like this is beyond its current capabilities. Considering all the right coverages for this type of operation, creating pricing that appropriately balances those risks with the actual exposure, and negotiating coverage amounts and premiums with the agent is orders of magnitude more difficult than what AI can master today – or even in the near future.
AI, like its brethren before it of rules and analytical models, will soon be able to automatically underwrite a broader set of homogeneous risks. But some of AI’s most interesting applications are not in replacing underwriters but rather collaborating with them. Insurers are experimenting with how humans and machines can work together to supplement underwriting and yield a better and more efficient risk evaluation.
Analytics, big data, robotics and AI represent a collection of emerging intelligent technologies that can transform underwriting today in three main ways:
- Intelligent Data: The job of underwriters is to synthesize and assess data for better risk evaluation. Intelligent solutions can help underwriters by evaluating large swaths of data that they could not do themselves such as evaluating thousands of news and Twitter feeds on a given risk. These tools can also improve the quality of the underwriting data by helping to verify or qualify data elements by analyzing images and web sites.
- Intelligent Processing: Intelligent tools can also be tremendously valuable in how we process business by aiding in automation and data driven decision-making throughout the underwriting process. Cognitive robotics can be used today to evaluate and route servicing requests saving the underwriter time. Win-probability and other analytics can help the underwriters profitably prioritize their work.
- Intelligent Enablement: The most important and interesting advancements in these intelligent tools are in helping the underwriters to evaluate the risks. This can include evaluating large swaths of data and identifying exposures underwriters need to consider, or creating comparative peer groups for the underwriter to consider in evaluating the risk. Other solutions pull together the companies best industry knowledge to help the underwriter evaluate the risk.
So rather than focus on the death of underwriters, it is time to really examine and explore how humans and machines can work together to create more consistent, more efficient, and higher quality underwriting decisions. Done correctly the results can be as satisfying as a great sandwich on freshly baked bread.