Gabriel Carrejo: Welcome, everybody, and thank you for joining us today. We have yet another power-packed experience, where we’re going to get hands-on with Agentic Automation. We’re joined by two of the stalwarts in the industry when it comes to expertise with Agentic process automation. We’re going to be sharing best practices.
Best practices based not just on theory, but based on tried and true, real-world success.
We’re going to get hands-on. You guys are going to get hands-on. It’s going to be an exciting time. This is our second installment of this. It was oversubscribed, and now we’re bringing it to more folks. So buckle up, everyone. Like I said, we’re going to have a power-packed day, and I can’t wait to get started.
Without further ado, I want to welcome to the stage two of my favorite minds in the Agentic automation space.
Today, we have Vinny LaRocca, Senior Director of Engineering at Lydonia, and my friend and colleague, Akshay Mehtani, here from Automation Anywhere. Vinny, Akshay, welcome to the show. Thanks for joining, guys.
Vinny LaRocca: Thanks, Gabe.
Akshay Mehtani: Thanks, Gabe. Hey, everyone.
Gabriel Carrejo: We’re excited. Actually, this is your second go-around. Vinny, this is your first rodeo here, but actually, really quick, give the audience just a quick primer of what was great about the last one. What were some of the things that were uncovered that you think this audience, this version, folks are going to get excited about?
Akshay Mehtani: I think what excites me the most about these sessions for the audience is, you know, not only do they get to understand how Lydonia and Vinny are using Agentic AI in action, how they are actually deploying it at customers, which is one part of the session, but everyone also gets to be hands-on with the product.
So as we do the second part of the session, getting hands-on, I think, is a really exciting part, because based on my conversations with so many customers, not everyone is able to get hands-on and actually access the Agentic platform that we have. So these two things combined, how it’s used with customers, but also being able to get your hands on the platform and use it, that, I think, is the great combination that comes to play over here.
Gabriel Carrejo: Yeah, there’s nothing better than rolling up your sleeves and getting into it, not only seeing it in action, but building it and putting it to action, right?
Akshay Mehtani: Correct.
Gabriel Carrejo: Guys, I want to do a quick refresher, just for everyone’s knowledge, mostly mine. Let’s just talk quickly: what’s the difference between generative AI, an AI agent, and Agentic Automation? Just at a high level, so we can level set what we’re talking about here when we talk about the APA system, and what they’re going to get their hands dirty with.
Akshay Mehtani: Sure, I can give it a shot, and then Vinny, I want your point of view as well, because you’re taking it out to the field as well. Think of generative AI, this is for the entire audience, as this large umbrella which has all these LLMs that give you text and responses to certain questions and things like that. But that’s all generative AI can do. How can I add another layer to say, “Hey, I want to be able to add my own enterprise data. I want to make this a little more customized to what I feel and what I need, to give me recommendations based on my real needs and requirements”? That’s where AI agents come into play.
So you see, we went from a very holistic, large view of the world into a small view of my organization, not even my organization, but my team within my organization. That’s what AI agents do. AI agents can help me be useful at my job within my organization, because I’m giving it more context. I’m taking that LLM, adding some context of my data, and helping it answer more specific questions that I would care about, that make a lot of sense to me.
And then if I add another layer on top of that, if I take my entire organization’s process, and, you know, think about if I’m in the banking world, a mortgage process, or if I’m in the healthcare world, processing explanation of benefit documents, think about those processes. If I take that entire process context and say, I want to orchestrate all of this using a combination of RPA, which is great at deterministic stuff and interacting with certain systems, doing if-then-else logic; AI agents, which are great at analyzing data and giving me information back based on what I care about; and adding humans into it, that is where Agentic Automation, or APA, comes into play.
And then I can take all these things together, put them in the context of an entire process, and get my efficiencies out there. That is what APA would do. So you take your larger picture, bring it down, and then take it into a larger process. Essentially, that’s how you want to think about it.
Vinny LaRocca: Yeah, I think I probably wouldn’t have too much to add to that. I think that’s well put. From our perspective, I would add that there’s really nothing we see at this point that can’t be done faster and better with agents when it comes to automation. Traditional RPA is still being used here and there, but we’re seeing less and less of it as time goes on, and as the Agentic capabilities get better and better. We’re starting to use more and more of it to really handle end-to-end processes without having to do descriptive or prescriptive coding.
Gabriel Carrejo: Yeah, it really is amazing, even in my own world from a marketing perspective. Generally, when you break down the idea of work, forgets even process, just work, our jobs, my job, your job, Vinny, Akshay… there are core tenets to how we approach it, right?
Previously, with deterministic or even some of the generative AI, as Akshay mentioned, there were personal productivity tools that could do a little bit, and that’s fine, that’s well and good. There are some siloed AI agents for specific platforms and whatnot. But in the idea and context of work, I’m working in multiple systems at once to get something done, to do my job.
To me, that’s where the Agentic Process Automation piece really shines. You’re not bound within the confines of one system, or limited to, as Akshay said, just a text output. Text output is good in many cases, I need a framework for a slide, for example, and that’s fine. But if I really think about why I’m building that slide, in what context, and from a business outcome perspective, then that becomes an entire end-to-end process. That involves other people from other functions and other systems. That’s what we’re talking about, I think.
Vinny LaRocca: Yeah, that’s a big difference between the GenAI capabilities when they first came out and everybody was worried about hallucinations. You had these generalist models that could pull something out of a document for you, but could also write you a Shakespearean poem. And it’s like, well, I don’t really need it to do that for my processes. I need a more specific, tailored tool to build end-to-end process automation, not some generalist model. And I think that’s a big piece of what Agentic brings to the table: grounding it in your information, your process, your documentation, your SOPs, and being able to guide it in a way that eliminates the fear of hallucination that we used to have to deal with.
Gabriel Carrejo: Yeah, it’s, we used to, we used to, at some point, we had a marketing campaign. I think it was mostly internal, but it was this idea of the art of the possible, and I was like, well, now it’s the art of the impossible, or at least the previously impossible. As you mentioned, Vinny, like, there’s pretty much nothing that can’t be automated or made better with the Agentic Process Automation System and this new approach, right? Which is super exciting. Before I turn it over to you, Vinny, as you’ll be kicking us off today, I wanted, you know, I was doing just a free fit of research. I want to say that you’re basically a rocket scientist, based on your previous experiences. Do I have that right?
Vinny LaRocca: I was originally, majored in mechanical engineering, and I was a propulsion systems engineer.
Gabriel Carrejo: There you go.
Vinny LaRocca: More into computer science, yeah. So I went from that to robotics. We’re actually, about a decade ago, was really focused on applying AI to robotic systems and manufacturers, and you can imagine, like, pretty rudimentary stuff at that point, like C-sharp, you know, very new, I guess I’ll say. And yeah, that’s what I was doing prior to coming over to Lydonia and kind of transitioning more into the pure tech space.
Gabriel Carrejo: Awesome, awesome. Awesome, well, we’re glad to have you, and we’re glad you made the transition. So, without further ado, Vinny, I’ll let you take over. Take it away, Vinny.
Vinny LaRocca: Yeah, you know, I’m the Director of Engineering for Lydonia Technologies. We are really a business transformation firm purely focused in AI automation. And where we differ a little bit is we tend to be hyper-focused on outcomes. And what I mean by that is really making sure that we’re going after the right types of processes, we’re deploying things that make sense, and quantifying it before we actually go and do it. And one of the core pieces to the way that we do this is our Agentic Accelerator program. And, you know, obviously we’re going to talk today a lot about the actual technology, which is a big piece of this, is picking the right technology partner to move forward with for success. But one of the things that we found is there’s a lot of other areas that you really need to nail to be successful here. And these are really the three pillars here that you’re looking at.
Identify: find high-value use cases with defined ROIs and level of, and creating a backlog so that we can, you know, programmatically sort of attack the highest to lowest ROIs.
Enable: letting the workforce know what’s available to them, what tools have been built or bought so that they can utilize it correctly, and even, in some cases, citizen development capabilities so that people out who are, you know, doing these processes every single day can start to build some intelligence against what they’re actually doing.
Manage: doing things like show-back models and governance, and what is the security posture of all of this, just making sure that you have a good program to continue to run this over time.
So the first one that I’ll really dive into is the identify piece of this. And the goal here, again, really is to go across, you know, different lines of business, figure out what everybody’s doing every day, find some high-value use cases that we can define ROIs behind, build that backlog. This is kind of an example on the screen of what you would see from an output perspective. And then again, just make sure that the use cases we’re going after is actually going to drive tangible business value, not just we’re building a shiny object because it seems fun, or somebody hates doing something. Right, so we really want to make sure we’re driving value back to the business.
The next area, here we go, is enable. Right? So, couple different pieces to this one. The first piece is going to be training, right? And that’s going to really be split into almost two different things. The first one being training on any sort of citizen development, as we’ll refer to it, which is giving the line of business no-code tools that they can utilize to start to automate mundane things that they’re doing, just to become more efficient at what they’re doing. The other piece of the training here is really more around the tools that are available to the teams out in the field, making sure that they understand what they can use, what they should be using to just perform their job quicker, faster, better.
And then the second piece of this that we see is really critical here is the community piece. So, especially when it comes to, like, citizen development, for example, we’ve found that setting up a citizen development group and then allowing the citizen developers to take out tickets with the IT team, what that ends up doing is we end up having our IT teams spend all of their time supporting citizen development and not being able to get their level of development done. So they end up with a massive backlog, and then everything stalls out. So this is one of those things that just over the years of doing this, we’ve figured out that actually the best way is to set up an internal community where citizen developers can talk to each other. They meet weekly, you know, they do internal trainings and knowledge transfers, and then that leaves the IT team to go tackle more difficult, higher level of effort type of automations without them getting bogged down. So it’s a sort of, you can have your cake and eat it too here, and we’ve kind of figured out how to structure that correctly.
The last piece is really mostly around setting up a program around all this stuff. Making sure that you have executive report, setting up a real structured center of excellence, show-back models, making sure that we have best practices installed, that we have ways of governance to ensure that those things are being adhered to and are happening correctly. You know, all of those things you really need to have a successful program over time and be able to scale it successfully.
And then, you know, what that kind of looks like from an execution perspective, if you flip to the next slide, is really these three steps. Right? So the first thing is we want to make sure that we’re setting up the program right. We’re scoping everything out of what’s going to happen. We have success criteria in place, we have different stakeholders who are holding different responsibilities, and everybody’s kind of firing in the same direction.
The second piece being execution, going and doing that discovery, finding somewhere usually between 100 to 200 high-value use cases that we can go attack. Putting those in an order that makes sense, working with the teams and the stakeholders to ensure that we are accurately ranking things based on what the value is and what their priorities are. Ensuring that we have things like governance models, and then ensuring that we’re building a program to continue to do that after we engage. So continuing to do intake, continuing to look at our governance model and update it and make sure that we’re being secure and we’re following best practices. And again, just making sure that we are set up to do that 2, 3, 4, 5 years down the road, not just at a point in time. And that’s really where we see this stuff be successful, is when we set it up as a program, not as a project.
And then, I believe there’s one more slide. Yeah, and then I’ve kind of touched on some of these, but these are really your core benefits from doing something like this. The first is going to be speed, right? Even just the use case identification part of this. This whole program runs 8 weeks. And in that 8-week time, we’re going to get in the ballpark of 100 to 200 use cases, we’re going to set up an entire program around this stuff, and we’re going to get everybody who you want involved trained and enabled. And so, really, this is a way to do it quickly and to make sure that you’re scalable in the future.
So, with that, I’m going to turn it over to Akshay. So, Akshay, I’ll turn it over to you.
Akshay Mehtani: Thanks, Vinny. So, oh, let’s go into the product deep dive now that, you know, we’ve seen how we can get this in action, how we can make this look cool, and do all of that that we showed us. So, for the priority dive, let’s take a look at what we have to offer, right? So, I want to start off by presenting this, like, not all AI agents are equal, and you know, in today’s day and age, we see AI agents all over the board. The biggest thing out here is just figuring out what’s best for what needs to be done.
And there are AI agents that are good at personal productivity, so think about tasks like summarization, tasks like writing an email, writing a blog. That’s what I would call personal productivity. And what we’ve noticed based on our research with our teams is that this kind of leads to a 10-15% impact. It’s not a great impact that it leads to.
Then we have app-specific AI agents over there, which are AI agents that sit within a certain enterprise application, think about Salesforce or ServiceNow. And as long as your end-to-end process sits within that application, it’s great, and people love it, it works really well. But there’s another drawback to this, that your entire process needs to be contained within a single application. And being process experts, doing this for the last two decades, we know that no single process sits within a single application, and that’s where Automation Anywhere comes in. That’s where we look at long-running, mission-critical processes. Things like order management, loan processing, client onboarding, and things like that, which go across applications, across systems. They have to interact with legacy systems. That’s where we see the maximum impact. That’s where we want you to focus on, and that’s how I want you to think about AI agents, even after the session. How you think about it is that’s essentially how you want to go about it. So that’s one on the next slide.
Awesome, and so what’s that? What’s the platform that makes all of this possible, right? How can I? I’m saying, hey, I can take you across applications. I can automate your most complex processes. I can interact with legacy systems. And the next most natural question that everybody asks is, how? This is how I can do that, right? We have what we call as the automation system, which is the one that you’ve been using over the last, you know, 5, 10 years, is at Tech and Hull, that’s the most, that’s the system that I’ve been working. And that’s what gets you the deterministic flow. If this, then that, interacting with different applications, interacting with different systems, legacy systems.
To that, we have added a whole security and governance layer on top of that. That sits over there for you to kind of do monitoring, for you to kind of get your ROI, and we’ve, over time, added more than a thousand integrations with different systems. Like the SABs and Sales Forces and ServiceNows of the world. But what we have done over the last 2 to 3 years is created this agentic system. This agentic system has a bunch of capabilities now. You can read documents of, you know, using document automation, and this is not just limited to invoices or purchase orders. This is all across different kinds of documents. You can do this with unstructured documents, structured documents, semi-structured documents, complex invoices, commercial invoices, all of those things. Document automation is your one answer to be able to extract data of all those kinds of documents.
If you want to be able to build automations faster or get a faster go-to-market. We have Copilot for automators, where it’s like GitHub Copilot, but for automations. You can easily just, in actual language, come in and say, hey, I want to build a bot. That picks up data from this Excel file, puts it into SAP. And there we go, boom, you have an automation that does that for you.
We have Generative Recorder that helps you automatically fix different automations in case a UI changes. That assists you through the process to automatically fix it up without you having to go in and write additional code. A few wise change. We then have this ability to help you build out custom process agents. With AI Agent Studio, you can connect any large language model, or small language model, or fine-tuned model into the platform. And interact with that in the context of a process. And with enterprise knowledge. You can bring in your own enterprise data also into the platform.
All of these things on our agentic system are powered by what we call as the process reasoning engine. The process reasoning engine is our secret sauce that powers all these different capabilities and functionalities that we have to be able to bring automation and AI to you within your business context. This is what helps take all of your information from an enterprise level. Accept a goal, and create a goal-based region where you can give it a goal. Based on the information it has, it creates a plan for you. And then helps you to execute that plan and reconsider, or rework that plan as needed.
All of this is built with human-agent collaboration in the platform. You can interact with all these agents, all these capabilities I just talked about. At any time, you can just say, hey, I want a human to interact with this, and it’ll look at a human infraction screen right then and there. So, that is the agentic process automation system. That helps you automate these long-running, mission-critical processes we saw on the previous slide. Like, order management, loan processing, client onboarding, so on and so forth. So, that’s the how.
And this is a process reasoning engine that I was just talking about, right? At the core of this process reasoning engine that essentially works, that essentially helps with the agentic system. At the core of this is our process models. So, process models, UI models, this is our two decades’ worth of information that we’ve packaged up together. And given it to you, so that you can kind of make sense of it, and you can use it to your best uses and best application.
We then have customer process contacts. This is your information, which is locked in. Kept within your environment. But just used to help make it better, help make it better for you in your context. We had memory and learning. So now, if I give my process reasoning engine a goal. It will then create a plan. Help you execute the plan, and then reflect the plan. All of this, with having humans in the loop. With human-agent collaboration happening in real time.
And these tools that you’re talking about are not just any tools. These are AI agents, these are AI skills, these are RAG knowledge bases. But you can all integrate into this platform using the process reasoning engine. Let’s go on to the next slide. And I’m not saying this, you know, just something that we’ve developed and we’re kind of, you know, playing around with. We’ve actually deployed this at customers in real time. We have done this at KeyBank for in their operations space. We have done this at Boston Children’s and for Prior Arts. We’ve done this at Petrobras, you know. Helping them automate taxes. You’ve done this at Ximena with invoices. Jaguar Land Rover for dealer opportunities. And at Alight to help them go with claims.
So, all of this are agentic opportunities that we’ve developed over the last 2 to 3 years, and it’s not just limited to these 6. We have more than 1,500 use cases that have been deployed across 6 major industries. And are unlocking transformative business outcomes at scale for all of our customers. Right? And that is the end of our product deep dive.
Gabriel Carrejo: But to the audience, I hope that you know, you’ve, you’ve seen not only what you could do with this particular example. I hope it really inspires you to think through what other processes you can attack. As we said at the top, this idea of the art of the possibles. More like the art of the previously impossible, is now possible. So another job well done, Akshay. Again, looking at the comments here. Looks like everyone shares my sentiment.
To those that participated today, thank you for joining us. I encourage you to connect with the team. With Vinny, with Akshay, with myself. On LinkedIn, feel free to ping us at any point. We’re always here to help. I encourage you guys to, if you haven’t already, join and participate in the Pathfinder community. And again, reach out to, reach out to us directly. This is exciting times, and Akshay, like I said, not to be a homer, but being a part of this, and watching folks like yourself, and folks that are participating, be able to do this, and really define this generative age, or the agentic age, is truly, truly something special, so thanks to everybody for joining.
But thank you, and thanks to the team at Lydonia for partnering with us. You know, at the top there, just to reiterate, Vinny. Vinny and the team at Lydonia are doing amazing things to help you programmatize. In the Agentic era, to get speed, speed to delivery, to scale. All of those things, so you’re not stuck in POC gel, or pilots, or whatever. You’re getting to real solutions. Real fast, with real impact and that’s what we want to help you do, and that’s what’s possible, so hopefully you all saw that.
We’ll be in touch soon.
Akshay Mehtani: Awesome. Thank you, everyone.
Gabriel Carrejo: Thanks, Akshay, and thanks, everybody, for joining. See you soon. Thanks for joining.
