Enterprise Software and Agentic AI: A Discussion with Vista Equity Partners
Robert F. Smith Q&A Session
On Government Systems and AI
I've always kind of marvelled at how government systems functioned. I knew there was a human side because when I needed to talk to somebody in the government, I could get a human being on the phone and they, in general, were good. I ran an SBA company for a while and was really happy with how things worked in Michigan and in Washington.
Recently, we went through this period of time with Elon Musk and DOGE. But when I'm listening to you and thinking about agents—which seem to me to be really replacements of human beings in terms of managers—I may not have understood it correctly.
The Impact of AI on Knowledge Workers
Question: Are the processes you're talking about potentially going to replace millions of jobs? And if so, Trump should have called you and said, "You know, what do we do?" How do you think AI will impact the workforce? Are we going to lose millions of jobs, and if so, what does that mean? It's not just in government—now when I'm on the internet, sometimes I get someone asking "Do you want to chat?" I always figure it's replacing a person.
Robert F. Smith: The short answer is it is going to impact every single knowledge worker. Every person that you've called in the past is a knowledge worker. And in some cases, it will augment and enable them to be more effective—pull up data and serve you. In some cases, you won't be able to tell the difference.
We have in some of our companies, for instance, we're now doing what we call "deflection rates." We can deflect customer service calls. One of my companies has a deflection rate of 65% in a consumer-based business. We only have a couple of consumer-based businesses; most of ours are B2B.
In this case, it's used for gyms. You have a problem—you booked a session to go to SoulCycle or whatever it is, and you showed up and the trainer gave the seat away to his girlfriend or whatever, and now you're upset and you want your refund and all that sort of stuff. We now have a 65% deflection rate and a higher net promoter score—a higher satisfaction rate with our customers. And as a result of that, we don't need 65% of the customer service people any longer.
Global Policy Implications
There will be certain countries that are going to have policies that you are not going to like. I know certain countries are saying, "Yeah, that productivity is great, but you're not going to fire people." And certain countries are going to be like that. Other countries are not, and other countries are going to say, "Let the free markets end up where they are."
There's no question: there are 1 billion knowledge workers on this planet who make up about 40 trillion dollars of economic activity. That's just their salaries and wages. All of them are affected.
Now, are all of them going to lose their jobs? I don't know. Are some of them? Absolutely. Some will have to get retrained, reskilled, repurposed. But that is an absolute fact that is going to occur based on what this technology is capable of doing.
I showed you that video. This works today. In essence, that's seven agents and one administrative agent that will out-work that one person—who Brady had a picture of—by a hundred times. Do you still need Brady? Should you keep Brady? Those are questions that we're all going to have to wrestle with as a society.
Data Transformation and AI Readiness
Question: What about the data that's out there in the world right now that you have in filing cabinets? Sometimes your outcomes are only as good as the data that you have. It seems like we're in a transition period where there's still a lot of data in filing cabinets or in 1980s-style databases. What's the transition look like from where we are today to where that becomes usable data?
Robert F. Smith: Part of when we build our factories, we build a whole system for what we call "factory readiness." In many of our cases, we now have the ability to take this unstructured data—you can take a filing cabinet, you can just feed it in. It doesn't have to go through OCR and it becomes part of the dynamic.
We did that for our contract administration. I had this built internally. Kieran's running around here somewhere, and he and his team did that. I now can plug our system in—if we're buying a company that's got 10,000 customers and 40,000 contracts, before we'd have to find lawyers and give it to them, and a few million dollars later, we'd have some understanding of the topography of the contracts.
We can plug it in and literally in hours have that completely ingested. It can be in paper form or data set form, with a higher fidelity. I know exactly which customers you should raise prices on, which ones have these terms.
Question: OCR is at the level that it can just read that today?
Robert F. Smith: You don't even need OCR. We literally can scan, and the understanding of what that contract actually is—and put it in the context of what we're interested in—works seamlessly.
In software, our relationships with our customers are governed by contracts. Understanding those contracts when you're buying a company is one of the most important things you can do. We now have an AI engine that does just that, built for our own use and purpose. That gives you some sense of the capability.
End of Q&A Session