Finland is routinely named the happiest country in the world. Why? One Finnish sociology professor believes it’s “a cultural orientation that sets realistic limits to one's expectations for a good life.”
Setting realistic expectations is also the secret to making intelligent decisions around Insurance Document AI. According to a straw poll conducted during a recent webinar hosted by Roots Automation, three-quarters of underwriting teams either haven’t started or are just researching their AI journey. While this figure may look bad on the surface, it suggests these underwriting teams are ideally placed to set realistic expectations before they commit.
First, the hype. One challenge for tech implementers in every organization is that the general population has been exposed to massively inflated claims about how AI will change the world.
Publicly available general-purpose large language models (LLMs) such as ChatGPT can generate impressive bodies of text, and AI image generators can conjure amazing pictures from simple text prompts. However, there is a gulf between those examples and fulfillment of the promise that AI will make white-collar work redundant or lead to exponential gains in productivity.
“We can support decision-making already,” says John Cottongim, Roots Automation’s Co-founder & CTO. “That’s using the traditional ability of machine learning to replicate the raw language abilities of people – speeding up data analysis, for example. But that’s definitely not the same as automating decisions.”
Setting clear, achievable goals for AI investments in underwriting is about focusing on the speed of data throughput. “The things that have provided immediate value for us are where we have a lot of information and need to find out what’s important or to summarize that data,” says Jennifer Krawec, Head of Global Risk Solutions Incubation at Liberty Mutual. “There’s clear value if you have an underwriter spending five hours going through years of documents that can be boiled down to seconds or minutes.”
Krawec notes other current use cases include identifying specific information from a large body of unstructured data (such as faxes, emails with attachments etc.), quickly classifying large volumes of data, and testing documents against defined checklists, which is ideal for regulatory or contractual compliance.
“Wins” like these serve to whet insurance’s appetite for more. A 2024 survey conducted by Cap Gemini revealed that 70% of insurance executives saw underwriters as potential broker/agent relationship and product development consultants. 63% of respondents believe underwriters, freed from the burdens of data admin, can add more value to the sales process. And we may be looking at the tip of the iceberg, given the expectations for exponential growth in data available to insurance underwriters.
“Think about the mountains of data that are coming from connected assets, imaging, geospatial data and all the rest,” says Sean O’Neill, senior partner at consultancy Bain & Co.
“Much higher-frequency data than we’ve ever seen before has great predictive capabilities. You’ll see a shift over time from a bias of looking backward and historical data towards more recent and behavioral data for underwriting effectiveness.”
Again, realistic expectations are what will allow insurers to merge into this ever-expanding flow of information as AI builders and insurance businesses work through questions about bias, governance, accountability and security.
Choosing technology partners that weigh all those factors—and have proven context-sensitive templates to accelerate achievable gains for underwriting right now while learning and scaling for the future—is the surest way to anchor expectations and keep underwriters, IT/operations leaders and their CFOs happy.
To help guide your underwriting team toward success in setting – and reaching – goals for technology deployment, view “Unlocking Value with AI-Powered Underwriting.” This hour-long session, available on-demand, delivers more exclusive insights and knowledge on AI for insurance from Sean O’Neill, Jennifer Krawec and John Cottongim.
Click now to explore critical use cases and success stories from insurers using AI today to deliver measurable improvements to their underwriting performance.