Vinny LaRocca: Thanks for joining, everyone. We’re going to be talking about agentic orchestration today. If you have any questions, just make sure to put them in the chat, and we’ll get to those at the end of the presentation. I have with me Logan Emerson and Jacob Ortega. Jacob is one of our senior engineers with Lydonia, and Logan Emerson is in charge of AI with UiPath. So, we’re going to start out. I’m going to pass it over to Logan, and we’ll go ahead and take it from here.
Logan Emerson: Sounds good. Thanks very much, Vinny. So, we’re going to be talking about agentic orchestration today. This is our newest product, and it’s actually available in public preview now. We’re going to show a QR code on how to get that into an Automation Pro trial at the end of this deck. We’ll see Agent Builder demonstrated by Jacob, and then I’ll show Maestro, the agentic orchestration tool that we have just come out with. Now, these are both preview products, but they will be released in the upcoming April release.
For a slight bit of review for anybody that’s new to the concepts we’re talking about here, we’re familiar with robots on the right side here. Robots are rule-based. We trust them with business-critical operations, giving them credential stores to do work and get stuff automated. We’ve been trusting them for about a decade now to do this sort of thing. Agents are new to the automation mix, and there’s a lot of hype around agents. There’s maybe some smoke around agents, but not a lot of fire. UiPath is here to really generate fire in a positive way—not in a negative way with fire. So, agents are more goals-based. You give them a role, you help them understand what they ought to be doing, and to contrast agents and robots, think of all the maintenance that might need to go into coding a robot to do a complex task.
So, “Hey, I need you to think about doing this. If this is the case, then do this. If that and this other thing are the case, then do this.” Developing that can take a while. Although Lydonia is excellent at developing automations with robots, maintaining it is also difficult. If something breaks or there’s new business context, a new operation comes into the mix that affects how that work should be done—whether automated or manual coding—that can be difficult, and maintaining all that code is difficult as well.
Agents, on the other hand, you can give them a little bit more of a fuzzy idea of what they ought to be doing. Give them your business’s context, give them robots to actually do the work, and then the agents will think and say, “How do I need to execute this work to get it done with the automations I have available to me?” The human-in-the-loop mechanism: if a human needs to be brought in, agents help connect these dots. And we’re talking about enterprise agents—trustworthy agents that get the job done basically the same way they would be done on rails, with a lot less effort.
Jacob’s going to show us a really good demonstration of how one type of agent in the IT space is made. So, two numbers I want you to think about as we go to the next slide: 72 and 90. And I think they’re both small percentages, or they’re really underrated. So, 72% of organizations found that their real-world, mission-critical processes were becoming more complex to maintain.
I think every process is becoming more complex to maintain as you buy more software to get these things done. You have to figure out, “Okay, we just bought new software. We just bought this other automation platform. Or we just use IDP in this other new way.” New things are coming into the mix all the time. We’re trying to optimize all the time, but it doesn’t necessarily mean that we’re simplifying these processes. It’s complex to maintain these. So, we need a kind of 10,000-foot view of, “Hey, where are my automations? Where’s my software? Where are my systems? Where are my people in this process? And where might these new AI agents that I’m eager to get started with enter the mix?”
Now, 90%—90% of IT executives in the U.S. say they have processes that would be improved by agentic AI. I think ITSM is involved with every IT program, and I would say 100% of IT executives should be saying that agentic AI could be helping them out. ITSM is absolutely a critical process and one that is very much benefited by having agentic AI in the mix. We at UiPath use agents in our ITSM project or product, and our customers are using it for that as well.
So, three questions that often come up in my conversations with executives from the C-suite down to individual contributors at our customers at UiPath and our partners at Lydonia. A lot of businesses have vague goals on how to take better advantage of artificial intelligence in 2025. They know they need to do it, but it’s vague on how they’re actually going to get that done. So, I’m hearing from a lot of IT teams and automation programs that they’re kind of nervous to start adopting agents because they foresee running into challenges integrating them seamlessly into their processes that already exist.
Also, a lot of organizations that I speak with have spent time automating individual tasks, but now they’re looking to take that next step of connecting a lot of task automations into one cohesive, maybe end-to-end process. And then the big question: How do I maximize return on investment? I’m buying AI, I’m getting AI into my mix, but how do I make sure that I’m actually getting a return on that software investment so I can hopefully make more investments and improve more processes? How am I going to do that? Well, Maestro is a big part of the answer to that.
So, Maestro is our product for agentic orchestration. What you’re seeing here is a way to adopt agents effortlessly and with control. How you can transform automation from these scattered tasks to seamless workflows, and you can simplify processes and the way that they’re managed. Now, this is just a screenshot. We’re going to be seeing the product live in a few minutes, so I don’t want you to focus too much on the details on the slide here. But as an example, this is a full workflow that includes an SAP automation, a human-in-the-loop, some decision points that might include business logic. There’s context that’s important to understand whether this flows through this lane or through this lane.
You have integration service connectors. You have agents in the mix. It’s complex. There’s a lot of different moving pieces here. However, you can see the whole thing top down here, and you can click in and kind of tear into this process at each of these nodes. Now, you may notice also, “What is this? I haven’t seen this in the UiPath platform before.” This is called BPMN. It’s business process model and notation, and it’s how we’ve had a lot of requests from customers to, instead of having manual logs, understand, “Hey, this process is going to be working like a workflow. Can I just look at it like a workflow?”
The re-framework did that, but that was pretty niche and for our most advanced developers. Now, anybody can start building processes and skeletons of processes to understand, “How am I going to get this work done?” So, at the bottom here, we have the ingredients for a process. We have agents, automations or robots, manual tasks completed by humans, and then your systems—your software systems, your ERPs, your CRMs. All of these together make up automated processes and partially manual processes.
What we do with Maestro is model, implement, operate, monitor, and optimize. And it goes in a circle. So, model—you’re going to design that process in BPM. Then you’re going to implement that process, integrating the humans, the robots, and the agents. Also, third-party agents. You don’t need to use UiPath agents. If you already have OpenAI agents or you have an Agent Force agent, those can come into the mix too once the product is fully released. Then, operate—you can see instances of the work actually being done across that BPMN workflow and say, “Oh, shoot, this particular trace of work is stuck on the human-in-the-loop. Logan’s sick. He didn’t do the human-in-the-loop piece of this, and it’s just sitting there.”
Well, maybe I can forward that to another person, and I’m only able to do that if I’m looking at this kind of top-down and seeing that work is stuck, and it’s waiting on Logan. There’s a bottleneck on Logan, and that’s just one instance. Now, we can use historical data with many instances in the past to understand kind of a heat map. “Hey, where does this tend to bottleneck? Where does this tend to get stuck? Where does this tend to take the most time? Or where does this tend to fail?” We could even see failures. So, monitor there, and then once you’re monitoring, you can then optimize. How are you going to optimize? You’re going to model again. You’re going to implement, you’re going to operate, you’re going to monitor, and you’re going to optimize. So, it’s a cyclical process. And this is kind of that management that becomes a lot easier using our platform
What does this mean for IT leaders and COE leaders? Reduce total cost of ownership. Scale AI with trust. We care very much about security, governance, and trust.
And then you can lead with data. So, really, you’re going to have a lot of data at your fingertips to understand how your processes are managed. How are they going? Are they going well? Do we need to update this? And you have actual data to understand how many hours are spent here. Can I invest the IT time or time in Lydonia to update this process so it actually becomes better? Is it worth our time to update this process? Maybe it’s not. Maybe it’s working good enough, and we can invest in other areas.
This is a high-level of a purchase-to-pay. I will go briefly here. I think I’m a little over time on my slide so far. So, purchase-to-pay—anytime a business needs to buy something, maybe it’s consulting services, maybe it’s software. We need to request something to be purchased. Then you create that purchase order. Then you receive those goods from the vendor. That vendor will also send an invoice that needs to be processed, and then you receive goods or services and need to pay for those services, which are outlined on the invoice.
Now there’s a lot of matching that needs to happen. The purchase order needs to match up with the invoice. You want to make sure that you received or were billed for the same things that you actually asked to get. There’s reviews. Is this even a compliant piece of technology? You would refuse a purchase requisition were it not. And all of this, what I want to really get clear here is this is a complex process. It’s a huge process that every business has to do when they buy things, and robots are involved, people are often involved, and now agents can be involved where people used to be involved, and agents can direct your robots and even escalate to humans when needed, simplifying and making this process a lot easier to understand what’s happening in each instance of us needing to purchase something.
So with that, I’m going to let you know that again. This is available in public preview now and it’ll be generally available in our next major release, which is the 25.4; that 4 stands for April, so it’ll be the end of April. And if you want, you can scan this QR code to get an Automation Cloud Pro trial that has access to agentic orchestration. We can also continue the conversation, either through Lydonia, through UiPath, and understand what are your needs? With this we can start scoping out some projects, doing some POCs, that kind of thing for you.
Vinny LaRocca: Thanks everyone for joining. Feel free to reach out to us if you want any more additional information or if you want to discuss use cases. But yeah, thanks again.
