My sales guy, Mike, lectured me recently: "Why aren't you talking about Technology Enabled Policy Review in your blog posts?"
It's taken me awhile to warm up to the idea of talking about our technology for a few reasons.What is this crazy insurance technology we created?
It's hard to talk about a technology when you aren't sure what it is.
For some time, I wasn't sure what we had created with RiskGenius. Our team has played around with a lot of names for the RiskGenius technology. It's both fun and daunting to create a brand new technology category, something that has never existed before. Back in 2014 and 2015, we called it Collaborative Contract Review. There is certainly a collaboration component to our software, but when we went to market with a beta version, users said they also wanted more, they wanted the machine learning that we were promising to build. So we built the policy machine learning and we settled on "Policy Analytics" to describe what we had created. But that name wasn't correct either. When people hear analytics, they think of Dashboards and Business Intelligence. So then I took a cue from other machine learning companies and started calling it "Machine Learning for Insurance Policies." I've always hated the phrase "machine learning" though -- it suggests technology that is unapproachable and fully automated.
I believe we have settled on the correct term, finally: Technology Enabled Policy Review.
If you have ever worked in the legal industry or followed legal technology, Technology Enabled Policy Review (TEPR) will sound familiar. In the legal industry, "Technology Assisted Review" has changed the way litigators review documents. Instead of reviewing millions of documents, lawyers can now use fancy machine learning algorithms to sift through and identify the most relevant documents in seconds. As a recovering attorney, I always kept this technology in the back of my mind; I wanted to apply it to some process in the insurance industry but I wasn't sure how or where.
Then the idea struck -- Technology Enabled Policy Review.
"Technology Enabled" properly describes how our technology and algorithms work. Machine learning requires human input. We have invested ridiculous amounts of time perfecting our Parsing Algorithm and our Categorization Algorithm. This investment of time will never end. We will continue to improve our algorithms by feeding new data, looking at results and adjusting our rules and indexes. It's the nature of the beast (imagine if Google stopped indexing the Internet).
However, "Technology Enabled" also describes the end result for RiskGenius users. These users are enabled to review insurance policies faster. Humans still have to review the policies. By categorizing clauses for our users, we just speed up the process for RiskGenius users. Imagine if you could instantly review "War Exclusion" clauses across fifteen commercial general liability policies -- that's how our software enables users to review clauses faster and smarter.
Why am I avoiding writing about Technology Enabled Policy Review?
You are going to see me writing about Technology Enabled Policy Review a lot more in the future. But like I mentioned, I resisted writing about our technology initially.
I was worried someone would steal our idea.
Going forward, in order to properly describe Technology Enabled Policy Review, I am going to have to describe our algorithms. For example, this article is the first time I have disclosed our Parsing and Categorization algorithms. I am okay with disclosure for a few reasons.
First, I recently had to write a memo for a prospective customer about our technology. I found myself dancing around our technology and how it works. "Maybe the client will steal our idea," I worried. But I decided to throw caution to the wind. I detailed how a policy hits our Parsing Algorithm and our Categorization Algorithm. I explained that we write new parsing rules for differences in formatting across carrier and broker documents.
I nervously sent the memo and something amazing happened. The memo started a conversation with the prospective customer about our technology. They asked questions and we provided answers. And then we moved forward to a contract and negotiations.
Since then, I have tried to practice what I call "radical transparency" with customers and prospects. For example, we don't cover all lines of business (yet). We will shortly be publishing a webpage that details what our software can and can't handle. It will also include an email newsletter sign up for updates and a schedule of anticipated release dates.
I am also okay with disclosure of our technology because what we have created is really, really hard to build. This where I have to give huge credit to Dan Burtchett (CTO) and Doug Reiser (CIO) for digging in and doing the extremely hard work of building RiskGenius algorithms. If you are looking to emulate the RiskGenius software, it will take you at least three years to build the algorithms. That's not bragging, that's the cold, hard truth of trying to create truth data from insurance policies.
Finally, I am okay with disclosing our technology because it's working. RiskGenius was not always working and I plan to write another blog post about this in the future. The architecture behind our algorithms is so much more complex then our initial build out. We have onboarded customers who immediately figured out how to break our algorithms (darn you Chubb policies with your weird line breaks!). We are indebted to those earliest adopters that helped us break stuff, learn from it, and fix things.
Why am I telling you this?
As I thought about our software journey, I realized something: the insurance technology industry is nontransparent.
I'm not sure why that is. But I see "radical transparency" as a market advantage for RiskGenius.
- RiskGenius is Technology Enabled Policy Review; and
- If you have a question about Technology Enabled Policy Review, you can find me on email, twitter, or LinkedIn. Drop me a message and let's talk.
What do you want to know about Technology Enabled Policy Review?