AI Agents Are Replacing IT Workflows — Not People. Here's the Difference

s1e31 AI Agents Are Replacing IT Workflows — Not People. Here's the Difference

Is your IT operation ready for AI agents? Not chatbots. Not automation 
rules. Actual agents that observe, plan, and act on their own.

In this episode of The Wired Garage with Pops, Steele and Pops go 
under the hood of Agentic AI in ITSM — breaking down how AI agents 
are changing incident triage, change management, and CMDB health in 
real IT environments.

🔧 What we cover:
- Why traditional ITSM automation is getting brittle
- How agentic AI thinks vs. how old automation rules work
- LIVE demo inside ServiceNow AI Agent Studio
- 3 real use cases: Incident Triage, Change Co-Pilot & CMDB Health
- Guardrails, blast radius, and why autonomy has to be earned
- The metrics that actually tell you if agents are working
- What executives are asking — and what IT leaders need to answer

💬 The quote that says it all:
"If you thought a bad process could run fast when automated... 
wait until you add AI on top of that."

Whether you're in IT ops, running ServiceNow, leading a team, or just 
trying to understand where AI is actually landing in the real world — 
this one's for you.

🔔 Subscribe for more no-hype tech talk from the garage.

📩 Got a story or question? Send it our way.

⏱️ Chapters:
- Welcome to The Wired Garage
- What Is ITSM Really?
- Why ITSM Still Struggles
- Why Old Automation Gets Brittle
- Agentic AI: Observe, Plan, Act
- Guardrails: Data, Scope & Autonomy
- Recommendation → Supervised → Autonomous
- Controlling the Blast Radius
- Specialized Agents vs. One Superstar
- Use Case #1: Agentic Incident Triage
- LIVE DEMO: AI Agent Studio in ServiceNow
- Use Case #2: Change Management Co-Pilot
- Use Case #3: Agentic CMDB Health
- Measuring CMDB Health Before & After
- How Executives Are Responding
- Metrics That Actually Matter
- Pitfalls & Guardrail Failures
- How to Prepare Your Team
- Rapid Fire Takeaways

#AgenticAI #ITSM #ServiceNow #ITAutomation #AIAgents #CMDB 
#ChangeManagement #ITLeadership #WiredGarageWithPops #ArtificialIntelligence

Transcript

SPEAKER_03 0:00

IT used to fix what was broken. That's how we made our careers, right? We waited for the call. We probably hated the call. Call user came at 3 a.m. But that's how we built our careers. Everyone needed us to help fix things. Now it's shifting toward a world where IT prevents problems before people even notice. Right? So the question is less about if this will happen. It's more about how fast companies can adapt and prevent these things from happening. And how many things can you prevent?

SPEAKER_01 1:18

The show where we pull back the hood on technology and figure out what's really going on under the engine. I'm your host, Steele, and as always, Pops is here with me, keeping it real. Today we're going under the hood of ITSM and AgenTic AI. Not the hype, but how these agents actually plug into incidents, changes, and your CMDB. We'll talk about some use cases. If you build or run ITSM, this one's for you. Whether you're an IT running a business or someone who just wants to understand where all this AI talk is actually landing in the real world, pull up a chair, grab your coffee, and let's get into it.

SPEAKER_03 1:55

Yeah, most people think ITSM is just about the help desk, like reset my password, fix my laptop. But IT service management is way bigger than that. It's the playbook for how the entire IT operation works. So it is now agenticai is starting to rewrite that playbook.

SPEAKER_01 2:13

And what we're going to talk about is how the agents are interacting with tickets, with workflows, with CMDB, where they help and where they can blow things up, and what guardrails you need. An example would be when an employee joins your company, they request a laptop through a portal. Behind the scenes, this kicks off a repeatable, automated process. Requests are logged, routed, prioritized, unfulfilled, all tracked for speed and quality.

SPEAKER_03 2:43

Yeah, so when you say ITSM runs this show, we're talking about everything from employee onboarding and hardware management to major outage response. So, Steele, why do you think so many companies still struggle to make ITSM efficient when these frameworks have existed for so long?

SPEAKER_01 3:00

We don't have enough time to answer that one. I I think uh like anything, people process technology and there is a maturation curve in every organization and maturity is different along different processes. And depending on the health of your CMDB, the health of your processes, when your last re-org was, that all creates complexity in managing your portfolio services. And if you are light in your CMDB space, if you are light in your service portfolio management space, if your guardrails are just how we feel on any given day, those are all things that are going to cause challenges when it comes to standardizing the services that your company has to manage. And to add complexity to that, we are now taking AI and putting it on top of all of that. So if you thought it uh a bad process could run fast when it was automated, wait until you add IT on top of that.

SPEAKER_03 4:07

Yes, I you know, you've bad data feeds bad data. So traditional ITSM automation is usually pretty static. So you've got hard-coded assignment rules, category, long, sometimes long, lengthy, too lengthy sometimes, category and subcategory trees. Maybe some scripted logic, it works until the till the environment drifts away and just and changes, or the business changes, and then the rules get they get brittle. So the gentic AI, I think, is gonna change all that, right? It's gonna help with evolving. Yes.

SPEAKER_01 4:40

That's where the agents hopefully well define as long as we well define these rules, but that's where the agents will come in and provide a different kind of automation where the rules are set and it's not just facts, then why? Rather, here's the goal, come up with a plan, and then execute on that. And then we have this feedback loop of continuous improvements to make sure that the agent is doing what's supposed to do and we're hitting the outcomes that we're supposed to hit.

SPEAKER_03 5:10

Okay, so here's where things get exciting, right? Agentic AI is changing how we think about automation and everything. Now ITSM is not just a smarter chatbot or a workflow rule. We're talking about systems that can observe, plan, and act in a loop. They see data like tickets, logs, CD, CMDB relationships, things like that. They form a plan and they execute actions within Guard Rails.

SPEAKER_01 5:36

So instead of that static workflow saying on category X throughout team Y, the agent's goal is something like restore service fast within policy. If it can read the ticket, look at monitoring alerts, look at CI relationships, and then determine how best to approach the problem based off of that information that's being provided to it. Agentic AI systems are designed to observe, plan, reason, and act. They interact with their environment and make decisions on their own within those guardrails that you set up. So if you imagine an AI agent monitoring your company systems 24-7, it doesn't just alert a human when something goes wrong, it takes that proactive step of diagnosing the problem, checking on change data. Did a change go off that caused this issue, kicking off an approved remediation plan, and then taking action. And then depending on what it's doing, assuming you have given it some autonomy, it can then go take action without having to wake up your folks at three in the morning.

SPEAKER_03 6:41

Which is the benefit, right? That's the benefit. Yeah. So the attack, the data service is huge. Incident text attachments, all these monitoring events, chatter, whatever else of these equipment's doing, change history, CI criticality, ownership, previous resolutions, right or wrong, right? Whether they really resolved it or not. The agent needs at least read access across that to be useful.

SPEAKER_01 7:10

Yeah. And I kind of to the point of read access, that's where we start getting into guardrails. And I expect that this will mature, this will evolve over time. But it's not too dissimilar from what we have today, especially regarding RBAC access, rule-based access controls. What data does the agent have access to that it's using to build its context, that it's using to execute on its plans? What is the scope of the agent? What can it recommend? Is it appropriate what it is recommending? Uh, and then what actions or responsibility, what actions can it take based off of the responsibilities that it has? What can it actually execute? And then how much of that can it execute without a human being in the loop or somebody hitting the approve button to go run it? How much trust do we have in that system? So when we think about GargayL's data access, what scope uh of responsibilities the agent has, and then what actions is it allowed to take, both uh supervised and autonomous are some of the key points there.

SPEAKER_03 8:17

Yeah. So unlike what traditional automation, which follows fixed scripts, workflows, things like that, these agents can learn from outcomes, correlate data across systems, and handle multiple steps in real time. So the payoff is massive. Early adopters are already reporting what a 60% reduction in service ticket volume. So you almost never want to jump straight to a full autonomy in production, though. You start with a recommendation mode, you do a supervised execution, maybe uh limited autonomous actions for well-understood responses and playbooks. Wouldn't you agree with that? I mean, you got to baby step this thing a little bit, even if anything, to build trust with your routine.

SPEAKER_01 9:01

Yeah, it's building that trust and putting a limit on that blast radius, controlling that blast radius. And just like anything else, you have to prove out the concept, make sure it works, and then make sure that where you have put AI in the loop, that it's proving out that value and it's achieving those outcomes that you're looking for. And it's not just one more thing that an associate is gonna have to manage. So it's rarely one big ICSM brain. Uh, you'll typically see specialized agents that perform probably one task, maybe two tasks at most. And that has to deal with the scope. A well-scoped, well-defined agent is gonna be pretty successful at its rollout, what it's trying to do, whether that's triaging uh an incident, checking on change risk, or checking on seem to be health. When you start trying to have an agent do a bunch of different things or a bunch of very disparate things, that's where it starts falling apart and where it can start getting really creative. And we don't want the agents to get too creative. So it's very important that they're specialized. And then you also get into the concept of agent ops. So teams that manage and tune these agents the way SREs uh tune SLOs and alerts. Really interesting landscape.

SPEAKER_03 10:22

So would you say instead of having a one superstar, you know, trying to be trying to do everything at once on a team or whatever, not he's on lean on the team, just one single superstar, he can't do everything, right? Can't play all, can't play defense all by himself against the another team, can't do all those things well. But if you have five people who are distinctly responsible for their area of the court in a sense, and they play well together, hand off together, communicate together, you're actually more successful as a team. Same thing here, right? You have these agents, and almost like you're you're also biting one elephant and the time's at a boil in the ocean. You're you're taking it, you're learning it. You're not only learning, it's not only learning what to do right, you're learning how it can do things right and how you can teach it to do things right. There's a lot of teaching, learning going on, and trust going on. So then you can build others and have these things work together. Next thing you do, you have a system. Not unlike your IT team, right? Your IT, you don't have your storage guy all the time also doing your desktop push, also doing your help desk, also doing you have teams of different people, infrastructure, network, storage. You have different people doing different things. And I think that's the same way you would treat these agents. Am I wrong in that or am I not? That's kind of the same way.

SPEAKER_01 11:37

That's how I think of it. And then if something does go wrong, it's far easier to troubleshoot that specific agent and correct that specific issue when that scope is very clear on what that agent should be doing.

SPEAKER_03 11:49

You see gain right away. You get sort of seeing some positive return right away instead of waiting for a long period of time of teaching and testing and evolution. So this is kind of wild. So IT used to fix what was broken. That's how we made our careers, right? We waited for the call. We probably hated the call. Call user came at 3 a.m. But that's how we built our careers. Everyone needed us to help fix things. Now it's shifting toward a world where IT prevents problems before people even notice. Right? So the question is less about if this will happen. It's more about how fast companies can adapt and prevent these things from happening. And how many things can you prevent? And so let's make this concrete. Let's hit a few use cases and let's talk through what these agents are actually doing. We're gonna start something new here with all of our episodes here. We're gonna call this under the hood. And so we're gonna talk about something, a use case, and then we're gonna dig into a small little demo per se, right? And we're gonna get a little, we're gonna get our hands greasy a little bit, we're gonna get under the hood on this one, okay? So, first one, age agentic incident triage. Classic story. Two POS incidents come in from uh we'll say stores in 15 minutes, all worded slightly different. Traditional ITSM sees 200 tickets. Maybe some rules route them to the same team, but it's still a mess.

SPEAKER_01 13:14

And if you add an agent in front of that, that could be the initial triage for those tickets, read the air codes, read the logs, look at the screenshots, understand what is happening. It could under that those 200 incidents, even though they're they're scattered, that they're actually related. And it could spin up a major incident, link the rest of those incidents as children to that. And that is just the amount of time that that would take for disparate teams to figure that out. That is just an incredible amount of time saved. So the use case that I'm gonna take us through, and I'm I'm gonna demo for us on the ServiceNow platform. We're gonna go through uh a canned incident that I have about a laptop battery replacement scenario. But I think it's really gonna start painting the picture, even though it's a simplistic example of just the power of what these agents can do with very minimal prompting and very minimal tool use. So in the platform, this capability is called AI Agent Studio. And in AI Agent Studio, we have use cases. Think of these as workflows, and then we have the AI agents. Think of these as all the required team members to make a use case work. And the use case or the workflow that we're gonna look at today is called steps for issue resolution, depending on how up to date or how new your instance is. If you're using a company provided one versus a learning instance, which I am here, this may look a little different for you. But this is just a good example of how to get your hands dirty with AI Agent Studio inside of ServiceNow. Or using an agent to solve ITSM issues. So this workflow is going to help us build a plan for resolving a specific issue. This incident uh about needing a laptop battery replaced.

SPEAKER_03 15:36

So what I'm gonna do is I'm gonna go to So put a use case in here real quick. So let's say this laptop battery, right? So often a lot of us have hardware where we buy a hundred of these units and they get this dispersed to different offices, but it's a bad batch. It could be now maybe laptop batteries, maybe it's a motherboard, right? Or something like that. Something that's similar. Something that's similar, right? And I think this is those use cases here where you really don't want 10, 15, 20 help desk agents taking in 200 calls about these 200 laptop batteries that need to replace. You would rather this agent seek this out as these things are coming in pretty quickly, or even say, wait a minute, these are all part of a certain batch. And let's it's pre it's preemptively solved this while these other, while they're our people are doing bigger and better things, right? So I think this is what you're trying to show.

SPEAKER_01 16:32

Yeah, exactly. And you'll have to forgive me because the this is going to be a very simplistic example, but you've hit right on the point of this. If you are able to set up an agent to handle a use case like that, it's crazy return on investment and time.

SPEAKER_00 16:52

And then if that hits at a very unfortunate time of the night, you don't have to hit your team members who are trying to get sleep so they can help deal with this in the morning. In the agent, we have a couple tools that are associated to it.

SPEAKER_01 17:12

It will look at similar incidents, it will get the details of those incidents. There's other tools that are available to it. And I will, once I go through the example and I show you the tests, we'll talk more through those. And then it will also look at data that's on the platform in terms of knowledge articles. So if you've gone through and you've documented how do I solve this issue, the agent's gonna find that information. So let's just walk through a test here.

SPEAKER_03 17:40

What if you had like um LOMs or SLMs out there from manufacturers like Dell or somebody else? Is is that something that can happen or not? Is that is that not, you know, are there data sources like that that are made available that can be searched or seen?

SPEAKER_01 17:59

The possibility for that, there is an AI search capability uh in this platform, and you can connect that to external sources. So if you have the data that's in internal, or if you connect to an external source and you index some of that, um, that is a possibility.

SPEAKER_03 18:20

You'd have to index it and have it under your umbrella in a sense on already for this to work, right?

SPEAKER_01 18:26

I would say yes. Now there is a tool that allows you to do a search on the web. So there are some use cases where you can get external, truly external to the platform and search and bring information in. But if you're using the external content plugin for AI search, it will index just a very brief minimal amount of that information. But it's an amazing capability and it'll be something that yeah, will help with that use case. So in this example, I'm gonna say help me resolve, and then this incident number here. If you wanted to just test just an AI agent by itself, you have the capability of doing that. And then this interface here is very similar to virtual agents or that chat interface. If you've seen that on ServiceNow, it's how you can interact back and forth with this agent as you're going through and you're testing through a scenario. You have this graphical interface, which will show the different agents that are connected to this use case. And then when those agents leverage and use tools, you'll see that as well. And then this is probably one of my favorite things about this interface is you can see all the reasoning, all the logic, all the information that was involved that this agent is using to be able to make its decisions, to take action. And we can see that it provided a resolution plan and it also linked to its sources too. So there's a knowledge article that details hey, if you have uh battery failure, here's what you need to do. My motherboard did fail, and I had to reach out to uh that manufacturer to have a tech come out and fix it and replace it, which is cool.

SPEAKER_00 20:19

I didn't have to go anywhere.

SPEAKER_01 20:21

But the next step, or how you would take this one step further, is to add a tool that would save this to the notes and then close the ticket. And I've seen use cases like that, and there is one that is coming out in a couple of months, probably around May. I think it's a L1 service desk agent that has a bunch of pre-built capabilities for solving issues. I think of this as it's not necessarily a tier one, but a tier zero. So having that first touch to help you solve those menial trivial issues like your password resets, like your VPN issues, so that your agents can then focus on the harder things where we don't have where it's not a good fit for an agentic agent, or we just don't have the capability built yet. Massive, massive value here. Uh hopefully this does not date me, but this is something that I wish I had back when I was on the desk. The last thing that I want to mention is that if you can't measure it, you can't manage it, it's also really hard to have that conversation about value. Out of the box um with AI Agent Studio, there is a dashboard that helps you track those metrics. You can also create your own, but there's a bunch that are are pre built out of the box more than you're seeing here. And if you have performance analytics slash platform analytics, but I think you need performance analytics, uh As well. There's actually more on this dashboard. This is a very, very large dashboard when you have those capabilities on. So really cool stuff. A great way to kind of get your hands dirty, test some things out, do the POCs, prove those use cases, and then also be able to prove the value to you. So great.

SPEAKER_03 22:22

Nice demo, Steel. Good job.

SPEAKER_01 22:24

Let's get back into those use cases, shall we?

SPEAKER_03 22:26

I think change management, right? Yes.

SPEAKER_01 22:30

Yep. So this second use case of change management kind of naturally is the next progression for triaging those ITSM incidents. So if you are, if you could imagine an agent kind of sitting next to your change manager, but it does all the boring pattern matching, that's where this use case is going to fall into place. Where if you've got an emergency change that's coming up, if there's a conflict that's coming up or something that's super risky, I think this would also be a great use case for an agent.

SPEAKER_03 23:08

Yeah. So yeah, an engineer, say they want to want to roll out a change. So instead of them manually filling out all the different, I call U16 white form boxes, you know, on a form, the the agent can can draft the change, description, affected CIs, the risk score, an implementation plan, and a test plan based on similar past changes.

SPEAKER_01 23:31

Yeah, it can look at CI criticality, dependencies on the CMDB, affected CIs, historical change failures, blackout calendars. And then it can propose is it medium risk? Here's a safe window to make the change, here are the approvers that you need. Really getting the value out of that foundational work and that foundational investment that you made in shorting up your CMDB and your processing.

SPEAKER_03 23:57

There's what it is. You had it the key to AI is you don't just turn it on and it just perfect runs. There's other things in your environment you have to do and manage and CMDB and having that correct to have to have in there what are the blackout calendars, to have that information, you know, in there and what systems are are linked to what, what apps sit on what servers, so you know when this server is going down, there's a black, you know, blah, blah, blah. So I think I think those you hit around the head. We have to maintain the environment still. So you keep the humans in a loop. The agent can't push a like a high risk change without approvals, and it removes a ton of friction and inconsistencies from like a low and medium risk changes and enforces all of your your uh cab rules every single time. So we have a third one, and I think this is one of your favorites, right? Agentic CMDB health. Think of an AI CMDB caretaker that never sleeps. For me, it's like we're talking about CMDB. Who's gonna take care of this for me? I can't hire somebody to do that. So maybe I'll bring on me a little an AI caretaker who doesn't have to go sleep or doesn't have to go buy Mountain Dew or take a smoke break or whatever those things are, right? Doesn't have to go on vacation, you know, up north, or or what is it you did? You go on snorkeling and whatever else you did, right?

SPEAKER_01 25:26

Night, night snorkeling. I think that eventually after enough token usage, we'll see AI agents taking smoke breaks.

SPEAKER_03 25:33

Yeah, you're right.

SPEAKER_01 25:34

This won't be the kind that we're thinking of. Yeah. The data centers will start smoking. So having an agent that is able to look at that discovery data, look at tickets, monitoring those usage patterns, and based off of data policies in the system and also that data, hey, these 120 CIs look wrong. The naming conventions aren't there, they don't follow what we have set in our knowledge base of how we manage CIs. Relationships don't seem correct.

SPEAKER_03 26:03

They're manually entered, something like that, right?

SPEAKER_01 26:06

Or data is just missing that we need having an agent be able to go out and triage that and solve the problem instead of somebody coming into the system and then having that bad experience of, hey, my data is not there. And I think that's how you start essentially designing magic into the system is when all those pain points that used to be something manual we had to throw somebody at that nobody really had time for in the first place. This is an amazing solution where that is that pattern of we don't have the resources for it, but we know it needs to get done anyways.

SPEAKER_03 26:43

So I always wrote dashboard reports, right? Looking for empty fields, right? Tell me all the records that have these fields null, right? And that makes me, though, have to go into each one and go figure out why. And I still have to pointed it out which ones they were, but there's still a lot of work there. So technically, you can have this agent compare what's in the CMDB against like multiple sources of truth, discovery tools, cloud APIs, even incident history. And when it sees like a systematic drift of a whole class of servers getting misclassified, it can flag that and suggest the bulk correction saves me a whole lot of time.

SPEAKER_01 27:25

I mean, it sounds like magic when you say it that way, and it sounds like love to me. Yeah. You'd still so I mean you've got your seem to be health dashboard, and I just imagine that something flags in that system. Uh, it goes from your 80%, let's say down to 79%, or something comes in. That agent being able to take that first pass, look at those data sources, try and solve what the issue is, and then either rerun discovery, create a discovery schedule, bring that in is mind-blowing to me. Because it's C2B is a really, really big space. The architectures are all over the place. You have cloud, on-prem, private cloud, all these different systems. And to have something that makes that experience feel like magic, that I will be able to go snorkeling again. I'll I'll be able to get sleep. The agent will have my back. And I think there is a little bit of nuance there. Uh, because we talked about blast radius, we talked about guardrails. So maybe there's a layer where you have instead of the agent trying to do all the things, the agent can fix. Maybe we just pointed out the attributes. Hey, if if this attribute is blank, your job is just to make sure this attribute does not get blank. And if it is, go and fix it. Like owners or CPU is missing, etc.

SPEAKER_03 28:53

Yep. So so now you've got the agents clustering incidents, traffic changes, constantly poking at the CMDB quality. So how do you know it's actually helping and not just adding more noise?

SPEAKER_01 29:05

And I think that's where you have your metrics defined of these are the things that we look at. In the context of CMDB, you have a CMDB health dashboard. You have metrics that are associated to that correctness, completeness, compliance. What are those metrics? Are we hitting the targets that we've set? And then that we have essentially that before state. And then when we turn the agent on, we have an after state. And we we can measure, but we can also test specific scenarios too. So it's not that we're in the dark and let's turn agent on and hope it works. We can set up a specific scenario and then see a before and after.

SPEAKER_03 29:42

Okay, Steele. So we're we're rounding the bend here. We're in what we now in our new terms of our sections of our show, the tune-up area, right? Before we close. So this section we're going to do is how our executives responding to this. So they're not they're not asking whether AI can do more. They know it can. They're asking how fast. They're asking how safely can it do it, and who's responsible when it acts? Because when it does go haywire, they want to know who to call. Right. Um, they also have executives remotely responsible, responding with a mix of enthusiasm and caution. They see agentic AI as a boardroom level opportunity, right? They see, they see it as an opportunity that gives them an advantage, but they also need governance, they need clear decision boundaries, and they need to measure the ROI before they can let it scale. And in practice, there's there's got to be leaders who are focusing on these three things: cutting friction in the operation, redesign your workflows around these AI agents, and then build supervision into the model so humans still can control the high-risk decisions.

SPEAKER_01 30:51

In ITSM, productivity isn't only about how many tickets did we touch. It's did we resolve the right work faster with fewer airs and then fewer or less toll on humans?

SPEAKER_03 31:06

Exactly.

SPEAKER_01 31:06

So for agents, I would track things like automation rate on eligible work, reduction in ticket volume or duplicate tickets, MTTR, of course, and then also how often are those tickets reopened, reopened. So are they our agents doing the job? Change failure rate, and then we talked about some accuracy signals in the CMDB. So better relationships, fewer unknowns.

SPEAKER_03 31:31

Yeah, so big pitfalls, turning agents loose without any guardrails. We talked about some guardrails here, and we should probably have a session just talking about guardrails. Just there's so many of them to think about that people need to at least have cons at least be knowledgeable about. No observability into what they're doing, optimizing only for speed. You can absolutely create a very fast system that closes tickets with bad fixes and quietly increase the actual risk. Especially security risk.

SPEAKER_01 32:02

Yeah. And if you're only if your only metric is time to close, I'm sure the agent will happily slam the door on users. Exactly. You you have to balance speed with quality and blast radius or security.

unknown 32:16

Yeah.

SPEAKER_03 32:17

So bottom line, IETSM gives you structure, defined processes, data models, governance. Agentic AI gives you the adaptive execution, adaptive execution. Systems that can wash, decide, and act. The win is when you design your operations so humans and agents are both first-class operators.

SPEAKER_01 32:36

So, what can you do to prepare? I think about upskilling your teams, educating your executive team on the risks, on benefits and limitations, doing some demos, showing them how this works in a real-life scenario, immerse in the tooling and experiment. A competitive advantage is emerging, and that is that the power of is really in the hands of the problem holders, i.e., problem solvers. So the folks that are in the trenches are having to solve these problems on a daily basis, they are the driving force for the folks who are going to be able to help prioritize. Hey, we're seeing this all the time. This is what we need to go after and solve with agents. So I would say that, yeah, to your point, ITSM gives you that structure and agentic AI gives you that speed.

SPEAKER_03 33:30

So in close, we're going to go real back and forth just real quick, and we'll close here. So if you're experimenting with this in your own environment, especially around incident triage or change copilots or even your seam to be health, we'd love to hear what's working and what's scary to you. So send us your stories and your questions.

SPEAKER_01 33:47

Audit one ITSN process this week. Pick your messiest ticket category and ask, could an agent handle the triage here?

SPEAKER_03 33:55

Or identify one boring but critical workflow. Change a drafting of it, incident grouping, seem to be cleanup, and prototype an agent assist on it before going autonomous.

SPEAKER_01 34:05

And start tracking metrics, automation rate, duplicate ticket rate, TMDB accuracy. You can't improve what you're not measuring.

SPEAKER_03 34:14

Before your next sprint or plan cycle, pick one workflow, just one, and ask three questions. What does it need? What decisions does it make? And what's the blast radius if it gets it wrong? That's your agent readiness scorecard.

SPEAKER_01 34:26

So we appreciate you hanging out with us today. Thanks for listening to the Wired Garage with Pops. We'll catch you on the next one.