Enterra Solutions CEO Stephen DeAngelis on AI in Legacy Software

I spoke with Stephen DeAngelis, CEO of Enterra Solutions, about using an abstraction layer to enable AI to power legacy enterprise apps.
Among the topics we discussed: 

As you survey how companies are using generative AI in enterprise-grade business applications, what trends do you see in 2023?
What recommendations do you offer to companies to optimize their enterprise use of generative AI – securely?
How is Enterra addressing the generative AI needs of its clients?
The future of generative AI in the enterprise? I think it may be a more unpredictable force than some executives realize.

See below for a transcript of interview highlights.
Listen to the podcast:

Also available on Apple Podcasts
Watch the video:

This transcript of interview highlights has been edited for length and clarity. 
eWeek: Many large enterprise software applications are aging legacy applications. What about building artificial intelligence into one of these legacy applications? Is that actually going on, or is generative AI going into only the newer enterprise applications?
DeAngelis: For last 40 years leading companies like SAP, Oracle and Salesforce and others have built what’s called transactional systems of record. Those were systems that were meant to record transactions, not make a mistake at scale, but were never engineered to dynamically think about changeable data in the marketplace and help companies navigate in a real-time environment their business.
Today we have uncovered a tens-of-billion-dollar market niche for a thing that we call a system of intelligence. Think about a system of intelligence that sits on top of the data and process layers that reside within the transactional systems of record. Now you have a system of intelligence which creates the ability to dynamically sense, think, act, and learn about the environment that businesses are operating in. So you can navigate that, you can navigate a changeable landscape dynamically and bring the concept of autonomy to the enterprise.
So what happened was companies need to be nimble to a degree that they haven’t been able to be nimble previously. And the artificial intelligence that we were able to use [to build intelligent systems] allowed companies to analyze data, generate a recommendation, make a decision at the speed of the marketplace where you couldn’t have humans manually cranking analysis by hand. These technologies allowed the leading companies to navigate effectively those dramatic marketplace changes that we’ve experiencing over the last several years.
Also see: Top Generative AI Apps and Tools
eWeek: It sounds like it’s an abstraction layer over an enterprise application. So you’re saying there’s a system of intelligence – an AI layer above that old fashioned legacy application – and it’s actually interacting with it between, say, the human or the system and the application itself?
DeAngelis: So this is truly creating an intelligence layer that is that abstraction. I think of it as a common brain that sits on top of the transactional system, generating insights and sending instructions to the transactional system to say: system of record, do this and do that. And then it gets the results, learns from it, and then uses that learning to inform the next decision it makes.
The abstraction layer is system integrated to the transactional system of record. I mean if you think about it in a technology architecture, it’s above it, but it is systems integrated too.
So for example, when we would do a trade promotional recommendation for a consumer packages firm, the recommendation is autonomously generated by the intelligence layer. And then the instruction gets sent through an integrated connection point to the transactional system for execution. So there’s a two-way street connectivity to that system.
It is a very, very powerful combination – the next step in the evolution of enterprise computing is to leverage the legacy investment. You don’t want to replace the legacy investment that firms have had because quite frankly, SAP and Oracle, they’ve done great jobs in the transactional systems of record. This now lengthens the time and durability of those systems of record and allows companies to leverage their legacy investments.
eWeek: Other than this advance with AI, how can companies optimize their use of generative AI in their enterprise infrastructure?
Well, part of the challenge, James, is that generative AI is an emerging technology where you’re seeing the tip of the iceberg.
And quite frankly, I think that generative AI is going to be, to a large extent, the provenance of large tech firms. Because they’re going to generate the massive large language models and they’re going to have to, over time, validate the large language model.
Because right now the problem with generative AI in scalable enterprise applications is that you have the “garbage in and garbage out” problem on steroids. You can’t validate that the information that the gen AI is gathering from the internet is efficacious.
So large tech firms like Google and Microsoft and Amazon and others will generate the fundamental large language models and the ChatGPT capabilities. But those are going to take tens of billions of dollars to build and curate over the next decade. Much like Amazon spend tens of billions of dollars to redo the retail online, this is the same kind of mass investment.
But then you’re going to see firms like ours and others using generative AI in conjunction with other artificial intelligence and machine learning and data science applications to perform optimization in decision-making in very discreet but yet transformative functions.
eWeek: What about the future of generative AI in the enterprise? What sort of developments do you see?
DeAngelis: We’re building what I call intelligent agents. Think of that as anthropomorphized knowledge buddies that sit within an organization. So let’s say you’re a supply chain planner at a company. You’ll be able to – within the next 12 and 18 months – have a anthropomorphized knowledge buddy [that assists you]. Think about a version of [Star Wars] R2-D2.
So working next to you, it says, Hey James, this happened last overnight. Our production plans were affected by A, B and C. Here’s our recommendations for you to navigate that supply chain challenge and here’s the recommended course of action. Would you like me to execute that?
So that’s going to leverage the conversational AI part of gen AI, right? But it’s also going to leverage the other technological assets I mentioned to you earlier. So what you’re going to see now is we’ve been very accustomed to autopilot [on airplanes]. You hear planes flying with autopiloting, you’ve heard of autonomous vehicles, right? Now the concept of autonomy is going to be brought in pieces to the enterprise. Gen AI applications will augment, not replace, the humans playing roles along the continuum of the value chains of large, mid-size and small companies.
So you could spend more time thinking about the strategic application of what you do and how to use your decision-making ability better in your day-to-day operations. And I think it’s rather exciting.
Also see: 100+ Top AI Companies 2023
The post Enterra Solutions CEO Stephen DeAngelis on AI in Legacy Software appeared first on eWEEK.