Encapture with Will Robinson | E262

Using machine learning to extract and recognize data.

On today's episode of the Fintech impact, Jason is going to talk to Will Robinson, CEO of Encapture. The company brings machine learning to banks and lenders to help them understand what is going on with their data and make better decisions. Anywhere anyone signed, any place someone showing up and having to fill out paperwork or submit supporting documents to do something at a bank is where Encapture can get involved. 

Episode Highlights

  • 0.58: Encapture is a machine learning platform in the intelligent document processing space. They are good at finding and extracting important info out of documents. 

  • 1.29: Encapture makes it easy for the bank to identify key information, identify the correct documents that they need to complete a task. As per Will they save a ton of time and effort of banks around manual data entry or the manual stare and compare that a lot of them are used to.

  • 2.11: Encapture is a professional services or consulting company in the starting, and probably 15 years ago Will started building and Encapture their own product to fill in some gaps in this market around collecting documents and identifying important information.

  • 4.02: Will explains how they collect documents required for loan processing and helps to save time and manual effort. 

  • 05.22: Will talks about compliance reporting. He says that there are different regulatory agencies that require banks to report certain information on the loans that they give just to make sure that they are not discriminating in their lending process. 

  • 7.36: The big impact over the last several years is the ability to read through unstructured documents, says Will. 

  • 09.30: There are a bunch of different analytical applications for machine learning as well coming through large datasets within the bank to look for patterns like fraud or some of the discriminatory lending that banks are keen to make sure that doesn't happen within their organization, says Will.

  • 11.36: The fundamental premise of machine learning is that you train it on datasets. Larger datasets generally can lead to better results if those data sets are high quality and the system over time. If it has a proper feedback loop built into it, it can get smarter over time.

  • 14.00: Will explains why it is important to feed really good clean data into your model in the first place.

  • 15.45: Jason talks about how the relative tightness of Encapture's data set really does play in quite well into being able to have that human versus machine interaction. 

  • 18.10: Will explains how some of their prospects are familiar with machine learning, how it works, and they understand kind of the limitations of it doesn't do everything and it doesn't do everything 100% accurately. 

  • 19.00: Will provides detailed insights on how they feed a lot better data so they can make better decisions. Humans can make better decisions, or existing processes can make better decisions, so they get people around that. 

  • 20.15: Will says they have got a lot of room to run at this point in the in the lending space. They do commercial lending, auto lending both direct and indirect like through dealerships. They work with a couple of consumer kind of fintech lenders that are doing real time.

  • 23.37: Will says that they kind of take the mindset if we can go solve everything there is to solve with the current product they have, then they will think about doing more. 

  • 24.38: People are untruthful in the front end, but it does feel like there is a lot of swirl and buzzwords in our space. That can kind of clutter the value prop and sometimes make it hard for the folks who are doing great things and not just us, says Will. 

  • 25.39: As per Will getting people to really buy in and believe into where they are going has been harder than he had thought. 

  • 27.17: Will says there are so many problems out here. So many things to get solved and the way that we even solved them, we are getting better at that. 

3 Key Points

  1. Will shares how they are using machine learning to save lot of time and effort in data processing. 

  2. Will explains how training a machine learning system is very similar to feeding it data. There are different methods of training where you feed it massive datasets, like thousands or 10s of thousands of samples of documents and you let the machine learning kind of figure it out on its own and come back to you with results. 

  3. Will explains how they are not making decisions about credit or about new account openings, they are simply looking for the data in the documents and making that hunt and peck a lot more efficient. 

Tweetable Quotes

  • "Our focus of the last several years has been in developing our own technology and trying to find some compelling sticky use cases in the financial services space to apply our tech." - Will

  • "There are probably 30 examples that we can give as tangible to bank and so much of it is dependent upon who are we talking to in that meeting." – Will

  • "We can actually go through all different loan documents, go through the entire loan package, find all the key data they need, pull it out for them, and allow them to report it with a lot more accuracy and candidly a lot more speed than they would do it manually." - Will

  • "It's incredible how far machine learning has come to mimic kind of a natural conversation between a system and between a person." - Will 

  • "There is a super long tail of community banks and kind of smaller regional banks that would love this technology that has never been exposed to it at all." - Will

Resources Mentioned

Fintech ImpactJason PereiraFintech