Most Government AI Agents Never Reach Production | Scott Ditch

Speaker A: Consider what the phone has, uh, been able to do over the last few years, you know, and the changes it's done in everybody's life. What would happen if AI was more on the edge here on our device? Would that cause apps to start disappearing, you know, where you're now able to go straight to the phone and go, okay, well, I need whatever. And then that agentic workflow kicks off. Uh, we're now in a world where it's no click applications.

Speaker B: Hi, Scott. Welcome to Drag and Drop.

Speaker A: Hey, Faraz. Pleasure to be here.

Speaker B: It's great to have you. I've known of you for a while. I've been trying to kind of, you know, find the right time, and I think now's a good time to speak. With the space evolving, you know, rapidly. So, you know, it's great to have you on and speak to you. It'd be great to kind of start with a, uh, you know, I think within the Outsystems ecosystem, you're pretty well known. But, you know, for those who are not in the kind of community, it'd be good to get a bit of an, uh, introduction to yourself and your background.

Speaker A: Yeah. So, yeah, my name is Scott Dige. Been with outsystems for over five years now. Been working in the tech industry since, what, 2003. You know, essentially worked almost every single type of role from support engineering, professional services, had my own company, uh, things like that. Now I'm principal enterprise architect for Outsystems. And, uh, you know, I get to focus on those, that big picture. How do things come together, how do they connect? Then I get to draw that pretty picture, which is always exciting, to make sure we have that common language when we're talking about, you know, how do things fit together, put together those frameworks. So that way we have rules that govern how these systems come together, and when there's a need for a solution, how do we pick the right thing for it? And then also I do, uh, demonstrations and stuff of the Outsystems software. So really exciting time, you know, with, with, uh, AI to, to be in this industry.

Speaker B: Yeah, definitely. And, you know, I've been following kind of the things you've been doing, and I noticed you've been, you know, doing a lot of things kind of on, you know, AI in the US public sector. You know, so I thought today's topic and theme would be great to kind of focus on that, you know, and kind of hear your views on, you know, government agencies have been, you know, experimenting with AI now for a few Years, you know, but it does feel like, you know, like you said, you know, we're entering a real, very different phase of adoption. You know, so as we kind of move into this next phase, what are government leaders actually kind of struggling with today that you're hearing?

Speaker A: Yeah, it's the trust is the key thing that I keep running up against, you know, right at the forefront of these discussions. And so with that, it's like, okay, well, how do we tackle that, that trust? And a lot of it's bringing in a framework and being able to execute on that and then knowing what to use and when to use it.

Speaker B: Yeah, yeah. And, you know, many, many of these agencies, obviously they have dozens of pilots, you know, going on, you know, what's separating, you know, the organizations creating these measurable value from creating even more complexity.

Speaker A: Yeah, it's, it's going from that vision that you have in your head to execution. How do you, how do you get that vision delivered? You know, and then once it's built, then what? You know, you have a lot of AI pilots that are, you could use something like Claude to build an application, but then it's on the local host, uh, you know, and then how do you get it from. From there to where it's all the way delivered to the people that need it, you know, and it's deeper than that. You know, with AI, there's the agentic development life cycle. We're used to doing the classic SDLC where you build something and uh, you release it. Then you hope that you don't have any bugs or defects that you gotta go back and repair. But the agentic development lifecycle's much, well, uh, not much different, but it has that additional layer of tuning that happens after the fact. And so you have to be able to look at what you've created and then know, okay, well, what do we have to do to make sure that those edge cases, uh, don't cause problems? And uh, you know, you don't want to be able to have that agent to be able to go and just do Google searching or even worse, uh, doing what would be like an inject on the model and stuff and be able to pull up things that are completely out of left field you want to have control over.

Speaker B: Yeah. And with everyone talking about AI agents, do you think the bigger challenges is actually orchestrating people, systems, data, Ah, and agents into a single operating model?

Speaker A: And folks that know me well, that I've been working with for a number of years in the public sector know that I have kind of A mantra with this. And it's that software's about the people. You know, they're the ones that consume it. In the end that, that ha. And it has to work for them. And in the public sector, that's who we're serving. You know, it's about the people. And so there's, there's concerns around adoption and whether or not this AI is going to eliminate, you know, the individual's job, is it going to cause more work? What's that like? And so it's making sure that, uh, there's, there's good education, there's good training to those, those individuals that it affects. And further, uh, people don't want to talk to robots. You know, we're getting into an era where you make a phone call and you end up talking, talking to agents that, that, that are essentially our bots. And that's, it's not a bad thing. It's, it's saving time, it's saving money, but it's allowing the people that have to handle these problems to focus on the part that requires the people.

Speaker B: Yeah.

Speaker A: And if I'm a citizen and I, you know, need something done and I, and you know, I'm working with agents and stuff, we need to make sure that there's still a human side of it. You know, there are companies now that are like, instead of using agents and bots, we're proud to have people. And so it's, it's that focus on the human aspect. In the end, that that's who we're really doing these things for.

Speaker B: Yeah. And yeah, I think, personally speaking, like, I'll always prefer a human. Right. Because I can ask any question and get an answer. Where even, like if you say, submit an inquiry online, it's so difficult to get to the point where you can actually speak to a human being.

Speaker A: Exactly. You know, you end up sitting there just mashing one on your phone, hoping that you get to that individual so you can go, hey, I need my cruise rescheduled. Uh, you know, or hey, what's the status of my building permit? Things like that, you know, and uh, it's uh, it's always good, you know, nice to talk to that individual, know, like that, that that person has, uh, has needs and wants to help me out. Uh, you know, computers don't really have those feelings, uh, to, to go, oh, you know, I, I have empathy for this individual. I can tell by their voice that, uh, that this is a stressful situation. I need to listen a little bit. Computers are like, okay, we gotta solve this problem and move on. Yeah, and there's not, it's not like people don't do that too, but.

Speaker B: Exactly.

Speaker A: You get what I mean.

Speaker B: Yeah. Of course, you know, citizens still experience, you know, government as a collective, you know, a collection of separate agencies other than one connected system. So how do you see agentic systems helping bridge those gaps?

Speaker A: Yeah, you know, in some situations it's able to do that, but they're all there. It kind of always has to go back to the money. There has to be a monetary reason to connect these things or it may not necessarily make sense. The problem with using uh, some of these agentic systems and the AI behind it is the tokens and the cost and that seems to be rising. And you know, there are techniques and essentially using cheaper models for the right situation and being able to evaluate those models is kind of key there. But, uh, it has to kind of focus on the money side of things. There has to be a monetary reason to bring these things together. That's what I'm seeing. Because they would be great to use AI to connect these things. But it, it becomes prohibitively expensive in many cases.

Speaker B: Yeah, you know, government decisions will also affect, you know, things like benefits, taxes, health care, permits, and, you know, public safety. You know, how do agencies introduce AI into those processes while ensuring, you know, like, the human remains accountable for those decisions?

Speaker A: Yeah, you know, like I said, you know, that the human is always the big thing. And in this case we need to have that human in the loop and that needs to be part of that, that big vision as to how, how you plan to execute on AI. So you know, enterprise architecture is, you know, my jam, you know, being an enterprise architect. And so it's getting, it's having that as part of the way you solve the problem. You go, okay, well I want to use, uh, an AI model to solve this, but we don't want it making the decision end to end and then not having a human to be able to go, is this the right, the right way to, to handle this problem? You know, agentic AI needs to be a guide to help make decisions quicker and not just the end all, be all to making that decision. You know, uh, that, that, that, that's one of the risks here is with, with AI, you know, if it's, if it's the connective tissue, you still maybe want a person in there. And that's part of that, that vision here. You know, it, you shouldn't have AI just executing blindly. One of the challenges with doing that is and using AI overall Is what did it do? How do we know that what it's doing is genuinely what we want? And so having the ability to audit, but then also having that human to be able to make that final decision and helping them to make that decision quicker. You know, we don't want them to be kind of just automatically going yes, yes, yes, yes. You know, back a number of years back I worked in the data extraction and OCR space and one of the steps when extracting data is validation and you get to a point where the operator that has to validate what was extracted, sometimes it gets to where the confidence levels don't tell them like hey, you know, this is unconfident. They just are pressing enter and just accepting whatever the, the, the uh, was extracted. And we don't want that in an agentix space either. Where they're looking, they're not even really uh, reviewing. What we want is these AI models to be able to pull these things up and for the operator to be able to make the right decision based on the information provided. So they can go, okay, well this document needs to match this type and we've pulled out this date and it matches the date or you know, things like that. Uh, so that way they can visually see and the AI is assisting.

Speaker B: And you know, one concern I hear from you know, a lot of business leaders is that, you know, they may be creating the next generations of silos with disconnected AI projects. So how should agencies think about governing an entire portfolio of AI, uh, systems rather than just individual use cases?

Speaker A: Yeah, you kind of have to think big to kick butt here. The initiatives that come down drive the strategy and then you have to build that strategy into your enterprise framework. So it's about having that vision and now you can make that right selection of that right model and then be able to connect these things. You know, it's not as simple as I'll just uh, put, put some AI on top of it and uh, just uh, hope for the best. You know, you really have to consider that token cost and then the uh, the trade off of going broad versus narrow. Do you want to have tool usage? You know, not all AI models can do the same thing. You know, some of them are better at others. And so that's part of having that enterprise framework, having connective tissue that is able to, to bring those, those disconnected projects into an overall vision that then can be executed on in a repeated cycle, you know, cause that's the goal here is to have a repeated playbook or that enterprise framework that uh, that can Be used for many years going into the future and evolved.

Speaker B: Yeah. And many agen still, you know, operating mission critical systems that are decades old.

Speaker A: Right.

Speaker B: You know, so do agentic systems change how we think about modernization?

Speaker A: Yeah, you know, in some of the best cases and best situations I'm seeing AI able to rewrite and modernize. You know, we run into a whole lot of legacy systems in the public sector. You know, I've recently uh, ran into a customer that had a payment system that was uh, 30 years old, running on some pretty old technology. They were paying a lot of money to still have this uh, supported. And that, that doesn't serve anybody's purpose. If you consider the systems behind the scenes there. They were very fragile, you know, but we had AI examine that application and go, oh, okay, well this is what you would need to rebuild that in a modern platform and then bring that into our framework. And that was able to get that application and that greater system rebuilt and bring it into that modern era and start making it more future proof. So that way it can last another 30 years. But then there's the other side of it, uh, where we're lucky, where people have potentially worked for these agencies for a number of years. They've been taken care of really well, which is always an excellent thing where people can retire after many years of working for, for that same, for that same organization. But there becomes institutional knowledge that's lost. And so AI is fantastic at writing those documents. And so that way knowledge transfer can be hap, can, can, can, can be executed on and all that great knowledge that they've accumulated over say, you know, 20, 30 years of working there isn't totally lost and we don't have to pull them out of retirement and get them on the phone, hey, how does that as 400, uh, integration work, you know, huh? We have those things now documented and we have now a path forward and they can go happily into retirement.

Speaker B: Yeah. And then touching on that point about, you know, people being working for so many years, do you, sometimes there's two sides to it. Do you think? Sometimes it's okay, they've got that knowledge but they're reluctant to change and sometimes it is needed to bring in someone fresh and disruptive to kind of move forward on that side.

Speaker A: Yeah, you know, that's, that's always a key consideration. You know, you want to have that fresh talent. You know, like I said, it's about the people. And so not only do you want to retain those people who have been around for a number of years, you Want to have fresh individuals who think a little different than that classic mindset. You know, we don't want to be solving things with programming languages that are ancient. We want to be thinking in a way that is, that is future focused, particularly in the government or uh, the public sector, where these systems do last a long time as well as sometimes the procurement cycle takes a year. And by the time we get done with the procurement, things could have potentially changed. And we don't want to get stuck in a situation we were where we were focused on the hottest thing.

Speaker B: Yeah.

Speaker A: And it's now gone and dead.

Speaker B: Yeah. Especially with AI moving so fast. You know, something that was popular. Now in 12 months time, there's a new flavor out.

Speaker A: Yeah, exactly. You know, and we're seeing that even on, uh, kind of a micro level where these models come out and then there's another version here in a few months as they're competing against each other. And that model might be more expensive and you don't want to release it only to have to claw it back. So that's why having frameworks that look at these things is so critical. And being able to evaluate as a team and have those, uh, and if you're lucky to have that enterprise architecture team for you to go, okay, well, we have a vision. We're able to have these three to maybe half a dozen models that are able to be part of our bigger vision. And then part of that framework is how do we reevaluate these models? How do we have that constant cycle of, okay, does this make sense for us? Is there something new that we should be evaluating? What does it mean for our future and how do we manage the risk around that?

Speaker B: So do you really, do. Do you believe that, you know, I could really become that bridge between legacy and modern citizen experience without requiring these massive rip and replace projects?

Speaker A: Yeah, you know, uh, the rip and replace is all is, you know, rip and replace versus doing a, uh, lift and shift. You know, is, is always what needs to be considered. You know, sometimes you run up against systems that, that are very old. Luckily we're able to layer on things like web services and then communicate with these AI agents. And this is how we're able to take some of these older systems into a modern era in a way that is a bit more of a lift and shift as opposed to an all in one rip and replace. And so, you know, in that context of that as 400 integrated with like a DB2, you know, maybe it's time to take that as 400 layer off of it, uh, integrate with the DB2 side of things for a short amount of time and then as resources, our individuals that work for us free up and we're able to make that transition. AI is there already there and the people that are working with that top layer are already used to working with the AI and asking questions of it or part of that agentic workflow ideally, you know, because that, that, that's where we want them, is in that agent. That's where we want our end users focused is in that agentic workflow as opposed to chatbots and AI running on their local machine. We want them to have the, that task that they have and then steps to execute all the way to the end cleanly and fast and with high accuracy.

Speaker B: Yeah, yeah. And you know, in your role, you know you're speaking to many enterprises and their team, you know, and um, software development itself is changing so fast, you know, as you speak to these enterprises, you know, what are you seeing happen to the role of developers, architects and you know, technology teams as AI becomes even more capable?

Speaker A: Yeah. You know, for developers it's considering like uh, the tools that you have around, you know, things like Claude Curo, Cursor Codex, that and what they've done. You know, I can now speak just plain English into these tools and outcomes. Excellent. Ah, well for the most part decent code. You know, there's still potentially a lot of bulk and additional stuff that's put in there and uh, but being able to just even start with that is, is really cool. You know, when I started a number of years back, you know, I learned Java, you know, professionally in college and back then it was Java 1.4. I don't even think there were generics back then. Yeah, or they did exist but that was in the uh, C Sharp NET world and then Java stole that. But uh, you know, languages and the programming languages are layers on top of layers. So developers are to having to navigate that, those, those layers to be able to solve problems. And so approaching it where AI is an additional layer to this world is the, is the same way we've been doing it for a number of years. And so it's not just as simple as ah, uh, we just put AI on top of it and we're done. There has to be an analysis of what comes below that for critical things, you know, for some of the easier stuff potentially we may not necessarily need to dig that deep into it, but uh, that's the way it's been. Navigating that stack is key and knowing that it is an additional layer and for architects, those are my folks, it's being able to focus on how things fit together and using AI, uh, to be able to have better advice, research on those unknown systems that you might get from just requests. How do we put these things together, knowing where those integration points are and the nuances to that integration. For example, is it going to be best to use that uh, API or maybe an SDK? You know, with SDKs there are certain power that you get out of that you can't necessarily get with an API. But APIs are uh, uh, abundant, uh, and usually fairly easy to integrate with. Many systems where SDKs require a bit more traditional development and are, you know, a little bit more difficult. But you might get that extra horsepower, that extra speed, uh, functionality that you normally don't get, that you couldn't get with a REST web service unless you build stuff on your side. So it's knowing those trade offs, AI is able to help us architects go, okay, well you know, where are those edge cases? What might we not necessarily be aware of? Uh, and where do we go to read about these things? What sources do we need to be aware of? That's the way I see fellow architects and even myself using AI. And then teams need to consider AI assisting their Scrum and then their task selection. So that way things are now assembled in a logical way as opposed to the old fashioned way where, well, not old fashioned but where, uh, during that Scrum meeting we go pull a sticky note off of the whiteboard and we're going, this is the one thing that I'm working on today because it's the easiest thing, but because uh, now we have AI to go, this is what's logical to come to be built and worked on today over the next period here and that will logically fit together as opposed to, you know, we've all been on teams where there's that one team that everybody's waiting on, uh, to build their piece. So that way all the dominoes can fall in a line, but the AI's assisting to help make that decision for the Scrum Master and go, okay, actually instead of picking the easiest thing off the board, here's really what you should be working on as a team. You know, let's get that integrated integration to that UI UX layer as opposed to having our UI UX team build that. But then it has no functionality behind the scenes or you know, and we're waiting for that, that rule layer to get to our integration, you know, and we're blocked we don't want to get into that state. And so AI is helping those teams and that scrum master make those decisions better and guide them to have something that then can be rolled out in a way that makes logical sense and can satisfy the, uh, development cycles.

Speaker B: Yeah, yeah. And, you know, if AI can generate, you know, applications, workflows, integrations, and even agents. Right. You know, what becomes the most valuable skill inside these technology organizations?

Speaker A: It's, uh, being able to execute on that vision is key, is king. Uh, from my perspective, you know, sorry, execs, uh, that are listening to this. Yeah, definitely being the big picture person and being future looking is important. And being able to go, okay, well, this is where our company needs to go as a direction. But then being able as the people that have to execute on that, uh, is difficult. And having that, that ability to go, okay, well, the people that, that pay me have a, have a vision. How do I get there? How do I go, okay, well, they want to put AI on this, but what, what do I need to do that? You know, and, and it's having the ability to go, okay, let's ha, let's bring this into our enterprise architecture framework. Let's be, let's have cycles on that to make that, uh, executable, you know, that way we're not, uh, compounding risk. We're not causing those executives that have that vision to be up at night going, oh my God, I saw on the news that AI was causing this problem. Ah. And, uh, is this something my team is worried about? And the answer is yes, they are. But they've got that framework, they're able to execute, and we're able to go into the future without those executives losing sleep over it. Uh, and like I said, execution is king.

Speaker B: Yeah, yeah. And talking about losing sleep, you know, these government leaders are, uh, under increasing pressure to demonstrate measurable outcomes from AI investments. Right. So where are you seeing agencies generate the clearest return today? And where do you think expectations are? Still ahead of reality?

Speaker A: Yeah, you know, it's doing more with the teams that you have. You know, what we don't really want AI to do is just cause people to lose jobs and things like that. You want the people that you have to be able to make right decisions. Ah. And to be able to do more. You know, I come across these teams that are, that consists of, you know, a handful of, of developers as an example. What we want to do is use AI to make them feel like a larger team and for them to be able to execute on that Backlog that's been there for a number of years or the things that they just could never do because they're just not a big enough team. And for the teams that are much larger, uh, uh, it's being able to have organization and to be able to make right decisions. But when it comes to measuring and having actual metrics, it's how do we have those individuals execute on more tasks and do it with more accuracy? You know, consider a building permit, you know, for counties and cities and stuff, that usually has quite a bit of labor behind the scenes for the people that have to issue them. There's a lot of documentation that has to be reviewed. A lot of other systems in California, we don't want people building on earthquake faults and in floodplains. So we need to go, okay, well are they building on the San Andreas fault, you know, at this, uh, this particular spot? How do we go and look that up? You know, how do we use that, uh, that data, look at mapping and go, okay, well there's, there's high risk. So now we need to look at their engineering plan. Is their engineering plan able to satisfy a very large earthquake? And so that type of decision making is not easy and it's not exactly fast. But AI can help organize those different pieces of information into a way that now that person issuing the building permit can look at it and quickly go there in this high risk zone. Okay, I need to look for these types of things within their engineering plan and oh look, they've got it all. Okay, so we know that this is high risk. Now we know what types of inspections need to be executed. I think you're getting the idea that, that it's about being able to help make the decisions correctly and in a quicker amount of time.

Speaker B: Yeah, yeah. And you know, with, with technology, you know, evolving, you know, every few months at the moment there's something, you know, there's something new and then it's changing so fast. No, but like you touched on, you know, government procurement and you know, these modernization programs often, you know, operate on a multi year timeline. So how should agencies adapt to that reality?

Speaker A: Yeah, this is something I think about daily. You know, finding future proof solutions is difficult. I can name one I work for, you know, one. But I've been around for 25 years making sure we're future proof. But you know, you have to consider that evolution of development and how computing power doubles every 18 to 24 months as Moore's law dictates. And so that, and that's almost a whole procurement in the government world procurement cycle in that government world. So expecting that there's no need to adapt is naive. We have to be ready to adapt. We have to be future focused and going, okay, well if we go down this path today, what is it going to be in two years when we are actually at full blast? And then how do we evolve from there? And then when selecting a solution we have to consider the past of that solution. You know, how have they been able to evolve? And I'm not saying that things that are new are bad and aren't great. You know, we've got. Because AI is actually is evolving at extremely high rates of speed. Yeah, I'd be interested to see if there's, I should look up to see if there's a uh, new like revision of Moore's Law that applies to AI. I bet there is that somebody's uh, coined. But uh, you know, uh, it's considering that, you know, and also that, that again back to the people that work for you are key. You know, finding that, that right developer and being able to keep them engaged for a long period of time is key. You know, we want those people to work for us and to see a path to essentially change the world. You know, when we're working in the public sector, we're working with the people and you do have that opportunity as a developer. And so we need to keep them engaged and not have them go, uh, oh well this isn't exciting, I want to go work for a gaming company instead. We want them to feel that they matter because of the people that lives, that they, they impact. And that way they're able to stick around for a number of years and hopefully retire with, and then share all that great knowledge and be able to bring up people just like they were when, when they were potentially early in their careers and then have people that, that they mentor and then when they retire they have that next generation. And so it's a, it's a cycle that then reads excellence.

Speaker B: Yeah, yeah. And you know, with your role, you're traveling pretty much all around the country, you know, meeting, you know, customers. You know, when agencies are evaluating these AI uh initiatives today, you know, what signals tell you that they're thinking more strategically versus you know, they just simply chasing the latest trend.

Speaker A: Yeah, it's considering that worst case scenario, you know, and managing and then what does that mean? It's managing the risk behind all of that, you know, and so you have to, you know, consider how there was the, there's this citizen development fantasy that's Been going around for a number of years and I've seen some of the government agencies that I've worked with get, have got into a situation where it's caused ungoverned software shadow IT and then that gets taken over by the IT departments and they're like whoa, look at this application that was built. One, it's built on something that's no longer supported. Two, it has high risk for the day, uh, with the data that's, that's under the hood. And so frameworks are key to that and there, and so looking at the risk that's associated and then applying that framework. And so uh, there's the NIST AI Risk Management Framework. And so using that is one way to execute a long term plan with minimal risk. There's always risk with everything that we do, but it's trying to minimize it in a way that's repeatable as well as in a way that other folks are using. You know, definitely we want to have something that's uh, uh, a framework that's unique to our situation, our institution, our agency, our city, our county, our state. But we need to have something that's proven. And there are many frameworks out there. The NIST one is one that I like citing uh, and seems to be working really well for the people that I work with.

Speaker B: And you know, what's the biggest misconception government leaders have about agentic systems today?

Speaker A: Yeah, it's that I have agents, uh, you know, somewhere uh, within my agency and they must be working. But, but really when I probe into that I find that the majority of them just don't get to production that they're in this proof of, this never ending proof of uh, concept state. And then the other side of that is one agent to rule them all, you know, uh, or one model to rule them all. It's not as simple as that. You know, one model could be fine for one situation, but it's part of an agentic. When you're solving problems things are part of a greater agentic workflow. You know, I find that when I design workflows for the people that I work with, it's common that we have multiple agents and then within those agents multiple models that are part of what's going on behind the scenes. And what this is doing is it's helping to avoid the token apocalypse that's coming where you have tokens that are potentially that are rising in cost. And we don't want to get into a state where let's say we have an agentic workflow that's public facing. And then all of a sudden you have a rush of people coming to it. And we now have a very large bill with our provider. And so it's being able to manage that with having a workflow that makes sense and is able to break outside of using agents and models when appropriate, as well as go to the model that makes the most sense for the cost of the situation. You know, we might have a workflow that has the ability to be public and internal facing. So with the people that work for our agency, we may want to use a more expensive model to help solve a problem or answer a question. But for something that's more public facing, we may want to use a cheaper and more narrow, uh, focused model. So that way we don't break the bank, but we're still able to provide these types of, uh, technologies to the people, you know, the people that need it, but uh, and still provide high quality for the people that work with us. So that way we can have that measurable difference to the, uh, to the, our individuals as well as to our citizens.

Speaker B: Yeah, yeah. Ah. And you know, looking ahead, you know, obviously things could change, you know, but, you know, based on where we are today and where you see things going, you know, what will distinguish agencies that, you know, successfully operationalize AI from those that never get beyond these pilots and proof of concepts.

Speaker A: Yeah, it's. Can you get that agent governed and into production, you know, and it, uh, like I said, it can't just be these endless, uh, proof of concepts, you know, and agents are more than just chatbot. We have to consider tool usage and what else they can do, you know, but it's looking to that, to the future. You know, consider what the phone has, uh, been able to do over the last few years, you know, and the changes it's done in everybody's life. What would happen if AI was more on the edge here? Excuse me, on our device? Would that cause apps to start disappearing? You know, where you're now able to go straight to the phone and go, okay, well I need whatever. And then that agentic workflow kicks off. Um, we're now in a world where it's no click applications. And so if you're thinking that way, agents can be built right now that can handle that situation, you know, and so I work with a, uh, outside of the public sector, I'm working with a major tour company that says they want no click applications focused in that world. So that way they can just talk straight to an AI model and go, I need to rebook or I want to book an, I want to book a tour. But it's, it's being able to think into the future like that and go, okay, well if, if I have an agent, I can put web services on it. I don't even need an AI ui. And if I've done that, it's now separate, uh, from the applications. It could be put into anything or even brought into these models like an MCP server. Uh, and now we're in a world where that's much more future focused, where those apps start disappearing and we're in that state of that no click world with those edge devices.

Speaker B: Yeah. And you know, obviously, you know, you're at outsystems. The platform is evolving not year on year anymore. You know, it used to be, I remember in the early days when I joined in 2018, every year there'd be an announcement right. Where now it can't be annually. That would be too slow. Right. Things are kind of moving much quicker, you know, for you and the platform itself being there prior to this whole AI kind of era. And you know, it was already quite a surprise for enterprises when you're presenting this platform to them. That's really good on speed and on, um, development and you know, technical debt and things like that. And now with AI coming in and all these features with, with Workbench Mentor and with you kind of traveling all over the place and you know, in my opinion, there's still awareness gap still. Right. You know, in, in, in, in the U.S. so when you're kind of, you know, now going to Enterprises now versus three years ago, what's the how, how's those conversations and what's. What are people kind of, how are they responding to the platform and how it's evolving? And obviously now it's moved out of that low code space right now it's in a much saturated space of many players. So how are those conversations for you? Because obviously the way you're speaking would have evolved as well as the company evolved. So how's that been for you?

Speaker A: Yeah, yeah, I'm able to build stuff a lot more quickly and uh, it's always a lot deeper than just building stuff. It's being able to have the full picture. When Henry Ford envisioned the assembly line, he could have never seen where us at AL Systems have taken it. We can think of it as a complete factory and we've always been that, uh, for a number of years. We're not changing anything. But it's, at least from that perspective, we're Now a much more open platform where we're connecting with all these other great tools out there. In the same way a factory's assembly line works, you could have a lot of different flavors of tools within there. But the factory is the big picture because that's how you get things constructed and then eventually delivered to the people that need it. So the conversations have changed. And even for me, it's difficult sometimes to keep up with the engineering team and how quickly we're releasing great stuff like guardrails, the ability to make sure your agents are doing precisely what you want with evaluations and all kinds of great stuff. And so my demonstrations have become a lot more exciting. Uh, it's no longer the harbor tour. It's about having a good punch to the face as to the cool thing right away, where I build an application right away in front of people. That's what they want. So the proof, proof of concept happens almost immediately and then the conversation shifts drastically of, uh, oh, wow, you've built that well, now how do I test it? Okay, well, I have these evaluations. How do I make sure that the agent. How do I get agents into it? Well, it's as simple as connecting these things. And then here's our workflow and we've got that now into our application. And then how do I get this finally into a production situation when I have other people working in other versions? This is all stuff that I'm able to demonstrate now within an hour to an hour and a half. Where previously, you know, these things took a, ah, decent amount of time. But this is, this is the future. This is where we've gone and we have that full picture, that complete factory, and it's very exciting and it shifts, uh, like I said, the whole conversation. And people are now going, oh, wow. If I had, you know, my team focused in this perspective, I wouldn't have this development life cycle, this whole deployment cycle that is many systems that are brought together by custom code and individuals that are expensive to pay and other. And companies that this is what they make their money off of, is the glue between all of that. We're now that complete picture instead of having those fragile systems and, you know, we've all had those days, uh, in the past where we go to deploy and it's gotta be done on the weekend in the middle of the night by the Giant team. And we hope that everything goes according to plan and oh, shoot, we forgot a dependency. Okay, stop everything, roll back, call up the engineering team. Uh, what version is gonna work? You know, yeah, that whole song and dance is, uh, it is becoming a thing of the past. And that's one of the great things about Outsystems is we've solved that and done it with a future focused manner. And so we're now able to execute on all of that and even demonstrate it live in front of individuals. And so it's very exciting. And, you know, we're constantly bringing out new stuff all the time. And when I say new stuff, uh, it's connectors, it's solutions, it's looking at and having our pulse on the evolution of built of the next layer of that software stack and being one step ahead of it, as well as having our vision on what we did in the past and being able to use those things and leverage our history. Kicking butt.

Speaker B: Yeah. Uh, amazing, amazing. And finally for you personally, with things evolving and moving the way they are, what are you kind of most excited about for the future?

Speaker A: Yeah. Uh, it's very exciting to be able to do things fast and to have good advice when I need it. You know, sometimes I need things a bit more rapidly. You know, if a customer comes to me and says, hey, how do I bring this into this other system? I'm able to quickly, you know, punch that up in AI and it's able to bring that. But what I'm most excited about is that having those edge nodes, um, those no click applications. What does the future hold for us there, where we're just communicating with AI as opposed to having to fire up my Chase bank application and go, okay, well, did the check get deposited today? Oh, it didn't. Oh, no, I've got fraud, uh, types of things that I believe will be surfaced with AI when the governance is there, when the risk management is there. And all of that's coming together with mobile technology and Moore's Law, where these things are now doubling every 18 to 24 months. The power that goes into these devices will be able to handle, uh, a lot more AI there. And so this is what excites me. This is what, uh, I love to talk to customers about. And then how do we go from where we are today and the vision and those initiatives that come down from our leaders as well as if we are the leaders, the ones that go, uh, okay, well, where do we need to be in five to seven years? You know, when I write contracts and work with customers to do architecture, that's that going to last, you know, 12 years. This is the way we've got to be thinking. And so it's very exciting. Uh, I Over my lifetime of working in the tech industry, I've never seen it move quite so rapidly. We saw VMs and containerization concepts do rapid evolution and jumps within the tech space, as well as even agile development. My grandfather used to work for Hughes Aircraft and I found some of the slides in his briefcase from the 80s, and they were talking about this idea of agile development. He wasn't even called that for him back then, but those were the concepts he was sharing. And it was classified back then. But, uh, you know, and so this evolution of the, of the tech space to be able to, to move faster and do greater things, to serve the people that need it, because in the end, it's about the people. And so that's what I'm most excited about, is the technology moving fast and getting it to the people for.

Speaker B: Yeah, amazing. Well, I appreciate you speaking to me, Scott. It was a, uh, great conversation. Thank you for sharing everything you're seeing and how things are rapidly moving and I look forward to keeping in touch.

Speaker A: Yeah, you bet. It's been a pleasure talking with you today and look forward to more of these conversations in the future.

Speaker B: Definitely. Thanks, Scott.

Speaker A: Thank you.