This anonymous case study is part of an ongoing series that gives readers a behind-the-scenes look at some of the most challenging and impactful business problems in the insurance industry solved by SortSpoke. We keep the clients and their companies anonymous so they can share full details.
The quality of your decision-making is only as good as your data. You could have a fully staffed team of the most skilled underwriters in the industry, but if you cannot easily get to the most important information quickly, you cannot determine which submissions to quote on first.
Reliably getting the data you need into your underwriting systems is often challenging and time-consuming. And it is made harder by the variety of documents your team receives from brokers that they have to process manually today.
What is unstructured data?
Unstructured data is any digital information lacking a formal structure—meaning it is not easily enterable in a database or spreadsheet. In insurance underwriting, that frequently means text in documents (like PDF files), emails, loss runs, statements of value, financial statements, or other scanned paper documents, including handwritten ones.
Historically, data in all of these formats had to be processed manually by human staffers. That is very time-consuming and inefficient. And transcribing is very prone to human error. Those factors can seriously affect the quality of underwriting decisions and, therefore, corporate profit margins and strategy.
One major Canadian P&C insurance company wanted to improve its inefficient broker submission triage process. They were writing enough policies to stay profitable, but it took an exceptionally long time to find those profitable accounts using their existing manual processes and error-prone automated tools. This case study explores how they streamlined their intake and underwriting processes with SortSpoke.
Entrenched manual processes and staff wary of digitization
This Canadian insurance company was already successful. They had thousands of employees at offices nationwide, serving almost two and a half million customers. But one of the national P&C managers at this company could see just how outdated and inefficient many of their processes actually were—especially during submission intake and the handoff to underwriters.
They have a “funnel” process to move newly submitted applications from an initial triage and validation to their underwriters. “It takes ages,” the P&C manager said. “We have a team that just does application triage. This team’s entire workload is just entering information from applications and doing some validation checks. We have so many submissions that if we put that all on underwriters, they would spend half of their work hours just doing that data entry. So it's a huge pain point for us.”
The triage team manually entered unstructured data from submitted paperwork so their insurance software and underwriting staff could properly evaluate it. “Anything we can do to make that process more efficient is a priority.” She went on to explain that some of their business was highly time-sensitive. Some of their customers needed underwriting decisions and quotes turned around quickly for their own timely decision-making.
“We deal with tasks that could turn into between two thousand and one million dollars in premiums. There are different senioritis and different types of underwriters that we want to work on those various tasks. We often can't win that business if we can't quickly get it to an underwriter.”
Metrics revealed the scope of their manual processes’ inefficiencies
This insurance company had a data science team. As the P&C manager explained, that team created an in-house industry code predictor application to save their triage and underwriters time on industry searches. “We have our own industry codes that are specific to our company, and there's about 2,000 of them. They’re very technical. And we have an accuracy rate at triage of about 30 percent.”
Their software was not cutting it. But, importantly, the P&C manager recognized it was not just about the software. It was about how software tools supported workers who were critical to their workflow but might not yet be the most knowledgeable. “It tends to be the least experienced people in insurance that work in this triage role. And so you'll get tasks routed to the wrong areas. So, getting that right consistently and immediately will be a big difference. That will improve everything downstream.”
Making a business case based not just on automation but finding value faster
“We're using a lot of predictive models to prioritize submissions and enhance the quality of the information we're getting,” the P&C manager said. She had gathered a group of internal stakeholders, and they were trying to decide what business case for automation they needed to bring to their leadership. “A lot of our challenges lie in the need to enhance the information that we're getting, get access to information that we don't have, and analyze all of it better.”
After consulting with those stakeholders and her data science team, she brought her proposal to leadership. “I made the case for generating new efficiencies and an uptick in new business. With more efficient document processing, we’d be able to get to our profitable accounts a lot faster.” She received the thumbs up on her proposal and got down to work researching options on the market.
Get our free self-assessment tool to see if your company could benefit from AI-powered document processing.
Selecting SortSpoke for their automated document processing
The P&C manager and her evaluation team scoured the market for different AI-powered document processing tools. Many developers made grand promises for what their software could learn. “SortSpoke was one of the only vendors that were actually able to do what they said they would do. Other vendors just weren't capable of proving their claims.”
SortSpoke, on the other hand, could handle all of their unstructured documents just like it claimed it could. And it could do that without forcing this major insurance company to rebuild its entire IT stack from the ground up to accommodate it. “Integration is really big for me,” the P&C manager said. “SortSpoke integrated right into our existing systems.” After a short deliberation, her team decided to proceed with SortSpoke.
Working collaboratively on deployment
One thing SortSpoke engineers made clear to her was that to get the automated intake right, there needed to be some upfront configuration and training of the software. But having seen what SortSpoke was capable of in testing, that was not a concern for her. “What we like about it is that it has flexibility. It isn’t a problem for us that it puts training in the users' hands because they're the ones with the expertise!”
The AI training approach SortSpoke uses only takes a few weeks to get up and running. And you don't need to train a technologist to make the machine learning work. All you have to do is give your submission triage staff or underwriters an easy-to-use tool that learns from them directly.
“We talked to other vendors who said they have developers who could do the training on our submissions. But really, we’d still need to pair them with our users to teach the developer how to identify the pieces of information. So we would end up paying for it twice.”
Managing user perceptions of AI
While the P&C manager and her team were happy with SortSpoke right away, some of her triage and underwriting team users were initially less enthusiastic. She soon realized they were wary of AI tools and what they might mean for their job security. “Some people are more or less familiar with tech, and so they hear some of those key buzzwords of AI, but don't really understand what they mean, what it can do. People saw it and said ‘Oh, you're automating my job away.’”
The solution was just sitting down and showing those users what the software could do. It did not take long for the deployment team to show the wider workforce just how effective SortSpoke was at improving their workflows.
It was not replacing anyone. It simply eliminated large volumes of tedious data entry so employees could get to decision-making faster. “People always think they want 100 percent automation where nobody has to look at anything. But what they really want is automation that does more of the grunt work, so all you have to look at is what’s valuable.”
A wide range of benefits and positive user feedback
The P&C manager was pleased to get positive feedback from team leads right away. “Feedback has been pretty positive,” she said. “The system is intuitive, so there was limited training. People are really excited about the process and what they can do with SortSpoke. We’re seeing good adoption.”
Speed of submission intake increased 5 fold for their triage team. And document review times dropped by nearly 90 percent for underwriters.
There were some unexpected benefits that she had not foreseen, too. The two most important are better data on submissions they declined to write—something they rarely had before—as well as offering better career development for their triage staff using AI software.
New data on declined submissions
“Also, I can get better answers now when I have a question about declined submissions,” the P&C manager said. “The questions we usually get are ‘What is it that we're declining? Why are we declining it?’ But when we processed submissions manually, the team just didn't put declines into our system. Somebody just read an application, and they were like, ‘No, no, we don't write that.’ So then we never had good data to back up decline decisions. Now having access to that information, being able to answer those questions is going to make a lot of people happy.”
Career development for triage staff
“We're having our triage team use the system,” the P&C manager said. These are typically the least experienced professionals in her organization, ones whose skill sets and career prospects she wants to develop. By giving them access to SortSpoke they get to develop their competency with the newest generation of insurance technology. “They’re helping to train machine learning models. It's going to look really nice on their resumes to say they’re doing more than just data entry.”
Feature requests & “lightbulb moments”
The P&C manager for this insurance company was also particularly happy with how easy it was to get new features added to their SortSpoke deployment. “Honestly, you guys are generally really nice to work with,” she said. “We can request additional features and it gets on the roadmap. You guys are always faster than we are. You guys are really easy to work with.”
She foresaw a long lifecycle for their SortSpoke deployment. “I can see how we can build on what we have and add additional pieces of information that we're looking to extract. You start with a little bit, and then eventually, you get to literally every single field. That will be awesome for us as we continue to develop this product and see what else we can do. We'll start seeing people have these light bulb moments where they're like, ‘Oh. We get this. This can do so much more.’”
See how SortSpoke can improve your underwriting process
Book a risk-free, no-obligation discovery call with our team.
