This episode of the Tyler Tech Podcast, recorded live at Tyler Connect 2026 in Las Vegas, features a conversation with Jake Burns, executive in residence at Amazon Web Services (AWS), on how public sector organizations can approach agentic AI with a practical, results-focused mindset. Jake shares insights on common misconceptions about AI, why many initiatives struggle to move from pilot to production, and how leaders can identify low-risk opportunities for implementation. The discussion explores the importance of workforce readiness, organizational buy-in, hands-on experimentation, and skills development, while highlighting how agencies can build a foundation for long-term AI success and improve efficiency with existing resources.
In this episode of the Tyler Tech Podcast, recorded live at Tyler Connect 2026 in Las Vegas, Jake Burns, executive in residence at Amazon Web Services (AWS), joins host Josh Henderson to demystify agentic AI and explore what successful AI adoption looks like in practice. The conversation examines common misconceptions about AI, why many initiatives struggle to move beyond pilot programs, and how organizations can identify practical, low-risk opportunities for implementation.
Jake shares insights on workforce readiness, organizational buy-in, skills development, and the importance of hands-on experimentation. He also discusses how public sector leaders can focus on achievable use cases today while laying the foundation for more advanced AI capabilities in the future.
This episode also highlights emerging strategies for building efficiency in the public sector, with insights into how agencies are using technology and process improvements to do more with existing resources.
And learn more about the topics discussed in this episode with these resources:
Listen to other episodes of the podcast.
Let us know what you think about the Tyler Tech Podcast in this survey!
Jake Burns: And so, the ROI potential is higher than anything I’ve ever seen, and probably human history has ever seen. But it does require kind of understanding what it is and what’s required to get there.
Josh Henderson: From Tyler Technologies, this is the Tyler Tech Podcast. I’m your host, Josh Henderson. Today’s episode cuts through the noise surrounding one of the most talked about subjects in technology today, agentic AI.
As organizations explore artificial intelligence (AI), they’re increasingly hearing terms like AI agents, multi-agent systems, autonomous workflows, and generative AI. But for many public sector leaders, the challenge isn’t understanding the potential. It’s understanding what’s practical, what’s proven, and where to start.
Recorded live at Tyler Connect 2026 in Las Vegas, my guest is Jake Burns, executive in residence at Amazon Web Services –– or AWS. Jake works closely with organizations navigating AI adoption and brings a pragmatic perspective on how leaders can separate hype from reality while building the skills, processes, and organizational buy-in needed for long-term success.
In this conversation, we unpack the fundamentals of agentic AI, explore common misconceptions about automation and productivity, discuss why many AI initiatives stall between pilot and production, and examine the role workforce readiness plays in successful adoption.
We also look at how public sector organizations can experiment with AI in a practical, low risk way while preparing for more advanced capabilities in the future. There’s so much to cover, so let’s get started.
Alright, Jake. Welcome to the podcast. It’s so great to have you here.
Jake Burns: Thank you. Thanks for having me.
Josh Henderson: So, let’s start with the big picture here. There’s a lot of terminology being thrown around, agentic AI, AI agents, multi-agent systems. How do you define these concepts in practical terms, especially for a public sector audience?
Jake Burns: So, I think it’s a great place to start because I think people get kind of overloaded by all of these all the jargon. And if you can’t speak the language, it’s really hard to talk about it.
It’s really hard to follow, you know, thought leadership and content on it when they use those terms. So, it’s really not that, the terms are pretty simple, basic. So, like AI itself is just a very broad topic.
It’s existed for, you know, more than half a century. It’s really just, machines that, can do intelligent things like expert systems and so forth. Recently, you know, in the last few years, generative AI has been kind of the dominant form of AI. So, when people say AI, they usually mean generative AI or now agentic AI.
So generative AI is just what it sounds like. It’s AI that generates content. So that’s typically text, but it can be images, video, or really anything else. So primarily when people say generative AI, especially in the business context, we’re talking about large language models, we’re talking about text generation.
Now AI agents are kind of an evolution of that. So, if you think about what, like, text is, like, what do what do we do for a living? Most of us, we talk for a living, but there’s action that takes place in the world as a result of that, those words that we speak, you know, if you’re a lead if you’re a leader, especially.
And so agentic AI is kind of like the we take that text that’s output from these generative AI models and we give them access to tools. And tools are just things that can take action in the world, whether it is to, you know, send an email or whether it is to write to a database or consume information. So, when you give these generative models a goal, you let them iterate, and you give them tools, they become agentic, which just means they could do things in the world. And then agentic AI is kind of the systems.
You know, it’s more of the systems, whereas an AI agent is the thing. And then a multi-agent system is a more advanced version of that where you have multiple agents working together. And, you know, it could get a little bit more complicated but doesn’t have to be. So hopefully, that very simple way of kind of thinking about all those terms.
Josh Henderson: That’s a great place to start, I think. I’m just kind of outlining all of those terms for us, for our audience. There’s also a lot of misconceptions, it seems, with AI just because it it’s not it’s not new. And to your point, it’s not it’s not like it was created yesterday, but it’s become kind of a hype machine at this point. And so, there are misconceptions that come along with that type of hype.
What do you see as the biggest misconceptions right now when it comes to agentic AI specifically? Where’s that hype getting ahead of the reality?
Jake Burns: There’s quite a few of them. So where to begin? Maybe my favorite one. And I’d have a bit of a contrarian view on some of these things, but I do work with these systems hands on a daily basis, and I work with organizations around the world who are implementing this.
So, I do speak from experience. I’d say the biggest one is that it’s going to make your job easier, that it’s going to take work away, that it’s going to do the work for you. I mean, this is a little counterintuitive because the promise is that, you know, you’re going to have this new AI employee that’s going to do your work for you, when in reality, it kind of creates more work for you. If you look at the people who are working with these systems the most, they’re working very hard, but it’s worth it because you get so much leverage out of using these, and the better you get at using them, these systems, the more leverage you get. So maybe I’m working, you know, 20% more or harder, but I’m getting 10x or 20x the productivity.
That’s a pretty good deal. So, I think that’s one of the misconceptions, and one and that misconception leads to a major failure mode in implementations where organizations and people using these systems, they put in no effort and they just expect instant results, and then they become disillusioned, and then they say it’s all hype. Whereas at the same time, you have other people who are literally getting 10x, 20x productivity because they’ve put in the effort, they learned how to use the systems.
And, you know, when the former group hears about the latter group, they think that they’re lying, and they think that they’re hyped. And so, we have these kind of like parallel universes going on at the same time right now.
Josh Henderson: To everything we’ve been speaking about up till this point, it’s very clear there’s a lot of focus on fully autonomous systems. What are leaders hoping AI can do today versus what organizations are actually able to implement successfully right now?
Jake Burns: So, it really depends on the type of organization we’re talking about. And I think this is another, like, misconception or kind of disconnect within the industry. So, the things that startups are doing and, like, big tech, like, you know, Silicon Valley type of, you know, VC funded startups are doing is very real, and they’re fully autonomous, and they’re doing amazing things. Sounds like sci fi to most people.
Now the problem is that when, like, a government agency or a large owner enterprise tries to take on that because that’s what they’re told this technology can do, and it can do that, it’s a very big uplift for them, and they don’t have the skills. They don’t have the capabilities. They don’t really know what they’re getting themselves into. And they don’t really need to start there.
So, the kind of gap between what people talk about and what they’re doing is very large. When I work with these enterprises, when I see the systems that are working, the agentic systems or even the generative AI systems that are working, they’re very boring. Like the and they’re getting a lot of value out of them. The things like automating very simple workflows.
They’re not even fully autonomous. They might be like two or three steps. It might be like take like If you’re like take a PDF file, extract the information and put it into an Excel file and upload it.
That can have a lot of value in an organization if you have people manually doing that work and it’s error prone, but you can have AI do that, you can have other AI check the work and do it in a very safe and compliant way with guardrails.
That’s where the big value is coming from large organizations, and they’re succeeding with that. But when they try to do the same thing that the really small agile kind of high tolerance for risk organizations are doing, that’s when things a lot of money gets spent, with very little results.
Eventually, they’ll get to that point, but they should go for the, the low-hanging fruit, the high-value, low-risk opportunities first.
Josh Henderson: I wanted to kind of go back to something you mentioned earlier on in the conversation about how this idea that AI doesn’t necessarily reduce workloads, but it increases the value per unit of work. So, can you unpack what that means a little bit further for us and how leaders should be thinking about this idea of return on investment when it comes to artificial intelligence?
Jake Burns: It’s all about return on investment. So, the return on investment, you know, I have a formula. It’s basically like the quality of the tools multiplied by the skill, and there’s two types of skill. The skill in using AI, so skill your workforce has in using AI. This is domain expertise skill that they have.
So, a misconception if we’re going back to one would be that non-software developers are going to be writing enterprise software with AI. That’s just not happening. Right? They’re trying, but it’s there’s a lot of hurdles there.
Now actual software developers are getting a lot of productivity out of using these tools, but they because they understand the domain. So, it doesn’t have to be software development. It could be marketing. It could be supply chain.
It could be government policy. It could be whatever the work that needs to be done that you want to amplify and get more productivity and get more ROI out of, it’s that domain expertise. So that has to go into it as well. And then you have effort.
People underestimate the effort. And there’s a learning curve. There’s a very real learning curve with this. And the people who don’t accept that, maybe because they were told it’s just going to be an easy button or because they just don’t are unwilling to invest in that, they tend to get very little value out of it.
Because if you look at the best formula, it’s a tool times those things times iterations.
And so that one formula, you know, the quality of the tool multiplied by the skill and the effort multiplied by the iterations explains why some people get no value out of it even though they’re highly skilled but they put in very little effort so it’s multiplied by zero.
And others get actually negative value out of it. They create a mess where there wasn’t one, and still others are getting these real outsized returns. And so, the ROI potential is higher than anything I’ve ever seen, and probably human history has ever seen. But it does require kind of understanding what it is and what’s required to get there, and it is an investment.
Josh Henderson: Stay tuned. We’ll be right back with more of the Tyler Tech Podcast.
Every day, government leaders are being asked to do more with less, manage growing workloads, meet rising expectations, and address staffing and budget challenges.
Jade Champion: And while efficiency is a priority for nearly every organization, today’s challenges require a different approach. It’s not just about improving individual processes; it’s about expanding capacity without adding resources.
Josh Henderson: That means connecting systems, simplifying workflows, reducing manual work, and giving staff more time to focus on the services that matter most.
Jade Champion: Across the country, public sector agencies are improving operations through automation, artificial intelligence, integrated systems, and smarter ways of working.
Josh Henderson: Discover how agencies are reducing complexity, freeing up staff time, and creating more capacity to serve their communities. Visit the link in our show notes and download our free e-book, How Governments Build Efficiency at Scale, to learn from real-world examples, client stories, and expert insights.
Jade Champion: Now let’s get back to the Tyler Tech Podcast.
Josh Henderson: And now there’s also this interesting thing where, you know, AI pilots might succeed, but the deployment might fail. So why does that happen, and what separates organizations that successfully move AI into production from those that don’t or aren’t able to?
Jake Burns: This is a huge problem right now, and it’s one of the biggest specific problems that I’m dealing with when I engage with organizations. So, their pilots succeed, and then they get stuck in pilot purgatory, and they can’t get out. So, the problem is that the environments that we develop pilots in are highly organized, perfect, best-case scenarios.
And we put nondeterministic systems in there and they work. If you put nondeterministic systems in environments that have nondeterministic variables, then that you get things get very complex.
So, the problem is that they try to go straight from prototype to production, and there’s just this huge gap. You know, you’re dealing with sanitized data, and then you’re going to messy data. You’re going from controlled environment to kind of an uncontrolled environment, and again, you have this amplifier. It’s a mirror and an amplifier.
So, it’s amplifying all of this chaos that you have. So, what needs to happen is there needs to be a middle step. There needs to be kind of a bridge between your prototype and your production.
And this is what most organizations are missing is that bridge. That bridge is where you take these you know, the prototype, you just iterate and iterate and iterate and fail a bunch of times until you find, like, the great thing that you want to invest in. Once you find that, then you need to create, like, a best effort replica of what production looks like. So, take sanitized production data, put some messiness in there, and really kind of make the system robust and put guardrails around it, secure it.
A lot of times you’ll find that a lot of the, you know, agentic capabilities or the LLMs or the AI models are actually could be replaced with deterministic code and scaffolding, and then maybe only a small piece of it needs to have that, and so really kind of reducing, and one of my favorite is reducing multi-agent systems into a single agent, because most multi-agent systems actually work better that way.
Not all of them, but a vast majority of them and it simplifies everything. And then once you have that, then you can, you’re ready to take that workload and put it into production. And you need to be doing this at a small scale before you do it at a bigger scale.
This is another problem. People, they have a lot of groundhog day moments. So very common scenario I run into is we did a pilot and it was a simple chatbot and we’re like, you know, we got it working. It was hard but we got it working.
And now in production, we’re going to put this full agentic orchestrator that pulls data out of 20 different proprietary data lakes, interfaces with, you know, 10 different departments, 10 different systems, and, you know, it’s just going to just magically work. And I said, well, that’s a pretty big leap from what you just did in a prototype. So, take a very small piece of that, find where you can get the most value with the lowest risk, you know, the minimum viable kind of, like, workload that is big enough that people don’t say so what, but is easy enough to you can actually succeed and then iterate from there and do it iteratively.
Josh Henderson: And I feel like this entire conversation has kind of led to a big theme that you often touch on in in the work that you do, the importance of people and skills in this whole equation. Why is training and hands-on experience so critical when it comes to the effective adoption of AI?
Jake Burns: It comes back to the fact that skill is a necessary ingredient. I would say skill, effort, and differentiated data are probably the most important ingredients to be successful at scale for a business or a large organization. And people underestimate that skill.
They think it’s again; they think it’s going to be just an easy button. People sell it that way, but it never it never really is that. So again, it’s amplifying what you put into it.
So, if the people who are operating it or the people who are deploying it don’t have that skill, then you’re really like multiplying by zero again, right, or worse. So upskilling is super important. Now the problem is that the framing of that, before you upskill an organization, you know, and I was very fortunate that I went through this with cloud, so kind of apply these same lessons to AI, and they’re very proven. Before you upskill, you have to get buy in.
And buy in is a process. So, you have people out there, you have workforce out there that’s scared. They’re scared of being automated. They’re scared of losing their job.
They’re scared of not being able to be able to use these systems. They’re scared they don’t know that these systems even will produce value. They have all of these concerns that are not being addressed.
So, the buy-in has to come first and that I could talk forever about this, but there I have a formula for that and there’s ways to do that. The biggest part is kind of what’s in it for them. So, they have to see that there’s like a value that they’re personally going to get out of it.
And the good news is they will, but it has to be articulated. Then the upskilling. And then you mentioned another important thing, which is the hands-on experience.
So, if you just do the upskilling, then it’s like basically a paid vacation. If you just do the hands on experience, then you’re giving people a very powerful tool, but they don’t know how to use it. So there has to be equal parts theory and practice.
And so, I like to mix those together as much as possible. If you could do the upskilling while giving them opportunities to use it hands on, that’s when it sticks. So, you just get buy-in, then you do upskilling and hands-on experience at the same time. And that second part, it never ends.
The training should never end and the hands-on experience should never end, and it kind of feeds into each other.
Josh Henderson: That’s really good and valuable stuff. Now for organizations that are just getting started with all of this, what does a practical low risk approach to experimenting with AI actually look like that you’ve seen?
Jake Burns: So, you want to have a safe place to experiment because experimentation is going to be, you know, key to success. Because despite what, you know, industry says, we don’t have all the answers. We’re still figuring this out, you know.
I’m on the frontier of this, and I can tell you that we’re discovering things every single day, new use cases, new ways of working with these models. It’s a moving target. That doesn’t mean you should wait, though, because, you know, it is moving very fast, and you want to be a part of this.
But you need to have a safe place to experiment. So, you need a place where people feel safe. They’re not going to get fired for making a mistake.
You don’t have any sensitive data in this environment. It’s completely firewalled and segregated. I even go so far as with some customers to tell them spin off a company if you can that’s separate from your own company to kind of do this experimentation work because it’s really hard to be innovative in a bureaucracy.
Because you’re just constantly fighting friction. So, it’s almost better to have a whole like, much as possible, have a whole new completely segregated section. And it could be the same people working, you know, in production and working in this innovative, you know, kind of experimentation area, but, like, make it safe to do so. So put guardrails in there but give them the freedom to use any tool to experiment, to solve any problem, to solve no problem, just play with it.
Really, people just need to get hands on with it. The more people use it, the more they want to use it.
And the more they use it, the better they get at it.
And so, while the training is important, I would say the hands-on experience is even more important. And they will find the use cases aren’t predicted. They’re discovered.
And the same thing I would say is true for your best AI engineers, and I say this about cloud engineers. Your best AI engineers will be discovered, not predicted. It’s very hard to know in your organization who it’s going to be.
Sometimes it’s who you think it is. Sometimes the people who will surprise you. Because it’s very different archetype to be successful with this than traditional high-productivity workers.
Josh Henderson: Really great stuff. I’ve learned so much during this conversation. I know our listeners will learn a whole lot as well. But to wrap things up, I just wanted to get your advice on what public sector organizations should be doing now to prepare for more advanced capabilities like agentic AI as they move into the future.
Jake Burns: So, I think, you know, you just need to start using it. You need to start learning about it. You need to, you know, get your workforce bought into it.
I’d say that’s probably the biggest barrier. The biggest barrier is the workforce. A lot of times you know, but a lot of times you don’t know. A lot a lot of projects, especially, you know, in large organizations, what ends up happening is, you know, they go into a meeting with the boss about the new project, about AI or whatever it is, and they nod and smile, then they go work on how they’re going to, you know, undermine the effort.
Because they see it as a threat. So, their concerns have to be addressed.
Again, the great news is that this is good for them. It just needs the leaders need to understand why it’s good for their workforce, and that’s why it’s good for their organization and for them personally. And then you need to learn how to clearly articulate that and really get that buy-in. And it’s been my experience that if you can get buy-in from an organization, from a workforce, that they will solve the rest of the problems for you. So, I can give advice on how to control costs, how to do governance, how to improve sufficiency, how to find use cases and all that. But you’re going to do only one thing, it gets buy-in from your workforce, and they will solve the majority of those problems for you.
That’s really great. I think that’s a great place to wrap things up. Jake, thank you so much for taking the time to chat with me today, and can’t wait to have you back on the show and to talk more as things, you know, develop even further.
It was my pleasure. Thank you.
As we heard today, Jake emphasized that AI adoption is ultimately about much more than technology.
Success depends on people, skills, trust, and a willingness to learn through experimentation. While the tools themselves continue to evolve rapidly, the organizations seeing the greatest returns are investing in workforce readiness, creating space for hands-on learning, and building strong foundations before scaling to more complex implementations.
For public sector leaders, that means focusing less on the latest buzzwords and more on developing the capabilities, processes, and culture needed to take advantage of the opportunities AI presents both today, and in the years ahead. If you’d like to learn more about the topics discussed in this episode, be sure to check out the show notes for additional resources.
We’d also love to hear your feedback. Fill out the listener survey linked in the notes or reach out anytime at podcast@tylertech.com. And be sure to subscribe, rate, and review the show so you never miss an episode.
For Tyler Technologies, I’m Josh Henderson. Thanks for listening to the Tyler Tech Podcast.