by John Cambier
The tightness of the current labor market has, among other things, increased the urgency and budgets for companies to streamline and automate workflows. With 11.3M job openings in the US at the end of January 2022 — up 60% from a year ago — it’s now less about taking cost out of the business and more about getting business done at all. With the “Great Resignation” upon us, most companies will not be able to hire their way into solid growth.
The bulk of today’s US economy, nearly 77% as recently as 2018, is in the services sector; healthcare, retail, hospitality and financial services among the largest of those. Of these four sectors, financial services has, by far, seen the greatest strides in using technology to simplify and automate processes. For example, think of the ease at which you can access and move money from your bank account or trade stocks(or cryptocurrencies). Conversely, think of how painful it remains, for both the customer and service provider, to open a bank account, apply for insurance or, heaven forbid, obtain a mortgage. Why is this?
These activities all rely on the collection, transformation and analysis of unstructured data. For instance, when we at IFP renew our D&O insurance each year, we have to upload financial statements, fund documents (contracts)and a completed application with various information about our business holdings. All of these are PDF documents and none of them are formatted the same way or with the same data from one firm to the next. This makes it a very hard process to automate as humans must read the submitted documents to extract the needed data, most likely re-entering that data into a spreadsheet or system of record where it then can be compiled and analyzed.
Some of you are probably thinking “this is what RPA (Robotic Process Automation) is for!”. Well yes, but RPA doesn’t have any intelligence. These tools do not have the ability to look through a document and pull out the salient parts if the document does not have the exact same format every time (unstructured data). So you might then suggest that you teach an (ML) model to look for and identify the salient parts of a document for extraction? The issue with the (potential) application of AI/ML to this problem is that, until now, you have generally needed a (very) large dataset on which to train the model and many of these tasks are serving the “long tail”; i.e. a very large dataset does not exist.
This is where our newest portfolio company, Sortspoke, shines. Its intelligent document processing platform helps companies extract any data from any kind of document. With its intuitive user interface and powerful ML engine, Sortspoke can begin to identify and extract data from documents with as little as 10 training pieces. This means that in less than one day, Sortspoke can start to increase the efficiency of humans, ultimately by as much as 10x.
Headquartered in Toronto, Canada, Sortspoke was founded by Jasper Li to solve a problem that he had seen throughout his decade as a consultant with PWC where he helped companies streamline and transform operations. When we saw the pitch at a demo day last Fall, we saw a company playing in a large and growing market with compelling customer adoption and strong founder/market fit.
As we worked to understand the market and competition better, it took a good bit of research and conversations with existing and potential customers to understand where Sortspoke fits and why (we believe)it has meaningful differentiation. You’ve got general purpose AI/ML platforms from giants including Amazon, Google and Microsoft on one end of the spectrum and RPA platforms from fast-growing providers such as Automation Anywhere, Blue Prism(now SS&C Technologies)and UiPath on the other. In the middle you have various venture-backed companies including Indico Data, Hyperscience and Workfusion.
In speaking with customers across the insurance vertical, we found that many (if not most) have tried and failed with one or more of these competitors. Though many of them advertise the ability to work with unstructured data and quickly get to productivity, the reality invariably falls short of the promise for many applications. When given the chance to try Sortspoke, customers are amazed at: a) how easy it is to use and b) how quickly it becomes self-learning and productive.
Jasper and his small team have accomplished an incredible amount with the money raised thus far. They have a great product and happy customers with strong organic growth. We’re excited to have co-led a USD $4.5M Series A with a major US insurer to provide Sortspoke the capital it needs to go after the market aggressively.