Palo Alto Studio Connection Silicon Valley and Wall Street. I'm John Furrier here with Dave Vellante, my co-host.
Welcome back to the CUBE studio here at the New York Stock Exchange. I'm Gemma Allen, co-host of NYC Wired's mixture of experts, and my next guest has spent 23 years building a company that many enterprise credit with turning robotic process automation into its own category. They are now trying to do the same thing again at agentic AI. Automation Anywhere is running north of a billion IT service requests through its autonomous service desk. CEO of Automation Anywhere, Mihir Shukla. Great to have you.
Great to be with you here, Gemma.
So, maybe just to start, we talked about kind of RPA there a little bit in the intro. It's a term we all knew well for a long time. It's had a little bit of a downer moment though, right? It's being replaced with agentic process automation. Help us understand—help level set with me and the audience—what is actually happening.
Okay, how about I set us up on—so today Automation Anywhere's vision is to solve one of the hardest problems in the AI industry, which is your AI spend is here, and the level of autonomy that AI is achieving is here, and that gap is widening. We know this through other large technological revolutions, from electricity to internet, that technology by itself doesn't create the impact; it's the level of autonomy if you achieve with that, it will change the entire operating model of the society and the companies. So that's what we have to achieve with AI, and that's what Automation Anywhere is trying to do.
Our answer to you, as you rightly said, is that agentic automation is how we are doing it, relying on our heritage of automation with machine learning and other models, RPA and other things. But now with agentic, we are taking a few steps up. Now with agents and the combination of other technologies, we are able to reason what to do, decide what to do, and sometimes take a human input, and then execute an end-to-end process with governance and security.
Today, Automation Anywhere stands for autonomous work. That's what we stand for, and we do it in two, three different ways. We offer a platform, Agentic Automation Platform, that's the number one platform in agentic automation today. We offer out-of-the-box agentic solutions across functions like IT support, HR, and many others, but verticals in healthcare, manufacturing, and banking and financial services. And we do it across thousands of customers, about 30 to 40% of the Global 2000.
So let's talk about your first point there, which is that spend on AI and tokenomics and everything, it's here, right? It's constantly headline news. And actual realized value within agentic and enterprise is somewhat of a debatable topic, right? There is certainly a wedge to overcome. When we think about the parameters of agentic for enterprise, we're used to a world where we could tell, run a script and tell a system, "Give me this data, do this." The world of agentic though is far more than that, right? It's about actual execution. So when you think about that gap, what are the parameters of what we are actually talking about here? Are we talking about systems operating collectively and completely autonomously, agents making decisions that humans made two years ago? Help me understand where you see that gap being the deepest in terms of what the market says versus what enterprise realizes.
Please. How about we do it with a few customer examples?
So take an example, and I'll give a few—maybe I'll give three versions of it. So there are customers who are identifying one process, two processes, and trying to get going. Then there are customers who are saying, "I want to do a large process end-to-end, order-to-cash, reimagine my entire business around that process." And there is a third class of customers who are saying, "I'm reimagining the whole department and how do I operate." So you get different varieties of how people are thinking about it.
Take an example of NHS UK, their healthcare system.
One of the largest trusts there came to us and said, "We want to reimagine the entire workforce of 18,000 healthcare people in their system, and what would it look like? Could we get 70% of automation of everything that was today happening manually or through other old technologies?" So we came in, and we looked at their scheduling, how emergency room visits happen, how do we plan compliance along every aspect of healthcare, and looked at the entire problem and did an agentic workforce management solution for them. Now, that's an example that ultimately saves a lot of cost because you are preventing temporary staffing that was required. If you plan early, you don't require that, and ultimately it improves patient care. That's one example.
Another example: let's take an example of Petrobras, Brazilian oil and gas company. We took their autonomous tax, which is in the finance world. So we took a complex Brazilian tax code—I'm told it's one of the most complex in the world—and we computed taxes for a subset of their business. In three weeks, we were able to save them $120 million.
In both of these examples, the way these solutions work is a collection of agents that are autonomously figuring out certain things, like the tax code, figuring out what the thousands of pages of tax laws mean and trying to guide what is the right tax law to apply. A set of tokenless technologies, so more deterministic technologies like APA, RPA—
Machine learning comes in. Multiple of these technologies come together to create that autonomous process end-to-end. And for those of us who have been doing it for a while, this much is clear: that the answer isn't just one technology all the time, because if you have a deterministic answer to something, making it probabilistic is probably a bad idea. So a collection of these technologies creates a seamless autonomous experience and changes a complete operating model.
So if we take the NHS as an example and talk about this world of agentic, we know that historically in an enterprise or an institution like the NHS, there were challenges with line-of-business migrations to the cloud, right? That happened across many government organizations the world over.
Credential access, depending on your domain accessibility for a clinician versus a manage member of the executive—systems didn't really connect or talk to each other. There wasn't the same level of interoperability, and a lot of that was because the governance wasn't in place historically, the structure, perhaps even just the actual workflow itself had never been fully mapped or fully discovered, right? When we think about agentic, it seems as though a lot of problems were solved in almost like a flick of a switch. How were some of those legacy challenges, like structured data, clean data, accessibility, systems connecting and talking to each other—are they solved at the point of the agent? Where does this, I guess, magical formula—
That's a great question. So I guess some problems are solved with agentic, and some is what is the hardest part that we solve. So the part that you rightly said, that healthcare, banking, they're not on the cloud yet as we know fully. So much of this needs a flexibility to operate wherever customers are and integrate with systems that they use. So one of the unique offerings we have is to be able to orchestrate across thousands of applications. In an NHS scenario, there are hundreds of applications that we have to integrate and orchestrate a work across it. So that's a hard problem to solve that we solve.
The second piece is, is data clean? Some of it has been solved with agentic because you could, for example, we have an unstructured data solution that is powered by generative AI that could take a large amount of unstructured data and make sense out of it relatively quickly. Now that piece of solution, because it has gone through millions of these, it keeps getting better every time. So today it has seen more than any human being has ever seen. It has removed the need to rationalize all of the data. You can throw all the unstructured data at it, and more often than not you will come up with the right answer, and you save all the time to clean up the data. So that has gotten easier after three years of processing it.
Governance and security is not solved, as we have seen in various news articles, so that's something we bring to the table.
I'll share one of the announcements in context of this. One of the announcements we made earlier this year, we announced the capability of a next level of autonomy that enterprises could get with security and governance, and we partnered in this scenario with Nvidia, OpenAI, Cisco, and a few other players, and we brought multiple capabilities together to create a full governance model, but a level of autonomy that is unseen before.
With that, we showcased—I'll take an example. Let's take an example of an outage in an IT system. Sometimes some of the issues take a month to figure out what caused it. Now, an agent, governed and secure, can go through your system and within an hour figure out what the problem was, suggest you a solution, and if you let it, it will solve it as well. One month that took you to figure that out can happen in an hour. Like that, in fraud, in claims, in supply chain, there are specific areas where the secure, governed agents can achieve unimaginable outcomes that we never thought possible.
And in that scenario, are we talking about a collaboration with like a NeMo Guardrail? Help me understand.
NeMo Guardrails and Nemotron models are useful everywhere, but especially in international markets where people are going for more open-source models, that's a key partner there. OpenAI in that scenario was our frontier partner, although we work with all other models, but in this particular example, we closely partnered with them to show enterprises what is possible with some of the latest models, as you know now. Cisco brings Defense Orchestrator.
Okta brings identity. So multiple partners brought multiple pieces to it, and we brought our agentic automation execution and governance framework along with it to deliver an end-to-end solution to the customer.
I want to go back to something you said around trust and security, because I think it's a very important part of the enterprise conversation. In one marketing spiel, it's used as an advantage, and in one marketing spiel, it's used as a threat, right? So security is kind of playing this interesting role right now. Some folks would argue that certain systems of record didn't connect for good reason, right? Again, because you needed to have silos in place in certain institutions to ensure that there was no dilution of data, right? And no breach, I guess.
