Who's Building Your Software Now? IT, the Business, or the AI?

s1e33 Who's Building Your Software Now? IT, the Business, or the AI?

This episode of The Wired Garage with Pops digs into how software delivery is being fundamentally restructured — not just accelerated. Pops and Steele walk through the evolution from traditional IT-led development to a multi-persona development model, where business users (citizen developers), professional developers, platform teams, and AI agents all share responsibility across the app lifecycle.

The conversation starts with the "why" behind citizen development — IT backlogs, understaffed teams, frustrated users waiting months for simple solutions — then moves into how governed, low-code platforms let business users get in the game without blowing up the architecture. They draw sharp lines between citizen development (process-governed, platform-scoped) and Shadow IT (your cousin's 99-cent app running on Bill's laptop).

From there they transition into agentic AI — what separates a chatbot from an agent, how AI agents plan and execute autonomously or with a human in the loop, and why governance applies to machine personas just like human ones. The episode wraps with a live screen share walkthrough of ServiceNow Studio and App Engine Studio as real-world examples of governed sit-dev platforms, plus a broader call to action for teams to start experimenting now.

Keywords: citizen development, multi-persona development, agentic AI, low-code no-code, Shadow IT, ServiceNow App Engine, ServiceNow Studio, governance, blast radius, human in the loop, AI agents, workflow automation, platform ROI, digital transformation, pro-code vs low-code, guardrails, two-lane highway, delivery velocity, compliance by design, enterprise AI

Key Takeaways

• Multi-persona development is the new operating model — business users, pro devs, admins, and AI agents all contribute on the same governed platform
• Citizen dev ≠ Shadow IT — the difference is process: a pipeline from idea to production vs. winging it with whoever knows somebody
• The "Two-Lane Highway" principle — how much breadth you give someone reflects trust, character, and risk tolerance; guardrails build confidence, not restriction
• Risk, Complexity, and Blast Radius — three filters to determine whether a task belongs to a citizen dev, a pro dev, or an AI agent
• Agentic AI is a governed persona on the platform, not a side experiment — it plans, executes, and can operate autonomously or human-in-loop depending on stakes
• Executives want ROI proof before scaling AI — excitement is real, but accountability, decision rights, and governance have to come with it
• Pro devs should be advancing the ball, not building email management workflows — free them up for what actually moves the business
• The workforce shift is already in motion — IT pros have always had to evolve; this is just the next lane change, and language/prompting skills are now table stakes
• Low-code tools like Make.com, Zapier, and AI assistants are accessible entry points for anyone wanting to get started today

Transcript

SPEAKER_02 0:00

Citizen development. Why do I need multi persona development? It is a form of enablement. The example I like to use is if I'm a baker in a bakery and I can only cook two loaves of bread a day, but I can teach someone how to cook a loaf or two loaves or more and help them to fish, or actually that would be a better analogy, is teaching somebody to fish is very similar to that. Not everything needs that pro code super strategic tactical approach. We can put someone on guardrails and then they can go and create that loaf of bread, go fishing and self-serve or help solve a problem that exists, that the only reason why it still exists is because you just don't have enough delivery capacity.

SPEAKER_00 0:47

I like the baker one because if say you can only do 12 loaves a day, but you've got more people wanting to stand in line, wanting, but if you can, let's say your magic is in how you I don't I don't know what you what your term is, but build the dough, right? What's in the what's in the dough, right? That's your magic. And maybe your magic is a recipe. You can just sell the dough in the recipe and let them cook it, right? But they're following your, they're taking your ingredients and following your plan, but they're building their own loaf of bread. And I think that analogy is perfect because that's what we're doing.

SPEAKER_02 1:28

Yep, and it and it scales. And your platform is delivering value at an outsized pace, and it's also incentivizing usage of the platform, and it is a form of return on investment to you, so it is helping to maximize your investment in the platform as well.

SPEAKER_00 2:44

So we start with the citizen development where business users can safely build apps, automate work on governed, like low-code platforms, then move into a gentic AI where machine personas begin taking on parts of the work themselves.

SPEAKER_02 3:00

Big shift is not just faster delivery, it's a new operating model. We're talking about how humans, platforms, and AI agents share responsibility across the app lifecycle, how governance keeps the model safe, and how leaders can turn all of it into a real competitive advantage.

SPEAKER_00 3:18

So in our fire it up section, this is like what is sit dev really? So citizen development is about letting business users build useful apps and automations with low-code, no code, no code tools, say that fast, while IT keeps the guardrails in place. So in practice, it helps organizations close the app gap by moving simple workflows and forms and approvals and some department tools out of the hack log and into the hands of people closest to that work. So if you if you prefer the term, and I do, multi-persona development, the idea stays the same, but the framing just gets a little broader. It emphasizes that business users, pro developers, admins, and platform teams all contribute to that delivery on the same platform. So the main change is not the technology or the delivery, it's the operating model and how you describe the collaboration.

SPEAKER_02 4:12

So thinking through that, what problem is citizen development actually solving inside large organizations? Why? Why do I need the citizen development? Why do I need this multi-persona development?

SPEAKER_00 4:26

And my examples of the places I've worked in my career, I run into the teams that are developing this are smaller than the amount of people or groups that want the work done, right? So you spend a lot of time in like design meetings, discussions, documentation, which all IT people hate to do, right? And you it it's it's just a long effort. And I think your user gets tired. It's like, you know, I wanted that six months ago. You know, now I'm found another way around it because their business keeps going while you're sitting there trying to figure out how to design it and get it going and code it and test it, and and then they want to make changes of the form because you didn't listen to the meeting, to them first time, you know, all those different things. So to me, this this actually lets them take it 100% of the way if you trust them, 80% of the way if they're decent, or at least 40 or 50% of the way that you capture their idea in some type of guard it, guardrailed, governed format path and let them get involved. I think they they end up also, sorry to keep this for a cool one, but I think they also learn how difficult it is to like build a form and get it to look right. I mean, they won't be asking for so many changes, right? Because now they kind of understand what work is involved in it. What do you think?

SPEAKER_02 5:43

I I think there is part of that to it. I also think it is a form of enablement. The example I like to use is if I'm a baker in a bakery and I can only cook two loaves of bread a day, but I can teach someone how to cook a loaf or two loaves or more and help them to fish, or actually that would be a better analogy, is teaching somebody to fish is very similar to that. Not everything needs that pro code super strategic tactical approach. We can put someone on guardrails and then they can go and create that loaf of bread, go fishing, and self-serve or help solve a problem that exists, that the only reason why it still exists is because you just don't have enough delivery capacity. So I think it's a good way to argument out because it is a good way to argument.

SPEAKER_00 6:36

I like the baker one because if say you can only do 12 loaves a day, but you got more people wanting to stand in line, wanting to, but if you can let's say your magic is in how you I don't I don't know what you what your term is, but build the dough, right? What's in the what's in the dough, right? That's your magic. And maybe your magic is a recipe. You can just sell the dough in the recipe and let them cook it, right? But they're following your they're taking your ingredients and following your plan, but they're building their own loaf of bread. And I think that analogy is perfect because that's what we're doing.

SPEAKER_02 7:16

Yep. And it and it scales. And your platform is delivering value at an outsized pace, and it's also incentivizing usage of the platform, and it is a form of return on investment too. So it is helping to maximize your investment in the platform as well.

SPEAKER_00 7:35

So, how do you define the difference between citizen development and shadow IT?

SPEAKER_02 7:40

Well, shadow IT is a much cooler name. Uh very much. I I think the best or the most concise way to differentiate the two is process. If you have multi-persona development, citizen development, there is a pipeline process from A to Z of how something gets into the funnel and then something gets to production. With Shadow IT, it's who's who's my person who helps me solve the thing, and they kind of figure it out. So I throw out a problem. Hey, I'm trying to do this, I've got a 99 cent app, my cousin, uh, or hey, have you tried this? It's running on Bill's laptop. So your outcome may vary, but most concisely is process. What kinds of apps are best suited for business users to build themselves?

SPEAKER_00 8:40

I think anything that is sort of electronically tracked, like through a Excel spreadsheet or some type of filing that is not business effective, takes too long, you don't know where it stops, you don't know where the problems lie, you know, it's not it's not really working, right? So a lot of workflows are workflows are the easiest ones. Um, apps that keep track of things like if you have, I don't know, an insurance cover, your insurance carrier, and you want to keep track of all of the insurance policies of your clients and you have expirations date, you can set it to where 90 days prior to that. I want to be notified that it's going to expire so I can call them and talk to them about all the different packages that meet their needs. You know, just different things like that. Then you you can also automate some of those things. So it'll send the email for you. Excuse me. You know, just I think apps like that, very simple ones, things that's in the head of the people, that's really what you want. Nothing so thick that it's, you know, that million-dollar platform. You want something simple that's going to make them happier on the platforms they have today.

SPEAKER_02 9:51

I think if you've got a very defined set of inputs and you have a well-defined set of outputs, and you can just put that in a workflow, and it's not something that there's already a solution to somewhere else in the organization, you're you're probably onto something.

SPEAKER_00 10:08

And I think with some of the tools in building these, they they use the same logic like today when you do a Google search or you do some kind of a prompt in perplexity or Chat GPT or something. Normal people are use are learning how to phrase those prompts to get the right response. They can start phrasing the right needs into these engines that these low-code engines that can help them build three-quarters of the way through it, in a sense. So I think they're learning some of these traits that allow them to do this anyway. So I think maybe it is not today, but if you train, if you have some training on your users on just phrasing and prompting, I think they'll be more successful in those tools. So, what should be the line between citizen developers and professional developers?

SPEAKER_02 11:01

Boy, uh I'm sure there'll be a lot of answers depending on who you ask on this one. The way that I think about it is risk, complexity, and blast radius. If they are working on something that is very risky to the business, it might be simple, but if it goes wrong, it brings down your network.

SPEAKER_00 11:18

Yep.

SPEAKER_02 11:20

If it is super complex, multiple integrations, multiple systems, probably not the most well-defined set of outputs. Probably need a project manager and all kinds of stuff for that. Product, product manager, product owner, project manager, all the acronyms, PDMs, PDOs, PJAs, or PMs. Um, and then blast radius. So if we think about what the security implications of this is, do are we considering OAuth, OIDC? Are we using LDAP? Is our security architecture need to be considered here? And then when we take into account all three of those things, we consider the solution. Are we making the best decision with the right folks to provide the best outcome? So risk, complexity, blast radius.

SPEAKER_00 12:11

I agree with that.

SPEAKER_02 12:12

What governance and review process keeps this model safe at scale?

SPEAKER_00 12:17

I think it's the, to me, it's the guardrails of what they're allowed to have exposure to, you know, and draw off of andor change. You know, you don't want them in the ACLs that'll change other apps. You know, you want them in a scoped, scoped world that's cut cut away to them in a sense. If it goes global or something else or ties in other things, you got to make sure there's still a hefty guardrails in there. Otherwise, you just can't let them in. Um, I think that's the type of governance that you got you mentioned the others where security and all these other teams are going to be involved if you cross certain lines. Either in complexity with scripting and all those sort of things, or if you're getting into um exposing this out to the outside world, or I've got sensitive data, all those sort of things, you bring other teams into it there. I go, but the biggest thing I think to governance here is you you want a two-lane highway in a sense. You want guardrails to keep them safe, keep them confident, and what you'll do is you'll build something very solid.

SPEAKER_02 13:17

Do you want to add anything to two-lane highway? So I I've heard that before, but for folks who might have not heard of it, what what is a two-lane highway?

SPEAKER_00 13:24

I don't know. I've always used that term. I've got back in my earlier days, I had people I trusted more, and they had four-lane highways. And I had people with two-lane highways, and I had somebody with an old dirt road. That's the only about the only thing I could trust them with. You know, it's just how wide of a sp of off kind of off the leash or whatever can you let them go? And you know they're they're sound and decision making. They won't bail on you, you know, and try to sneak something through. Can you trust them? Their character, all things play into that. And I think that's something you just gotta you gotta do. I think to me, that's what that means. Maybe other people would define it something differently, but I mean it's how much breadth I give you left to right, in a sense, to make your own decisions.

SPEAKER_02 14:14

Gotcha. I think part of the governance and and maybe even part to that analogy is also velocity. So part of the theme of enablement is velocity. And I think if you put model, a governance model and a review model in place, you set expectations up front, you have those defined inputs, you know what you can expect coming out of whatever the effort is. So there's probably some idea, maybe a review with a product owner or product manager. Should we do this thing? Yes, here's a set of steps. And you know at the end of the day it's gonna be a service request or some kind of record producer or something. I think that helps you build trust and keep trust because you've defined kind of where the boundaries are, and boundaries kind of keep everybody safe, also makes it really easy to build trust. Um, and then probably having some kind of re retro or continuous improvement process to just check in every now and then. Do we have enough training? Are we upskilling folks as techniques shift? I think that's kind of been like a sleeper thing is so much of the platform and the recommended practices have shifted over a decade plus. And keeping up with those is kind of the difference between building for the future and just kind of building where you are now and having to move tech debt later.

SPEAKER_00 15:45

Yeah, I I think you're because your users get smarter, right? In the beginning, those simple workflows is good enough, but their dreams are growing, their needs will grow and expand, and you need to keep up with that. A two-lane highway may not last long, you may have to go four lanes, not because of going left or right as much as is handle the amount of cars, the amount of work, and the speed, right? So that's one thing. And I think what that gets into is building there's like there's companies out there that have all these different automation APIs and tools that'll tie one system to another, you know, tie like buffer, like I'm looking at something that ties my inno reader with buffer, you know, takes news feeds that I like and track, tag them, and have it automatically create a draft in my butt in the in the buffer application so I can then post it. You know, those sort of API workflows, and I don't have to think about it. They just give me the plug-in, they just give me the little what way to work. I think you use your senior developers, your professional developers, to build those to add to their catalog of abilities that they can do. So it gives them more they can do. You're still constrained, they're still working in an area that you've already prescribed, but now they can do more. And I think if you keep feeding that, that's something that I think is sustainable.

SPEAKER_02 17:05

So intragentic AI. Because it's now not just sit dev. We have AI through in the mix. And so one step further by adding that machine persona. AI agents don't just help with prompts, but actually perform parts of the work. That means that agents can plan, call tools to do work, execute workflows, have extended capabilities, and run continuously alongside business users, pro developers, and platform teams. For a ServiceNow audience, the key idea is that agentic AI becomes a governed part of the platform story, not a side experiment. ServiceNow's Agentic AI framing emphasizes AI agent orchestration, workflow execution as a first-class capability on the platform and in enterprise operations. I think that's absolutely on the mind or near top of mind for every executive today that owns one of these platforms with these capabilities.

SPEAKER_00 18:12

Yeah. So if we back up a little bit and kind of go into some of the episodes we've already recorded, but just ask the question. So just in case someone doesn't know, what made what's the difference between a Gentic AI and your traditional co-pilot and chat bots that are out there?

SPEAKER_02 18:27

Sure. So I think when Jen AI came onto the scene, it came out in the form of like Chat GPT, for example. You could type to it, get a generated response. It was your therapist, it was your confidant. No, just kidding. But you you it would generate chat, uh text for you. And then kind of the next step was, huh, we could add a button and then have a generated prompt. So then summarization kind of came into play. And then, hey, we can summarize. What else can we do? And then all these skills started developing. And then, hey, we've got all these skills, but what if we had something that wasn't deterministic? What if we could have an agent plan and then go do something if we agree with that plan? And so planning, executing, doing the work, maybe it needs to get an approval or just acts autonomously. That is where agentic is different than a predefined skill and different than just a chat bot that's not really doing any of the work.

SPEAKER_00 19:34

So agentic AI, if running full loose, can just run on its own and there's no man in the middle, there's no handling of it. You can schedule the agentic AI to act actually acknowledge or recognize scenarios and act according to how it is developed to act.

SPEAKER_02 19:53

Yeah, there's your wallet is in the middle.

SPEAKER_00 19:56

Yeah, yeah.

SPEAKER_02 19:58

So I would say that depends on the parameters around the agent. Some agents designed to be human in the loop or have a human in the loop, just making sure that the output is correct, you have that trust, you're hitting approve. Maybe it's an agent that handles money and there needs to be some oversight there versus an agent that might just be generating resolution notes, handling tickets, or solving a VPN issue that we've seen a thousand times.

SPEAKER_00 20:28

We have to know how to reboot the server just because it did this 10 times. You you reboot or re-restart the service and it's fine. Why wake the engineer up and make them do that simple task? And you know it's contained, it doesn't take a lot of risks, there's not a lot of there's a blast, right? You took blast radius and your wrist and things like that are small. And I think, yeah, yeah.

SPEAKER_02 20:49

And that's a happy engineer who's getting some extra sleep. Oh yeah. He can pick up donuts the next day.

SPEAKER_00 20:55

Unless he's worried about his losing his job because now he doesn't he doesn't wake up in the middle of the night and take care of that problem.

SPEAKER_02 21:00

Now he's the one building the agents.

SPEAKER_03 21:03

Yeah.

SPEAKER_02 21:04

Um so which parts of the CLC process, so software development lifecycle are most realistic for agents to own today? I think I will take that question. I think every part of the process is gonna be touched. If we think about um planning, I mean there there's gonna be an inception part, but I think even at the inception of ideas or the building of a backlog, some of those may be spawned from things that agents have noticed, either running on the platform or security gaps. Like there's gonna be a a self-fulfilling loop there. Development's already impacted it. There's no question about that. It's already happened, it's real, it's here. Engineers are using it, not just for development, but also code reviews. GitHub is really. Good example. There's integration there where you can have an agent look at your code review it. There's a company that launched not too long ago that is doing something very similar through testing, through quality of it, quality assurance, having an agent review all that. Even uh kind of near the end of the process, after I mean you already have CICD, um, but let's say something fails or is not the happy path, you have agents that are picking that work up at triaging.

SPEAKER_00 22:33

And so, and I don't disagree with you, but I have a hard time with them running 100% of a path of anything because they they can be led astray. Just because someone complains of something doesn't mean it has to be fixed.

SPEAKER_02 22:55

I think that's where you still have that. It depends on your how your pre your planning process is. But you still have that planning, that collaboration with your product managers and your product owners, your release train engineers to make sure that the backlog is being groomed and that the business requirements, the outcomes that we're going after, what are the goals for the quarter? I think that is all still being set. And then there still has to be somebody at the end of the day that is making sure that what was delivered and what is executed and in execution and has been deployed to prod is doing what it was supposed to do. So I think that there's still going to be, if you imagine that being the guardrails for what for this, that's what I would kind of paint the guardrails as. Each agent should have its own scope and is generally trained to be very good at the one thing that it's doing to prevent hallucinations, to prevent issues. And so if you have a specific agent that is doing your coding, you would want ideally a separate agent that is doing the testing and so on.

SPEAKER_00 24:39

Would it be a different agent or a different package? Because aren't they using the same sources?

SPEAKER_02 24:46

Wouldn't it be like the same model?

SPEAKER_00 24:48

Yeah, yeah. Like it it's building it. It's like let's say ServiceNow develops a product and we can use it to, you know, to go ahead and just play, and they do with build, right? Do I want it to also do the code check? Or do I want something else third party that's not sharing in the same soup in a sense?

SPEAKER_02 25:08

Separation of duties.

SPEAKER_00 25:09

Yeah, it's separation of it's a purely separation of duties because it these guys are these AI agents are actually smarter than humans, right? They're going to use everything to the nth degree, but they're going to also test the same for nth degree, and it's going to pass every time, you know, where there might be some holes. There might be something it didn't see or know of that it didn't think of. It's not going to think of it all of a sudden use a different agent. You know what I mean? You need to have a different platform, I think, different sources, something. I don't know. Maybe I'm wrong.

SPEAKER_02 25:41

I think the most critical thing to that is that you have a model that is fit for purpose with an instruction set that is fit for purpose, that has has been evaluated, has the trustworthiness, and is following the rules essentially for what it is reviewing. Whether you put that in service now, you put that in a separate platform. Um, I think that is the most critical piece is that whatever is doing the testing is scoped and built for purpose built for what it's doing. And it's probably not the same agent that is handling HR onboarding or scheduling delivery of a loaner laptop, for example.

SPEAKER_00 26:22

Right. So let's skip in our fired up section here down to our last part. The big picture. Excuse me. The big picture is that your operating model now includes both human personas and AI agents across the app the app lifecycle. Instead of treating Ray C as like a static process chart, then this becomes a more practical way probably to show who sponsors, governs, builds, reviews, runs, and improves the work when developed. And it's distributed across business owners, pro devs, platform owners, and machine actors. So like a useful podcast framing is that the real shift is not just about more automation, right? But clearer accountability and a hybrid human plus AI delivery model. So the platform team seems to own the guardrails, humans stay accountable, and AI agents increasingly become responsible for the execution inside those boundaries, right? Do you still feel racing is useful when agents, AI agents are part of that workflow?

SPEAKER_02 27:25

Yes. I think at any point in time, someone has to understand who's responsible for what? Because as the famous quote goes, cannot hold a computer accountable for management decisions. Yeah. So how does a multi-persona AI agent model change the way teams think about ownership, if at all?

SPEAKER_00 27:51

I would say I think it's the same way as any other system that they trust in their data center today. There's so much like you know, VMware and and all these, you know, the virtualization of servers, how they build servers, and there's automations there, and Microsoft and with Azure and the AWS and all these things that are going on. We trust machinery, machine learning, different things to bring us answers and we act upon it. So I think that multi persona and the AI AI agent model, I think we still own it. We still want to control it and make sure it works correctly, but we're we're happy to have these tools that are helping us through it. The world is just exponentially getting more difficult to manage, to keep it secure. Your users are smarter than they were before. And I think it's just it's difficult with in with money, or by there's no IT team says they have enough people, enough skill. So I think we appreciate these multi-persa developers who are helping develop their their own workflow. We're happy with AI agent models, and just like any other tools we buy. That's my call.

SPEAKER_02 29:03

So if you're in danger. I I I agree. I I think that that's the root of it, is that there's never enough resources. So anything that can kind of expand that, whether that's citizen development, whether that's an AI agent kind of helping with development as well, helps kind of release some of that burden as long as there are those guidelines and guardrails so that you're not spending extra time cleaning up a mess or trying to uh clean up that blast radius.

SPEAKER_00 29:30

Yeah. So which life cycles are best suited for AI agents versus citizen developers or the pro developers? I think we already covered that a little bit. You said all of them.

SPEAKER_02 29:40

Yeah. Uh so when I think about the life cycle, and I I guess my brain goes to the CSDM and CMDB when it comes to life cycle. So kind of from inception to retirement. So is this in kind of to frame this or maybe add some clarity to this, when we say life cycle steps, are we thinking about kind of the development process? When should something be handed off to that AI agent highway versus that SID dev highway versus that pro dev highway? Okay. So I think that for AI agents, again, maybe the blast radius is really controlled. It is a process where you have a pretty decent level of trust. And maybe it's human in the loop, maybe it's autonomous, but you've got purpose-built model, you understand what the inputs are, you understand what the outputs should be, and you've evaluated it. You have a framework to evaluate. You've proven that trust. And I would say solving problems that to a pretty high degree of certainty, you know there's going to be a successful outcome. People aren't perfect, AI isn't perfect, pretty similar there. I would say for sit dev, if you just don't have the resources, but you have very well-defined inputs, very well-defined outputs, and it's not something that we would send to an AI agent, I would then put it in the sit dev or that sit dev lane. And then for pro devs, let's be honest, they're gonna go and try and figure out how and go make an agent to go do it. I would say that's probably where they're gonna start first. And then if that doesn't work out for them, then they're gonna go develop. And again, it's that risk, that complexity, and that blast radius, taking the guardrails into consideration.

SPEAKER_00 31:26

Yeah, I I like it because I think the easy stuff can be done through this multi-persona development. Um, and let's not kid ourselves, even if the citizen developers get the these tools in their hands, AI is right in the app.

SPEAKER_02 31:41

They're just they're starting with the prompts, yeah.

SPEAKER_00 31:43

They're starting to the prompt. It's just natural language prompts. But it helps. No, no, no, no, no question there. But I do want to put my pro-developers into advancing the ball and let us keep moving ahead and get rid of these other issues that are problematic. Instead of, can you work on a workflow that'll help me manage my email? Well, no, you can develop it. And while I'll put my developer on something more important that's actually going to advance the business or protect the business. And I think that that's where those values are. So the executives are responding to an agentic AI with both excitement and caution. They they see real boardroom value and the efficiencies, the workflow redesign, the faster execution, but they also want clear governance, defined decision rights, and proof of ROI before sealing it broadly. The strongest pod, the strongest thing to think about here is the executives are no longer asking whether AI can do more. They're asking how fast can it do it, how safely can it do it, and who remains accountable when it acts.

SPEAKER_02 32:48

Are executives actually thinking about AI today? I think you laid out kind of the framework or the groundwork for probably pretty much every executive. And I would say a lot of the stuff that I've seen follows that. It's are we seeing ROI for our investment? Are we getting value out of it? And then build versus buy is a very big one here as well. And then what metrics are most important in those conversations. So anything that helps tell that ROI story, uh, I think is pretty critical here. And then I think showing your wins as well, showing showing that low-hanging fruit, showing those wins, and then just helping to educate and get that that message out there also helps build trust. Where I think that has been at least in the past a challenge, and that's kind of where all the experimenting was. But once you got agentic and you had open claw, now all this stuff can do work. Now it is all right, we've provided the investment with our LI. Where's our return?

SPEAKER_00 33:53

So I think you mentioned what the RI is, right? So what decisions now remain human-controlled? High risk, I guess, is the biggest one. Like you mentioned before earlier in the in the in the episode, high risk is one.

SPEAKER_02 34:11

I think this is a good question.

SPEAKER_00 34:14

What applies to business, what doesn't apply to business, start roadmap, things like that.

SPEAKER_02 34:18

I think I think outcomes, priorities, anything that's inherently risky, just off the top, those are probably the main ones. I think there's still going to be human decisions around architecture, at least for a while. Um should should we build it this way? Um, I think the trick around that is with architecture, we try and develop building blocks or patterns for how we should do things. And then usually those patterns of those building blocks are from different parts components, they're new. And then we try and figure out how we can add building blocks together to create something new. So I think there might be some of that. If it's something that is new, you don't have trained data. I think depending on how good inference gets, there'll be models that will look and we'll try and build put those building blocks together, but it's gonna be a human that would interpret did it do what it was supposed to, is the outcome correct? So I think there's always going to be some kind of collaboration or partnership, especially where dollars are involved.

SPEAKER_00 35:25

But you're but the good thing is you're giving them a process. You're not saying, you know, um, you got a computer, bring your own tools. I don't care what software you use, I don't care how you save it, I don't care, you know, you're not doing that. You're setting them in an environment, in a contained environment. It's almost like it's I can't say that. I almost say to say it's like like a playground, but you know, it's it's it's it's protected to the most degree, and and you're governing it. It's not going to get out the door completely before you see it, right? So you're giving them the tools, you're giving them the platform, you're giving them, you're teaching them, you're training them, educate them how to prompt, how to, how to your AI engines that you brought in are actually developing these apps and and you're governing it how it gets out the door. So I think it's safe that way, right? It's not like uh the Wild West, but um I don't know.

SPEAKER_02 36:22

It's yeah, it's gotta follow guardrails, essentially. Yeah.

SPEAKER_00 36:26

Yeah, you just can't let them loose in, you know, and say, go go do what you want to do and tell me how it worked. You know, that's not the way this works. You wouldn't do that with any part of your technology.

SPEAKER_02 36:36

How do workforce roles change when people move from doing work to supervising work? Is everybody a manager now?

SPEAKER_00 36:44

Isn't everybody isn't isn't that the rule of IT now today? Hasn't it changed already? I mean, is your IT tech of 1975 the same as 1990, or the same as now? I mean, have they evolved to where you have to be a little bit more business savvy, you've got to be able to communicate, you gotta have empathy. We've talked about this I think in previous episodes. You know, that your your technician isn't the geek in the corner that you feed a Mountain Dew and chocolate bars and they can't leave the room and they're sitting there for 12 hours. You know, that's not the life today. So I I think your workforce changes only in, and I think you're used to it now. You're we're used to change, right? We're continually moving from we had physical computers, then we had virtual computers, now we have cloud computing. I mean, we're kind of like used to changing, and I think you just gotta continually change.

SPEAKER_02 37:40

And I mean, not the most apples to apples example, but we have calculators, we still have math teachers.

SPEAKER_00 37:50

Yeah.

SPEAKER_02 37:51

We we have tools, we have a new tool that helps with orchestration, helps with automation, helps with planning. So tool and toolbox in that case. And so for folks who are getting through the work day, it is now a new tool in your toolbox to do something quicker, to automate something, and maybe it lowers that barrier where that knowledge gap and that skill gap was preventing you from achieving that outcome. So I think that would be the biggest thing.

SPEAKER_00 38:25

And language art becomes more important, you know, you have to be able to convey what you want and you know, and and and are looking for in a sense too. So what what should teams do now to prepare to prepare? Is it is it tooling?

SPEAKER_02 38:47

I hope it is already in flight for teams. Whatever it is, already in flight. So essentially, whatever platform, whatever agents, whatever models, whatever tools, I mean, how did we prepare for um I don't know, getting bruises on a playground? Go experiment. Nothing kind of prepares you for the bumps and bruises. There are a lot of training and knowledge resources out there, depending on your core business platform. ServiceNow, for example, has now university. I think Salesforce has, I think it's Trailblazers or something. You have knowledge out there that will teach you not only about the platforms, but about how those platforms interact with the AI, what their AI roadmap is, their strategy, and even from a platform agnostic standpoint, what AI agent capabilities are out there, what tools, applications allow you to use them. You think of ChatGPT, you think of Cloud, Perplexity. There's a whole plethora of different clients out there, and then models. Each of those will let you choose different models and just getting out there, playing with them. Pick something simple. If you're a planner, a thinker, uh, an accountant, ask it accountant questions. Then take it a next step further, ask it planning questions, then take it a next step further, say, how could I do X, Y, and Z? And you have to experiment, you have to build that muscle, and then just find learning resources, YouTube, if nothing else, or listen to a podcast.

SPEAKER_00 40:26

Yeah, yeah. I mean, do you pay that? Yeah, and there's also like user groups. What's the one that Matt Gosson's part of with the high schoolers? What's that one?

SPEAKER_02 40:36

Okay, so uh I think that's Interalliance.

SPEAKER_00 40:40

Interalliance. So if you're young and still in school, you know, there's probably groups in your school district or in your region, something like Interalliance, where it takes high school students and and indoctrinates them or brings them in, you know, into businesses. Or there's ones like technology first. We had Andy Hickey on talking about technology first, and whereas a lot of I a lot of IT groups that are sitting there talking and building groups, just so you can sit and talk about the issues together. You know, not just you're not in it alone. So find a vertical, uh like if you're in insurance, try an insurance vertical. If you're in legal, try a legal vertical, but find some chat room of group of users that you trust and talk about these issues and talk about what they're doing, see what they're what they're accomplishing. I think that's a value as well, more than just buying a 99%, 99 cent app on the Apple store and trusting that what they're telling you is the truth, or we'll take you anywhere.

SPEAKER_02 41:38

Yeah, go uh go compete in the tech Olympics. I think that's also a high school thing too.

SPEAKER_00 41:44

Yeah, yep. All right, let's finish this off here.

SPEAKER_02 41:48

All right. So the future isn't about choosing between IT and the business or between people and AI. It's about designing a model where the right personas, human and machine, can build, govern, and run work together with clarity and control.

SPEAKER_00 42:05

Yeah, so the future of delivery is it's not just faster, it's more distributed, it's more governed and more intelligent. So thanks for tuning in. That that's the real opportunity. That's not just more about output, but better ownership, better guardrails. Thanks for listening. If this has sparked some ideas of your own model, it's time to fire it up. Let us know, talk to us about it, send us some messages, subscribe. Thanks for joining.

SPEAKER_02 42:32

Cheers.