Pat’s Couch: Machine Learning Overview

Posted by Brinna Hanson on December 30th, 2019
Author
Brinna Hanson
Brinna is a marketing professional and recent graduate of the University of Minnesota. Brinna joined Smart Data Solutions in 2019 to assist the marketing department reach new heights with a focus on the HubSpot inbound process. From her time at Smart Data as well as at previous internships, Brinna has been able to gain knowledge in many different aspects of marketing as a whole.

Machine learning is when AI (artificial intelligence) is applied to a system to give it the ability to “learn” and improve on processes without having to be updated or programmed to adapt to change. Essentially, the system will learn from repetition and pick up on certain templates or processes that are similar and automatically take the appropriate actions.

It has incredible benefits such as eliminating human error and putting out the same amount of work with fewer people, which reduces labor and turnaround costs. OCR and AI work together, along with the program’s machine learning in many of our processes at Smart Data Solutions.

You might be asking, how can machine learning actually be utilized in the industry? What impact does it make on process improvement? Our co-CEO, Pat Bollom, took the time to sit with Machine Learning and OCR Specialist, Johnny Bouthilet, to talk all things Machine Learning here at SDS. Watch the video, or read our transcript below for a little snippet of what machine learning is, and how we use it at Smart Data Solutions.

Pat: Why is the use of machine learning so important here at Smart Data?

Johnny: It helps us optimize things, and gets it out of the hands of manual labor so we can automate things away. With that it will give a more standardized approach to things. Such as, the machine will always give the same answer time and time again, while a person over time might start changing answers or become biased towards different answers. It really helps us stay consistent and efficient.

Pat: What sort of results are we seeing?

Johnny: We only just started a few years ago with it but so far we’ve increased 60% in automation and everything is double keyed so that ends up being an 80% reduction in keying overall. So that means 60% remains 2 keys, and 40% is 1 key, so that gets us about 80% of total keys is automated away. (chart below)

Pat: What does this mean for cost savings?

Johnny: Everything is double keyed, so with the automation we’ve reduced 80% of all keying staff, so someone is overseeing 20% of all keying when we were looking at 100% before.

Pat: Smart Data’s biggest cost is labor. The more labor we can displace, the more profitable we are, the more savings we can pass on to our clients. So it has been substantial.

Pat: What other processes are you working on where you can introduce machine learning in those high volume processes to save money?

Johnny: The guideline is, any decision that takes about 1 to 5 seconds can typically be automated away using machine learning. And in general data science has plenty of applications that you can use machine learning to automate those simple processes in ways such as identifying documents, fraud detection, and double-check things to make sure they’re going the way they are. There are other statistics and data analysis that end up using that stuff as well.

Pat: I love it! So how many machine learning folks do you have here at SDS?

Johnny: Right now the team is about five, but steadily growing. I’ve been passing along knowledge and increasing the team as we grow here.

Pat: The big question is regarding one of the biggest issues in machine learning, which is the data labeling, the cleansing of the data, getting good data. How are you overcoming the issues with the bad data — the data that is not consistent, not quality?

Johnny: The first thing is, having great data is going to beat out any fancy algorithms pretty much anytime. So, the first thing we want to do is make sure we have a valid pipeline to help clean that data as fast as possible and as efficiently as possible. Having bad data in the model can result in pretty drastic drops as far as efficiency and accuracy, so we always make sure to have that. Unfortunately, it just comes down to that brute force just to make sure that everything is there, build pipelines, and make sure we have the proper data.

Pat: There’s so much more to machine learning than we can cover here, but if a client wanted to talk more about machine learning, who would they talk to?

Johnny: They can contact me!

Pat: Any last thoughts on AI, machine learning, deep learning or anything like that?

Johnny: It’s been said that artificial intelligence and data science is kind of the new electricity, in that people have seen electricity applied to all kinds of different fields. It’s still relatively new, so we are trying to figure out where it can best be applied but we’re still playing with applications and new ones that haven’t been tried yet. We’re excited to find some problems and try it out!

As you can see, machine learning is just the cusp of how AI can improve business processes, and especially our processes through Smart Data Solutions. From eliminating time-consuming manual work to double-checking data for accuracy, machine learning is truly the way of the future. We continue to improve upon it and develop new ways to save time and money for ourselves and our clients. If you’re looking for a way to up the ante, and rid yourself of cumbersome hands-on work or too many vendors, contact Smart Data Solutions for an intelligent approach to accurate data.

Leave a Reply

avatar
  Subscribe  
Notify of

Recent Articles

View All 

Why 2020 is the Year to Get Rid of Your In-House OCR ...

If you’ve been using an OCR (optical character recognition) technology software, then you understand the benefits of ...
Read Article

Healthcare Claims Processing Workflow: Tools and Processes ...

If we're going to talk about how to improve efficiency within healthcare claims processing, we should start at the very ...
Read Article

The Steps of IT Risk Management in Data Capture

IT risk management is defined as the procedures an organization puts in place to protect data from potential risks and threats, ...
Read Article

Why handle the heavy lifting of your claims workflow when a smart team enhanced by AI can?

Achieve peace of mind, reduced costs, and greater process efficiency by automating and consolidating your data workflow.

Start Customizing Your Solution