Author: Daren Rudd, Head of Business, Technology and Innovation Consulting, Insurance CGI UK
There is a definite buzz in the insurance industry right now aroundthe opportunities Artificial Intelligence (AI)represents for improving efficiencyand services. Almost every client conversation I havehas AI somewhere in there, especially when discussing IT or business strategy or the direction of the market.
Much of this interesthas been spurred by the recent stellar rise of GenerativeAI (GenAI) and Large Language Models (LLMs). ChatGPT and Microsoft Copilotare some of the better-known examples of this newer AI. This initial excitement stems from the almost human-like understanding these tools appear to demonstrate when used with unstructured information.
With tools like ChatGPT the power to understand language and meaning mean they are well suited to address some of the sticky challenges the insurance market has struggled with for years. In particular, they can help ease the load of the heavy manual processing needed to handle unstructured documents and data.
It is easy to equate this capability with the chance to reduce the drudgery of day-to-day administrative tasks for staff and customers. However,it’s worth bearing in mind that despite the focuson GenAI, this is only one of several different flavours of capabilities captured under the AI umbrella term.
Many of these have been around in one form or another foryears. At CGI we have been using AI and machine learning (another capability under the AI umbrella) for over 20 years, solving a variety of problems as diverse as understanding the health of sea grass fields under the ocean, helping the European Space Agency guard satellites from space junk, or building wildfire predictive models to protect property.
Having worked with these technologies for so long now, we find that the best way to understand the opportunity is to start first with the business problem. With that well definedit becomes simpler toidentify the right technologies to solve that challenge, and the potential return on investment.
But difficulties often arise when we start with the technology first.
Cutting through the noise
When speaking to underwriters in the speciality and commercial markets one of the major issues is the sheer volume of email submissions (enquiries) they receive. The challenge is how to find the business they reallywant to write hidden in the (hay)stack of emails and documents sitting in their inbox. One lead underwriter told me that her team had to leave 90% of emails unread. And she wasn’t sure that the 10% they did quote was the best business! There aren’t many industries which have to ignore 90% of business opportunities.
To help solve this, as part of our Underwriting Workbench we have used AI to review incoming emails and documents to help guide underwriters on what to focus on first. The underwriter remains in control, managing the data the AI is looking for, as well as deciding what to work on. The system also learns as the underwriter works and responds, to better help identify the right type of business.
Finding needles in haystacks
Another challenged faced by underwriters and claims staff is the effort required to find the right information hidden within complex and unstructured documents when assessing a risk or claim.
Insurance has long suffered from a heavy reliance on multiple documents in all sorts of formats and layouts. Even something as simple as an inception date can have many different formats and titles. Old school OCR (Optical Character Recognition) would get confused by that variation, needing much more consistency on where to find the expected information – even skipping to the next page would cause a problem. We have been working with newer AI based data extraction technology for several years now across a variety of document types. The introduction of the new Large Language Models (LLM)has meant a major improvement in the ability to return insights from messy data stores.
Do not press 1 for customer service
We have all experienced that sinking feeling when calling our insurer only to be offered a lengthy trail of options to navigate before we can speak to a human. Most of us have also probably experienced the frustration of an overly friendly chat-bot popping up on the insurer’swebsite, only for it to fail on the basics.All too reminiscent of Microsoft’s “Clippy” Office assistant.
The latest advancements in conversational AI, driven by the LLMs’ understanding of language and meaning, now offer insurers and brokers the chance to redefine that customer experience. With access to the right data and underlying systems it’s possible to have more natural conversations to guide a customer to the answers they need. This isn’t limited just to supporting the end customer. Our clients use this technology to speed up solving IT helpdesk queries, guiding their partners to the right product and proactively finding customer service agents the right information they need andfreeing them up to have a better connection with their customer.
Removing drudgery
One of the ambitions for AI automation is to improve the efficiency of existing processes. Multiple hand-offs between systems, manual workarounds, and poorly designed flows all create friction. The first step in these projects is often mapping the existing processes. Traditionally this manually intensive process required interviews and sitting with staff to follow their process. One big challenge is staff missing issues they no longer notice, having learnt intuitively over time to work around them.
We are now working with partners who have leveraged AI to make this process mining much more efficient and effective. With the agreement of staff, these new intelligent monitors canwatch a person work across different systems and steps, looking for where people get stuck or waste time looking up information or fixing data errors. These new AI driven tools don’t just map the process but build the workflows to help staff overcome those delays and accelerate the route to greater efficiency.
Digital brain trust
While manyof the uses for AI discussed above are focused on efficiency improvement through task automation, there are also more creative ways to use these tools to add value. We have been building solutions used to answer questions over a very specific body of knowledge, for example helping staff better understand internal operating rules to help compliance or increase sales through a better understanding ofspecific product details. The advantage over traditional FAQs is you don’t need to anticipate every question or write all the answers. Product development teams are already using AI to analyse and understand social media feedstoidentify frustrations being discussed and have led to new, niche products to address those issues.
Finally,AI can also be used as a sounding board to testthinking, speed up research and act as another voice in the room when discussing ideas as a team. Health warnings are attached here though, and it’s important to fact check and use your own experience and knowledge to sanity check answers.The creative process can be helped along by using the right AI tools with the right data sources.
Mindfully innovative.
While the opportunities are exciting, and it is important to innovate and test out the potential of these new tools, there are some considerations.
First and foremost, having an understanding of the compliance, security, ethics and potential bias of these toolsis an important step. Along with creating the guidelines to help staff understand how to use the tools safely. Many of the open sourceand public tools take the information provided into their own training data sets, presenting possible confidentiality as well as copyright concerns, something which is already raising its head as an issue.
Our second recommendation is to look first for the business value and ensure you understand the problem to be solved. Can you see a route to your return on investment. How will you measure those benefits, and how will you track the delivery of those benefits. While this is good standard practice for a well governed project, in the excitement of what new technology can do, we see that this critical step can often not be given enough attention.
We also advise our clients that AI will not solve their problems alone – truthfully no technology will! Ensuring access to high-quality, sufficient data is vital for the success of projects where AI is involved. To be useful an AI system will also need access to other systems if it is to enable automation of processes or create a joined-up customer or staff experience. This has been a major challenge for the insurance industry historically which is still riddled with legacy systems and messy data silos. For AI to be truly game changing, we still need to solve that problem.
Probably most importantly the people in your business, whether they are your employees or customers, must be very much front of your mind for any project involving AI. Humans will need to be kept in the loopand step in when the AI doesn’t know what to door monitor its behaviour. When the plan is to use AI as a potential advisor or information source, winning the trust of those employees working with these new tools will be important for success.
Our experience has demonstrated that to succeed organisations need to do much more than just flick a switch to a new way of working with AI involved. The critical success factor of any AI implementation is not the technology but consideration of human nature. Working on the cultural change, as well as the operational and business change, will be the cornerstone of maximising the opportunities AI brings.
Doing more than adding AI plasters
To this last point, while many initial AI opportunities will create initial efficiencies, as an industry we can achieve much more if we shape our innovation more thoughtfully.
It is important to look beyond AI as just another digital plaster placed over old legacy systems and processes. In my experience underwriters will jump at any chance not to have to use the clunky policy admin system they’ve been required to use up until now. However, there are only so many layers we can put over these old systems. Too many layers make future change costly and slower to deliver. We need to be smarter with how we use these new tools to make a difference.
But there is also an opportunity to stretch our thinking and go beyond using AI as slightly more intelligent and flexible automation tools. Today’s processes remain fundamentally limited by historical designs,based on moving paper from one team to another to complete a task. When we combine AI with better access to data, new sources of data (like IOT) and different ways to interact with insurers, we can rethink the way we provide products and services to customers.
What’s clear is deeper thinking is required around what AImeans for the role of the insurance industry and the opportunity it presents.
Our inflection point as an industry is not that we use AI to create a smarter proposal form but to build a new service model where the proposal form is no longer required.
Uma Rajagopal has been managing the posting of content for multiple platforms since 2021, including Global Banking & Finance Review, Asset Digest, Biz Dispatch, Blockchain Tribune, Business Express, Brands Journal, Companies Digest, Economy Standard, Entrepreneur Tribune, Finance Digest, Fintech Herald, Global Islamic Finance Magazine, International Releases, Online World News, Luxury Adviser, Palmbay Herald, Startup Observer, Technology Dispatch, Trading Herald, and Wealth Tribune. Her role ensures that content is published accurately and efficiently across these diverse publications.