Productivity isn’t just a corporate buzzword. Rather, it’s an increasing priority for insurance firms of all sizes.
Post-pandemic inflation has increased the cost of claims (even up to 10.5% in December 2022). With rising interest, insurance companies might explore different avenues to grow in a progressively challenging landscape. Enter generative AI: the much-hyped productivity tool.
When carefully implemented, we can trust generative AI to complete parts of the underwriting process. These parts can include:
AI’s speed and accuracy make it an attractive option for CFOs looking to send their productivity figures soaring. Let’s break it down.
The role of AI is to automate aspects of underwriting that would slow down experienced underwriters. Underwriters, ideally, should be spending their time on high-value tasks that AI cannot currently automate. Such tasks might include:
Natural Language Processing (NLP) is what really makes AI an adequate substitute for an underwriter’s administrative expertise. NLP refers to the ability of machines to comprehend natural language as spoken and written rather than just the information’s typography. Finely tuned NLP algorithms underpin many AI-powered tools that support productive underwriting. Let’s walk through a few examples of these tools:
In a fascinating article by Insurance News Net, Dale Gonzalez, chief product officer at Axio, notes the inadequacy of insurtech in 2024: ‘The connection between the client, the broker and the insurer is still very manual and ad hoc with lots and lots of rekeying of information’. We've written extensively about how the rekeying of information - also known as manual data extraction - can slow down teams dealing with large volumes of financial data.
If implemented successfully, data extraction tools will reduce (but won’t likely eliminate) the manual touchpoints associated with moving data through the underwriting process. Data extraction tools may occasionally flag errors (even if it’s only 1 in 100,000 data points). Then, you’ll need to review those errors for accuracy.
Certainly, the initial stage of extracting data from underwriting documents (e.g. policy applications, claims forms, bank statements, medical records, vehicle data, public records and so on) does not need to be completed manually. NLP’s capabilities mean it can parse uploaded documents and extract the required data, like a (human) analyst. Unlike an analyst, however, AI-powered tools can extract data in seconds.
Of course, the extraction speed doesn’t matter if the data quality is impaired. Maintaining the fragile balance between data accuracy and speed historically flummoxed AI developers. However, the emergence of technologies like Large Language Models (LLMs) demonstrated that speed and accuracy could form a formidable and productive pairing.
Accuracy is achieved when AI algorithms can verify internal consistency, i.e. all the data in a document makes sense. If a document isn’t internally consistent, then, generally, a data point has been incorrectly extracted. As a result, either the AI algorithms can correct the error or flag it for an underwriter to review.
Data validation is necessary when it comes to decision-making. For example, if you use the extracted data to assess risk and determine premiums, a robust data validation tool will ensure that the underwriting process is accurate, fair and compliant.
AI excels at identifying anomalies. By feeding data into AI, it can flag potential risks based on outliers and use it to recommend a decision.
However, high-quality AI for underwriters revolves around transparency. It should make accurate judgements and can also digestibly present how it arrived at that decision. Sting Fan in The Actuary suggests that compromised transparency and governance could damage an insurer’s prospects as ‘risk management failure is highly likely, and this could lead to flawed outcomes and undermine consumer trust’.
Personalised insurance offers an interesting use case for insurers looking to branch out from automating existing processes. Fintech Global notes a growing trend of consumers looking for personalised insurance offers. In other words, consumers want tailored offers deposited straight into their inbox. Instead of generic mass mailings, they want their data used meaningfully. They want to receive personalised offers tailored to their individual needs rather than being included in broader demographic groups, even if those groups are very specific.
Creating personalised insurance offers might seem like a pipe dream for companies relying on manual data extraction. But for firms using robust data automation, creating personalised insurance offers a collaborative and revenue-boosting use case for underwriting and marketing departments.
If looking for customised underwriting functions, it could be unrealistic to expect AI to deliver a completely accurate experience initially. With training, however, a custom AI project can quickly yield 99-100% accurate data. If you’re looking to process data from various underwriting documents, you may need to submit samples for training.
Setting up a feedback loop might seem like a lot of work, but you can work out the details with an insurtech vendor. For example, we generally keep training time under 48 hours.
According to insurtech firm Hyperexponential, over two-thirds (69%) of underwriters are concerned that AI will replace them by 2029. While reading this article, it may seem like AI is supplanting key functions of the underwriter. In actuality, AI can nudge underwriters into a more strategic role, where they can engage with data rather than just moving it around their internal systems. Underwriters, after all, are highly skilled, usually possessing (potentially underutilised) finance degrees.
Employers must introduce AI carefully, with set functions and plenty of opportunities to test it through the acquisition process. Winning over stakeholders is an art and not one that AI could replace.
The insurance industry is facing pressure from all sides. Staff shortages, inflation and greater competition mean insurers are seeking productivity hacks. AI-inspired innovation may be the answer.
Rather than rescinding their market share, companies are beginning to explore how they can implement generative AI for productivity gains. Rather than replacing underwriters outright, AI-powered insurtech cuts time spent on back-office processes (like data extraction) and offers new opportunities (via personalised offers).
While you can use an AI tool successfully in the long term, the transition to using one may require initial training – plus a focus on its benefits when introducing the technology. It’s also important to choose the right tool for your requirements.
A bit about Evolution AI: We automate data extraction.
More specifically, our solutions are designed to generate productivity by decimating the back-office functions that burden growing insurers. We’ve turned custom projects into major success stories for high-profile clients across financial services.
Questions? Comments? Contact us at hello@evolution.ai or book a demo with our financial data team.
