AI for insurance: automating bottlenecks

We have self-driving cars, spacecraft on Mars, and instant access to global knowledge. Yet, the insurance industry is still playing catch up.

While other industries are making the most of what AI can offer, so many underwriters are still spending hours and hours on tedious tasks every day. From navigating lengthy documents and extracting information, to data entry into multiple systems. 

Luckily, this doesn’t need to be the future.

In this blog, I’ll discuss how line slip extraction – a crucial process in underwriting agreements involving multiple insurers – can be improved with AI. I’ll explore how large language models (LLMs) are stepping in to speed up the process and what this means for insurers.

The manual burden of line slip extraction

Line slips are semi-structured documents that lay out the terms and conditions of a risk. They are effectively designed to act as a standardised contract between brokers, policyholders and insurers. 

Although they’re intended to streamline insurance placement, there’s still a fair amount of manual heavy-lifting involved.

Once a risk has been placed, the insurer has to input all the details into their policy administration system. This system monitors the company’s risk exposure, so everything that’s fed into it needs to be 100% accurate and complete – demanding time and precision. 

Now, imagine a typical FTSE insurance company selling tens of thousands of policies in the London market every year, with each line slip running over 40 pages. Manually processing all that data isn’t just inefficient – it’s a major bottleneck to scaling the business effectively.

Traditional approaches to break bottlenecks

Previously, the Lloyds Market Association has attempted to automate this with their Structured Data Capture (SDC) tool. However, insurers often find the tool isn’t accurate enough to rely on. And with processing times up to 4 hours, it can be quicker to handle line slips manually.  

On top of that, using the tool comes with a hefty variable cost of approximately £20 per slip.

The good news is that recent AI advancements, particularly LLMs like those used in ChatGPT, have opened doors to new solutions. 

How AI transformed a FTSE 100 insurer’s processes

A FTSE 100 insurer recognised the potential of AI to help fast-track core processes. They knew it could free their team to focus on more high-impact work.

To help them get started, we identified where they could embed AI effectively. But the real game-changer was our custom-built AI tool, designed to automatically extract data from line slips.

All the underwriters need to do is send line slips to a dedicated inbox. Within around 2 minutes they receive a dedicated email report containing all the key fields extracted in a structured format. Along with a reference to where the information came from.  

This allows for “human-in-the-loop” AI safety checks to take place before the information is entered into policy admin systems.  

This process can all be done with an overall accuracy of around 93%. Response times are within 2 minutes. And costs are roughly $0.25 per line slip. 

In the future we’ll develop the functionality to automatically populate the policy admin system in cases where confidence in the extraction accuracy is high. This has reduced the time it takes for a line slip to be processed from ~25 minutes to just ~5 minutes.  

When you consider that around 55,000 line slips are processed per year by the insurer, the operational time savings become really significant. And crucially it eliminates the bottleneck to scaling and writing more business.

What does this mean for other insurers?

Insurers can start implementing AI to enrich day-to-day operations that often take up a huge amount of time. 

When implemented the right way, LLMs allow insurers to scale operations without a proportional increase in manual effort, driving down costs while maintaining speed and efficiency.

In short, LLMs don’t just have the potential to speed up line slip extraction, they can supercharge expert decision-making and automate manual processes. Ultimately the results could be better risk management and improved operating ratios.

You can find out more about our applied AI services for insurance here. Alternatively, get in touch with a member of our team.