“The information you have is not the information you want.
The information you want is not the information you need.
The information you need is not the information you can obtain.
The information you can obtain costs more than you want to pay”
― Peter L. Bernstein, Against the Gods: The Remarkable Story of Risk
At the end of the 20th century, as web technologies began to go mainstream, the term “information superhighway” was 
coined to describe how the piping infrastructure of the internet would allow information to travel frictionless from portals to users and back to portals. As internet adoption exploded, the infrastructure evolved to support the tsunami of information we began feeding it.
Fast forward to today, and the infrastructure of the information superhighway is one of our historically great technological achievements. Our ability to download and upload gigabytes of data takes seconds or minutes to accomplish. As the volume of information continues to grow exponentially, the issue is no longer about the capacity of the infrastructure to allow information through. Just like when the internet began getting wide public traction, a new and more complex problem arose…finding the needle in the haystack.
The Evolution of Search
Information lacks value if you can’t find it!
Search engines were initially designed to create a directory or index of files that were stored on the internet. They quickly evolved to index the webpages themselves and then another evolutionary leap to index the hyperlinks and keywords within the web page…all in the quest to make it easier for users to find information (here is a well-written history of search that is informative and an easy read).
Google’s Page Rank algorithm changed search entirely. Before Google search results were problematic as sites could game-the-system with keywords. Way too often, I’d have search results come back from foreign countries in foreign languages that had nothing to do with my search request but just happened to score highly because of the keywords hidden on the page. Google’s algorithm looked for relevancy and found clever and statistically valid ways to compute relevancy. My first Google search was an A-HA moment. The clean, white webpage (no clutter, no banner ads). The simple input box, and the very fast and very relevant results, right there, right on page 1. There was no going back!
Lack of Progress in Search within Organizations
What has worked for the internet, has not exactly worked elsewhere. There is often frustration in search when it comes to finding files, finding emails, finding that announcement from management…What worked for internet search doesn’t exactly translate well to search within your organization. Existing approaches are rigid, and employees don’t get useful answers.
This is something we have focused on for a few years now. It is especially relevant in insurance, where there is just so much information that constantly changes in a dynamic environment. It just takes much too long to find the information we need in the time that we need it.
Users want a Google-like experience of entering keywords or a question and, in return, getting a (short) list of results that allows them to move forward. Unfortunately, within organizations, search is broken. We need a new technology.
Knowledge Management & The AI Revolution
ChatGPT may have ushered in AI mania in 2023, but AI has been around for a while now, and it is something that has been part of our solution for years. ChatGPT’s conversational abilities have inspired new ideas about how people can interact with data and information.
And what we’ve seen from the data in our platform is that the way users find information differs for each user.
Some like to use keyword search, as that’s the world they’re familiar with….essentially CTRL-F on steroids.
Others like to use filters, tags, and drop-downs to narrow down to the resource they’re looking for and find information that way.
Now, the promise of these new Large Language models is the ability to use natural language, where we don’t need to guess what keywords will get us the result we need. All we need to do now is ask simple questions, and these models can do that work for us. Let me explain
In a typical search engine scenario, you punch in a few keywords, and the engine returns results to you. But what if those results are not what you are looking for? With search engines, you start over and keep refining your search terms until you get what you need. You may add more qualifiers to the keywords, or you may try a whole new set of keywords, but you are essentially starting over each time you refine your search. This can be inefficient in most insurance operations, especially those with call centers and departments where speed of response is crucial.
What the AI craze of the past year has shown us is that we now have tools we can engage with as if we were talking to an analyst.
“What is the sub-limit for backup water and sewer coverage for the 2022 homeowners form?” or “What was our company’s depreciation expense for quarter 1 of 2021?”.
Hmmm…that isn’t what I was looking for…” add the depreciation expense from quarter 1 and quarter 2″.
Still not quite what I am looking for…” What is the delta between the depreciation expense from quarter 1 and quarter 2 of 2022 vs 2021?”. “What is the 5-year trend for the first two quarters?”.
If you can imagine the back and forth, you can imagine how much less friction there will be to get the information you need much quicker. This is the promise of AI.
What Not To Expect From AI
AI is complicated technology. Not all AI is custom-tailored for each task. This is especially true in insurance. Slapping ChatGPT onto your internal document stack will create a conversational-like feel to it, but, the ChatGPT models were not trained on YOUR data nor insurance. They can “hallucinate” or generate inaccurate information when the training data is incomplete.
As the title of the article suggests, there is just way too much data, knowledge, and information in insurance. We see it with our clients daily. When it comes to Knowledge Management, it is important to recognize where the inefficiencies come from and where the solutions are likely to come from as well. We have been big believers in how AI will help us solve this problem. But be wary of the quick fix. Not all AI is the same.
As you prepare your organization for all the information capture that you will need to preserve your institutional knowledge, think about how your employees are going to need to engage with the technology to efficiently get at that information. Remember: Information lacks value if you can’t find it!
