Most of today’s leading companies are now in the midst of digital transformation. While this is true across all industries, the financial services sector has been rushing to advance faster than most.
From banks and credit lenders to insurers and brokers, even the big players know that the disruption brought by fintechs in recent years means that change is now paramount to survival. Many strategies are being employed to make efficiency gains and improve customer experience, but few have been making as much impact as automation.
Photo: Sam Gross of ChoiceWorx giving a presentation on intelligent automation. (Photo credit: Loren Moss)
Same Gross, founder of ChoiceWorx and cofounder and chief technology officer (CTO) of the Digital Americas Pipeline Initiative (DAPI), is a thought leader in this realm with a long pedigree at companies including Unisys and CSC.
He has been involved in automation since long before it became a new-age buzzword and understands the evolution of its core concepts from small-scale timesavers into business practices that are now integral to the digital transformation efforts taking place across the globe.
To Gross, the most interesting development in recent years is the move to intelligent automation. “Intelligent automation is a mechanism that allows us to escape from that otherwise never-ending cycle of maintaining automation,” he says.
To gain more insight into this world and his perspective, Finance TnT Executive Editor Loren Moss sat down with Gross to talk about intelligent automation and why machine reasoning deserves as much attention as the other hot topics in the discussion.
Loren Moss: When it comes to technology, we have all these buzzwords: Robotic Process Automation (RPA), Machine Learning (ML), Deep Learning (DL), and Artificial Intelligence (AI), just to name a few. I have also heard you talk about organizations moving to adopt “Intelligent Automation.” What is Intelligent Automation in the context that you use the term, and how is it different from some of the other terms that we are throwing around today?
Sam Gross: First of all, let’s demystify automation. Automation is not new. It wasn’t first introduced to the market by an RPA tool. We’ve had automation for a long time. I would argue, if we think about, in the technology realm by itself, the first automation we had was something called “job control language” followed by the “scheduler.” These were automation tools that humans used by typing in commands on demand on a computer keyboard.
RPA tools use the same type of declarative model to create their automations. These require a human to actually decide what the next step will be — and then what the second and third step will be — and it requires us to create the declarative code, or case logic, to be able to handle every possible exception that we can think of. About 80% of all the code that we write for technology today is for the exception process. Only 20% is for what I like to refer to as the “happy path.”
READ MORE: RPA Implementation Proves Challenging, But Rewarding, for Insurance Companies
Intelligent automation is a mechanism that allows us to escape from that otherwise never-ending cycle of maintaining automation. It is driven by combinations of AI techniques — not a single AI technique — whereby the machine has a mechanism to learn how to make a decision around what to do next.
That’s done mostly in the realm of machine reasoning, with some machine learning techniques used, while moving from declarative code structures and driving automation to having machines learn and help you make those decisions — if not make them for you. That is how you move to intelligent automation.
Loren Moss: You use an interesting term and, knowing you, you didn’t use it by accident: machine reasoning. How is that different from machine learning?
Sam Gross: It’s hugely different. Machine learning is a data-intensive operation. You need to get large volumes of data in order to have an effective machine learning outcome. The problem is that most of the data we work with is actually flawed. It’s flooded because of the human factors that were associated with generating that data.
Machine reasoning differs because it does not depend on large vast amounts of historical data. Machine reasoning operates by understanding concepts, like events and states, and allowing humans to teach the system explicitly, for example, when this event occurred and include the other relevant pieces of data. Humans insert that data into a machine reasoning engine and then the machine reasoning engine has a mechanism to know where to start and then present the “next best action.”
“Let’s demystify automation. Automation is not new. It wasn’t first introduced to the market by an RPA tool. We’ve had automation for a long time.” – Sam Gross, founder of ChoiceWorx
What is a next best action? Next best action is very much like the recommendation engines that you see on Amazon. You know what I’m talking about, right? When you buy something, Amazon’s platform references the people who also bought that same product, and it knows what those people purchased next. That is a recommendation engine, which uses a set of mechanisms that understand the “distance” between different data sets. Obviously, the “closest” data helps make a better decision than the “farthest” data.
This is how machines are able to make decisions. And here is the thing about machines versus people. I can teach a person how to make really good decisions. But when they leave my company, all of that decision making leaves with them. All of that education leaves with them. All of that know-how leaves with them. It goes right out the door.
But when you implement decision making into the system through machine reasoning, that data stays there. How many times was a decision that was made a good decision? How many time was that decision a bad decision? When it was a bad decision, what where the associated elements? What were the elements when it turned to be a good decision?
That is how it learns. It learns by doing. The myth is that all technology must have a high level of precision. The fact is that humans don’t have a high level of precision. And we are insisting that the technology we use to replace them — if I could be so bold — has a higher level of precision than they do. But it is not necessary.
Loren Moss: When we talk about intelligent automation, is that the paradigm that we are aspiring to currently? Or are we there today? Are we on our way?
Sam Gross: We are there today. But it is early. The technology works. We know how to build it. We know how to deliver it. We know what kinds of use cases we can deliver with it. And we see it in different parts of our life. We are just not always aware that we are seeing it when we are seeing it.
Loren Moss: What should students who are interested in this — young people who want to participate in it and be part of this future — study today? How can they be preparing themselves so that they are qualified to help shape this future?
Sam Gross: I’ll give you one of my favorite little sayings. It used to be said that the meek shall inherit the earth. That is not true any longer. Mathematicians will inherit the earth.
If algebra is key, algorithms are algebra. You need to know how to write an algorithm to work in the field of automation, intelligent automation, and artificial intelligence. But it helps a lot if you really understand the science underneath it as well.
READ MORE: RPA in Action: Innovative Ways Five Financial Services Companies Are Using Automation