SAP Joule, Veeam DataAI, Red Hat Ansible AI Agents

Hey everybody, welcome to the Tech Drawing Gang for Wednesday. It's the middle of the week. Some folks call it hump day, but it's also the day after a whole lot of conferences this week, and it's that time of year and that season, but we have had events involving SAP in Florida, Veeam in New York, and Red Hat I think was in Atlanta this week.

And they all had something to say about, surprisingly, AI. And a theme is starting to emerge here, but I'm going to see if we can get into this or pull it out with our guests, who are, of course, Jon Swartz is joining us. John, good to see you.

Good to see you. Hi, everybody. Jack Gold, it's been a little bit of a dog's age since I saw you last, but Jack, how you doing?

I'm doing great, Mike. You look the same. Well, I hope that's good.

Andy, good to see you as always. Always a pleasure, Mike. All right.

And Teri Robinson joining us from New York, where I am as well. Yes. One of these days I'm going to go down to Manhattan and see Terry.

Yeah. Maybe we should do this together sometime. We could- Yeah, that'd be fun ...

share a frame. Yeah. All right.

" And I guess there's some truth to that, but I'm scratching my head a little bit about, well, is that all? Is that the issue here? Because to me it feels like a lot of this is that the AI is probabilistic and the business workflows are deterministic, and the two don't always meet in the line very well.

But John, what's your take on what's going on with SAP? They are still the dominant provider of ERP class software out there- Yeah ... so corporate America does depend on them.

They seem to be working overtime to show that they're relevant during this era. Is it my imagination or every few days they have some sort of announcement where they try to pivot towards why they're important, why they're essential in this new AI world. So in a sense, they're kind of remaking or reframing the company.

And I think what is happening and what they're trying to say is that enterprise AI is starting to look less like a model race and more like a control plane fight. So they're positioning ERP as the brain or the trusted system of execution for agentic AI. So their argument is their AI agents are only as good as the structured proprietary data they can access.

And this Sapphire conference, which I believe is in Orlando, and Adrian Bridgwater covered it for us, he was looking at how SAP is engineering their systems to work with external models and help prevent silos of AI. And what they profess is an open approach, acknowledges that while ERP is the heart of the business, the AI limb, so to speak, might come from elsewhere. And you referenced the CEO, is it Christopher Klein, I believe.

So he pointed out that businesses need higher reliability than typical consumer-grade AI. He says 80% accuracy is unacceptable for core enterprise workflows. So, and again, SAP, they pitched the Agent Hub as kind of like this central place to deploy and manage agents and orchestrate workflows that connect LLMs to enterprise business data.

In a sense, they're trying to lean on their strength to rationalize and pivot and position themselves as relevant in AI. And I'm not saying this as a criticism of them, but this kind of seems familiar to me. I'm not sure if you got the same drift from Red Hat, Mike, but it seems to me a lot of these companies, and I was just at the ServiceNow events in Las Vegas, they are kind of looking for a way to play into their strengths and explain to you and sell to you the idea of why they are relevant, and it leaves questions to be answered, but this is what they're doing, and I think they have a sense of urgency they have to do this.

I mean, Jack, what's your take on what's going on with SAP? Because it does feel, to John's point, like, well, they're just suffering maybe from a little AI envy. Jeez.

Model size doesn't matter, Mike. Well... Larger now.

Sorry. Well, look, they're coming from a position, in some extent, a position of strength. Because if you think about where the data actually resides in an organization, it's not in the LLMs that you're going to go license from OpenAI or Google or whatever, Perplexity, Anthropic, and so on.

The data, it's in how you've been running your business for the last 30, 40, 50 years, which is that's all in SAP stuff. And so they're positioning to say, "Okay, you may want to run AI. AI is going to be important.

All of that's great, but the only way it works for you guys is if you use the data that's already in our systems, and we're going to make it easy for you to get to that data, or we're going to try to make it easier for you to get that data," including access to the data, support of the data, securing the data, putting it into a knowledge graph that's applicable to whatever models you're running, building models on that data instead of trying to go out and hire or just buy an off-the-shelf model, which probably isn't going to do you that much good. And so that's really the underlying message that they're coming up with. It'sYeah.

All of these guys are big. OpenAI has raised gazillions of dollars, but at the end of the day, an off-the-shelf model isn't going to help you if you can't access your core data assets. I don't disagree with their core assessment about the value of the data, and they are the system of record, and it's the crucial data that the business kind of runs on.

But there's these things called systems of engagement, and that's where we collaborate and do things that are interesting, and that's where most of that AI action has been. And to be honest, for years SAP's been trying to be a player on the system engagement side with mixed success, maybe hardly any success, depending on how you want to look at it. And so to me, Andy, as you look at this, ultimately is SAP just, "Hey, there's a bunch of data that I'm going to invoke through an API, that I'm going to run through some sort of system of engagement application," but what's happening on the back end with the system of record is a database update, and so what?

Yeah. Look, SAP going to SAP. Get all the data, pretend like no one else exists.

At the top level, this makes sense to me. Jack's exactly right. Fifty years of enterprise data about purchasing, inventory, buying, selling, manufacturing, the whole kit and caboodle, literally.

So the data's there, it's informative, it's meaningful. You should be able to make something of that. But look, it's in SAP's DNA to want to hoover up all the data across the business, and then pretend like no one else has any data and they have everything.

And that's not going to work. They're going to have to change fundamentally in terms of how they exist in the world. I assume, by the way, they're going to want you to change your business process just to make their AI work as well.

Because business process re-engineering is really in their DNA also. So look, I actually think this makes a lot of sense, but you're right, Mike. This is back-end stuff.

This is system of record. I still think it's really meaningful. System of record informs what you've done in the past.

Sometimes you can use that to understand whether that was good action or bad action, and then you can novate that into a system of engagement and action, and do the right thing this time. Top line, yeah, I get it, but I don't think SAP's got it in its DNA to share their toys, and I think for that reason it's going to be a partial effort at best. I think there's a couple of things to unpack in there.

And I've never understood this either, but SAP has always argued that, well, we're going to create a packaged application that you don't need to customize because we studied everybody else's workflows and processes. And we determined exactly how this thing should optimally run, and you shouldn't dare customize this thing because we know best. And yet, Jack, I hear you laughing, everybody runs out and customizes this thing, and the whole point of AI, I think, is to create some sort of custom process that gives me some sort of competitive advantage versus me doing the same thing everybody else is doing the exact same way.

It's even worse than that. First of all, from an SAP perspective, I don't know of any customer at SAP that has ever taken an off-the-shelf SAP application. I don't think it exists.

So that's number one. Number two is agents are going to take over, and SAP is really scared about agents. Because if you think about what an agent can provide, you may still need the SAP back-end stuff, but the front end goes away, and that's really the sticky part for SAP.

It's not just the data. Sure, the data's important, as we talked about. But if I can deploy an agent that doesn't need to license 300 seats to purchasing, then that's a big revenue hit to SAP.

So they're going to have to change their model. Sorry, Terri, go ahead. They're going to have to change their model- Yeah ...

instead of just businesses changing it. I think you're really onto something there, and kind of look at SAP over the years and especially more recently with all the layoffs they've had. They certainly have been struggling to find a place or differentiate themselves.

I agree that some of this is a really good idea and the concepts are there, but I don't know. There's a whole wider ecosystem of data and applications that I think this doesn't take into account. Or I'm sure they've thought about it, but this is not part of the way they've announced and presented this.

Well, we talked about this a little bit last week when we were talking about SAP and their API strategy, and to Andy's point, they're trying to pretend that nobody else exists, and therefore you should pay a privilege to have your AI agent use their API to access your data, and that kind of creates this conflict in my mind. So let me get this straight. It's my data that I stuck in your software, but now I can't get to it without your permission is kind of frustrating.

But they're not the only ones who were saying that. Yes. I was going to say, Mike, my takeaway, there's this consistent narrative at these conferences.

They're customer conferences for the most part. So they're preaching to the base, the true believers, and it seems it's not just SAP, it seems like it's ServiceNow, SAS, Adobe, whomever, Red Hat, and they are clearly trying to appeal to what they have. It almost seems that they're trying to retain who they have.

They're afraid of losing their customers, so they're kind of grasping on. And in a sense, there's almost a sense that they're afraid of what AI represents, and they're trying their best to embrace it without it killing them or stabbing them. It seems like that's the takeaway.

There's almost this sense of desperation a little bit Maybe. I think that's super insightful, John. That makes me think that this is a land war.

This is a land grab. Everyone's trying to get theirs. Mm-hmm.

And traditional vendors across the board are concerned about how AI's going to cannibalize their business. Everyone from Google, obviously, cannibalizing Search, and so they put their AI results- Yeah ... at the top of the page now, right?

Everyone is bolting AI in because they're afraid that if you just give access to the data, you don't need me at all. And so that's actually really insightful. It is that land grab, and so I think SAP needs that.

They're trying to protect what they own. It is like a land war, Andy, you're right. It's like they're- Yeah ...

trying to hold on to what they have and entrench themselves because I think the possibility, and they know this, in the back of their minds is, "Oh my God, we are going to be going through a bunch of layoffs. " OpenAI has got into this enterprise consulting services industry with their latest move. Anthropic is a threat.

There's a lot of angst and anxiety, especially as the market watches them and looks at their results and wants to know how are you going to rationalize all this spending, whether infrastructure or whatever your product development, how are you going to rationalize that into actual revenue? There's- So I have definitely met in my lifetime, happy SAP customers, and what they're happy about mostly is it's a packaged set of applications that they didn't have to build, and they didn't have the skills to maintain it, and so they're running their business off of it. And I understand that, and it makes a lot of sense, but part of my soul also can't help but wonder if, I don't know, Jack or some of those customers suffering from some sense of Stockholm syndrome, and they're just kind of like- Yeah, hoping that nothing bad.

Sure. But let's also be realistic about this, right? If you're a big company, you're not going to pull SAP out of anything anytime soon.

You're running your business on it. You may experiment with an agent here or there. You may try to do something else.

Long term, maybe SAP becomes the next Oracle database that's kind of sitting in the background that you can replace with another database if you want. But at the end of the day, they're operational. They run your business.

And so it's going to be a long haul before they're pulled out. Now that doesn't mean they're not looking over their shoulder, of course. As John said, as Andy said, as Terry said, they all have issues with AI, but realistically, I don't see them going away anytime soon.

Yeah. Eventually. But eventually, maybe.

Or being smaller. Yeah. Mike, or those happy customers, they didn't have AI back when they signed on.

This is true, and it comes back to this phrase we keep hearing over and over again in the agentic AI era, it's headless. And people are starting to think about back-end services as headless services, and the AI agent is some natural language interface that I speak to, and then it goes and figures out where the data is and all the APIs are. And if that's the case, I may never need to look at another SAP screen again, not that they were pretty in the first place.

But Andy, what do you think? Yeah, look, I have actually had many happy SAP customers. I used to own the product line at a certain large enterprise vendor for the SAP support.

Lots and lots. And look, SAP absolutely provides value. They deliver.

And I do believe that all software that exists provides value in some way, even my competitors. But Terry's exactly right here. This is death by a thousand cuts, is where- Yeah ...

the potential exists. You're right, Jack. You're absolutely right.

Businesses are built on this, and especially that mid-size business that, to your point, Mike, doesn't have the ability to build, run their own software in the same way to be able to rely on that expertise. And yes, the business process re-engineering as well, that's been very positive. But I do still see that in the world of headless applications, and yeah, my customers are all over that, this becomes a very different value proposition.

This becomes more about data and data storage and data access and data throughput, data correlation, data aggregation. What it doesn't become is that front end. It doesn't become that smiley face that people log into.

The friendly, happy SAP logo on my screen at 8:30 in the morning when I come to work. And that loses business. That loses goodwill.

Over time, that loses functionality because when you don't see it in front of you, the functionality bleeds out as well. "Oh, I'm going to use this other tool. " And over time, does SAP maintain relevance?

I feel like it might not. You know, the Stockholm syndrome, can I ask you just a really quick thing? Yeah.

Stockholm syndrome, you're onto something because one of the things these companies do at these events is they cart out the customers, their partners from years ago, and it's almost like a nostalgia tour. "Remember how well we've worked together? " And in a sense, they're playing into the fact of being comfortable with the SAP product while they're retaining the customers.

But anyway, that's part of the dog and pony show. Right. Sure.

Yeah, I often wonder when those people show up at that conference to be part of that keynote, do SAP people show up at their conferences and also do that presentation on the other side? Yes, they do. You know?

Yeah. Anyway, but Terry, here's the funny thing about all of this in my warped mind. Okay.

" And they're kind of like suddenly selling a security vibe with a little bit of fear wrapped underneath it. So I can't help but wonder if I'm going to see all these application security vendors start to pitch this whole-Don't worry, I got you covered, and I'm Allstate. Well, perhaps so.

And when I was reading all of this, that's one thing that stood out for me, too. Now they are asking to, or saying that they're going to be the ones that protect your data. Can they do that?

Because I think with an approach like this, you're going to be increasing the attack vector, and I think there are going to be all sorts of issues. And yes, you're right about other companies coming up, but I just think from looking at SAP, there's going to be all sorts of security issues that arise from that. They're not a security company per se.

All right. I'm going to shift the gear here because, well, this is a natural break to what we're going to talk about next. But the second conference of the week was the folks at Veeam had a conference here in New York.

It's called VeeamON. And if you don't remember or know Veeam, they're the folks that came to prominence because, well, they mastered backup and recovery for VMware virtual machines. That was the thing that they rode up into some level of prominence.

ai, which is a provider of a data security posture management platform. Put a pin in that, because we're going to come back to that in a second. One of the things that they were talking about in New York is they've now developed a knowledge graph that provides all the connections and observability into where all your data assets are and their relationship to each other.

And they're saying that through this new platform that they launched, that you can apply governance policies to that data, including if it's being accessed by an AI agent. And what was interesting about this, and we talked about this briefly yesterday, is what they're really saying is that you shouldn't be thinking about putting the guardrails in and the security in for AI agents on the AI agent itself. Instead, we should finally go back and actually secure the damn data that we have in the first place that we never got around to securing ever.

So now it's just coming up in spades because AI agents are accessing data willy-nilly. So Terry, is there some irony in all this? And is data security the next thing?

Well, yeah. Plenty of irony. Now look, I love that whole promise of let's go back and secure the data, but I feel like I've been hearing that refrain for years and years and years, no matter what it refers to.

AI has created this really sense of urgency, though, about security. We've done all the stuff without guardrails to date. So it makes sense to go back to the data.

It really does. And you guys know I'm going to get a little cynical about stuff like this. Am I really hopeful that that's what's going...

I know. Am I really hopeful about that's what's going to happen? Mm-mm.

Not so much. But I think it's a noble play, if you will, to try to secure the data, do the data and AI trust infrastructure, like a new category here for us. I don't hold a lot of hope.

But Frank, it gets better. So they're also saying that as part of this, the cybersecurity teams and the IT operations teams will link arms and fight this fight together in a collaborative kind of way and will cooperate with each other now. I heard that part of the speech, and I think the first time I heard that it was 1995 and I was living in California, and it still hasn't happened.

But- And just about every major study that you see on this talks about that they need to come together, but they don't. They just don't. It's dogs and cats, guys.

Come on. Yeah. It's not going to happen anytime soon.

No. They're at each other, so they won't. But there is another piece to this that I think that they really didn't, in my opinion, hit hard enough, and that is as we move into an agentic world, what we really need to start thinking about is agent access management.

Just like we do for people, we're going to have to start doing that for agents as well. And so that's also about data protection, because Mike, I may give you access to my HR stuff. John, you might not get access to HR, but you may get to accounting.

Terry might get something else. Andy gets engineering. And so we're not seeing, in my opinion, enough emphasis on putting that identity management, that agent identity management play in place from a security perspective.

Right. Right. Andy, correct me if I'm wrong, but if I understand this as the way it's going to play out, let's say every employee has, I don't know, 10 AI agents that they're using.

I'm going to need a whole other army of AI agents to manage the AI agents that are performing various tasks. And so I'm going to have AI agents on AI agents because some of those AI agents will be looking at security issues and governance issues and observability issues and all the other things that IT people do every day that they probably don't get enough credit for over the years. But are we building layers and layers of castles here in the sky around AI agents?

Well, of course, we are. We've built layers and layers of humans, didn't we? Look, I'm just waiting for an AI agent to buck up and want a pay rise and have its supervisor offer a virtual pizza party instead.

I don't know. Yes. We're replacing all of the humanity with the AIs.

Yeah. Jack's spot on, by the way. This is going to be a whole new category.

Identity and access management for agents, agent access management, AAM, to replace IAM, I guess. But you're absolutely right, because we've got to do this. By the way, Veeam was a client of mine in a previous lifetime, 100 years ago when I was an analyst, and I was covering the virtualization revolution, and it was the same pattern.

We create the revolutionary technology, then we figure out how to manage it. VMware created this revolutionary technology, well, they adapted it, and then we had to figure out how to manage it. I had a line at that time.

First comes the revolution, then comes the management. This applies in geopolitics. This applies in technology.

We need the management. We need the identity and access management. We need performance management.

Like actual, not just speeds and feeds, but did you do the right thing? Did you meet your OKRs? All of this is going to happen.

And I actually think that Veeam is very smart to be there around this, because you mentioned trust and data trust. Look, backup and recovery... is data governance, which is data security.

This is a very straight line to me, and I think especially through this acquisition, that they have the opportunity to really be and provide this layer of data governance for AI that is the start of this management, normalization, standardization process. You're no longer looking for those magical blue sky moments, but looking for that activity, that work, that outcome, the output, the ROI, the boring metrics that we used to use before we used random users per day. Let's get back to that, and I think Veeam's in a really good position to establish that level of trust, management, security, governance, because honestly, it's what they've always done.

Here's where I get lost. I feel like we're all collectively looking at all these AI agents and building ones to replace tasks that humans do. And that's well-meaning, but I feel like every time we have a new innovation, in my experience, all we do is use it to do what we've always been doing, just slightly faster.

And I think we're suffering from maybe a lack of imagination. And Jack, won't all these AI agents just start performing these tasks in parallel and get blended with each other, and I may not be able to distinguish between what one AI agent is doing and where the other one picks up and leaves off? So, is the way we think about IT shifting from a sequential set of tasks managed by humans to parallel tasks managed by armies of AI agents?

Well, I think it's even worse than that, Mike. I think it gets to the point where we as humans all have our own workflows. Our best workflow is, mine is different perhaps than yours or John's, or Andy's or Terry's.

We do things a little differently. And so your agent may do something that is great for you but not optimum for me. Mm-hmm.

And so how do I personalize an agent for me? How is my concierge different than your concierge? And we're going to get to that point.

We're not quite there yet, but my clause is starting down that path, because I can build my own agent. We're going to get to a point where IT will probably have to build some universal agents for everybody. But at the end of the day, I need tools that are going to be able to build a workflow for me that makes sense for me.

Because we get to the same result, but I do things in a different way than you do, Mike, or you do, John, or Terry, or Andy does. Right. We're not there yet.

This is the whole model of... This is the Office model, right? " In the early days, it was all I'm an old guy.

When I had to build something, I'd write it on a piece of paper and gave it to an admin who would then go type it up for me. That was the thing. Who ever does that anymore?

Well, there are some execs who do, but we won't go there. But that's what we're going to have to get to. And so the real issue then becomes, how do I manage and govern those agents?

Because I may have 10, Andy may have 30, Mike, you may have six or seven, and they may all be at least slightly different, modified. That becomes a real long-term- There's no standardization, yeah. Nothing like that.

No. So where will that occur? Because as humans, I work with John.

If there's an issue, we'll go have a beer somewhere and sort it. Where do the AI agents go have a beer and sort it? Mm-hmm.

Well, we're going to go out and have a beer AI agent's coming pretty soon, so that'll be- ... coming real soon. This is the crazy thing is that these agents pretty much do the same things, or they're much more efficient, but there's such a nuance involved based on our personalities, our workflows, our habits- Right.

Yeah ... that it's infinite. And I think this is kind of starting to dawn on some of these companies, and they're trying to ram the square peg into the round hole.

It's going to keep escalating and escalating. It's just building upon building upon building, and it's like human layers within companies. It's the same principle in effect.

You think shadow IT is bad now. Oh, yeah. And this is where it comes back to what we were talking about in the previous block.

Whose agent is it? Is it my agent, or is it SAP's agent or Salesforce's agent, and they're just letting me use it? Because if it's my agent, I got a lot more say about what it does, versus if it's their agent, they're going to tell me what I can and can't do.

Yep. Are they going to really let it be all your agent, though? Because they lose control, right?

Yeah. I think that comes back to the heart about what all these companies are flipping out about, especially given SAP and its history in Germany. They like to be in control.

That's kind of where they've always been. Well, but I think there's some nuance here in terms of the evolutionary cycle we're in. I was looking at a certain analyst firm's wave the other day, and nothing is mature.

You know the wave diagram, right? Mm-hmm. Nothing has crashed.

This wave is not even breaking yet. Nothing is mature. So when you say we're lacking imagination and all that sort of stuff, I do agree, but this is stuff that will evolve.

And you think about, is it my agent, is it their agent? To your point, Jack, when was the last time you wrote your own word processor, right? Although I did read a journalist the other day wrote his own email program that just did what he wanted to do, used Claude to do it.

But this idea of bespoke software, yeah, right now we can do it. We're in the early stage. There is no management.

There is no standardization. Over time, we get back to specialization of labor. Over time, I don't want to use Claude to write my own email program.

I just want to get one off the shelf that's already got all the AI built into it. I think over time, packaged software does come backBut right now we're experimenting. We're trying to figure out what's our use case.

Look, when we first started to use, let's call them word processors, Jack was sending his emails to be printed or typed. And so we got the computer to do that work for us, so human replaced computing. What we didn't plan for was sending an API call to a cloud service to call me an Uber.

Right? We didn't know we were going to do that with IT, but then we figured out the things we could do. I think that's where a good thought around the evolution cycle.

We've a long way before we're pinging an API to call an Uber. We're still at the point where we're using AI, the equivalent of printing out my email form. VT-100s.

We're at VT-100s when we need to get to Windows. There you go. And there's three of us on this, I think, who know what a VT-100 is.

Yeah, you guys. Yeah. Hey, Mike, can I ask you a quick question?

Sure. During this process, have they recast themselves from a backup company to a data and AI trust company? That's exactly what they're trying to do, and that's kind of where that play is going to be.

And whether or not that's doable remains to be seen. But on one hand, you could say, well, who are they to make that claim? But then, you know what?

To their point, they said, this whole ballgame is up in the air now, so anybody can do anything they want because 90% of the data's in play and nobody owns it yet. Nobody's got a dominant position, so... And you've got the alternatives, right?

Are you going to trust OpenAI with your data or SAP or Veeam, right? Or AWS or Azure or Google. All right.

I'm going to shift the gear to Red Hat because, well, this conversation, you're going to start to see a pattern here. But Red Hat had its conference, and one of the things they were talking about that at least got my attention was, well, Ansible. And they were going to say, hey, we've got all these lovely playbooks that people have created and they're for automating IT operations, and we're going to let you expose those things to an AI agent through these MCP servers that we're going to provide.

And that will be great because it'll enable the AI agent to do something safely. Now, I thought about that for a minute and I said, well, that sounds good, but in principle, in reality, if I was smart enough to create the playbook in the first place, do I really need the AI agent? Because I already figured out how the automation should work, so what's the value add of the AI agent?

In my mind, it was always to automate something that I didn't know how it works in the first place, or to spare me the toil of creating the playbook in the second place. I don't know. Jack, am I crazy?

No, but there's another or. Or to get rid of people. Right?

To make IT more efficient. And I think that's what they're saying. Look, what's Red Hat's history?

It's about being able to get Linux instances run on any platform, anytime from a universal console. That's kind of their business model, right? And so what they're saying is, we're going to manage that more effectively.

You may not need to have 300 IT people, maybe you only need 200. You may not have 30 instances now, you may have 300 instances because they're going to ramp up. You're going to have instances running just on AI.

There's going to be an AI instance running on a VM somewhere. Or a Kubernetes somewhere. And so what they're looking at, it's the same excuse me, it's the same discussion we just had previously about SAP, is how do I become an AI agent-driven company and still keep my customer base?

Mm-hmm. And what they're saying is, look, we're going to tie all this together, and it's going to have MCP capability. There's going to be an API.

You're going to be able to run agents that do all the stuff that you did manually before, but you still need us at the back end to run Linux and Kubernetes and whatever else you're running on our platforms. I think that's really where they're going with this. All right.

John, I know you're waiting for this, but so Red Hat is really trying to introduce or remind us all that we should have trust issues with AI agents. And boy, doesn't that- Right ... sound familiar?

Everybody seems to be banging a drum out there saying all of a sudden we all need to have trust issues with AI agents because, well, they're just That's the new selling point, right? It went from, embrace this idea, this concept, to do we really trust them? We can't trust them.

Yeah. Right. And so, for me, it was interesting.

I liked some of these use case examples they mentioned. I think there was one with ISS and there was one with Nissan, but it kind of brought me to the point of, so if AI can run efficiently on a satellite or a car's firmware without this cloud back and forth, does it shift AI from a chat assistant to a critical system component? It's just like these concepts, and again, they're veering on how they approach or they anticipate or how they define an AI agent from something that is kind of generic we all can trust, to something that we cast a little bit of doubt on and we pivot to the security or the governance of data issue.

And I think what they're doing, they're trying to grasp a way to monetize and kind of evolve their products and where they're going. And I find it interesting, and I also think it's shrewd on their part. We see through it, but I don't think the customers do.

Well, maybe. But I also think- Many may, many may not ... Andy, walk with me on this one a little bit, though.

So my problem with playbooks in general is that they're perfect states of something. If the IT environment doesn't change, that playbook should run just fine. There is no IT environment that doesn't change.

So then, the odds that that playbook's actually going to execute are 50/50 maybe, and over time it gets less and less because the IT environment kept changing. So, are playbooks going to be relevant in a world where AI coding tools are updating our IT environments every 45 minutes? Yes, and absolutely.

And by the way, I love this for Ansible because all of the observability tools-- You know this is my game, right? I do observability monitoring performance, all that, right? Infrastructure is my jam.

And with all of this stuff changing, and you're right, no data center, no shop, no cloud service remains the same for more than, these days, a couple of minutes. But this is exactly why I love this for Ansible. " Right?

A human will go and check out the last ticket that was opened. A human will go and read the chat logs. This is stuff I'm building right now.

The idea that I can understand a problem, but I don't know how to fix it without going to all this context. Where's all the context coming from? Well, part of where it's coming from is the runbook.

Where's the runbook today? Maybe it's Confluence. Maybe it's SharePoint.

Half the time, it's in an Ansible script or a Puppet or a Chef or maybe a Terraform or something like that. If I want to remediate a problem, I've got to take an action. I've got to reconfigure a switch, reboot a server, fire up another pod, whatever it may be.

I need an automation tool to do that, and so why wouldn't it be Ansible? How much information do I need to pass to Ansible to make them do the right thing? Do I need to check on whether they're doing the right thing, or does Ansible's own internal AI agent go out and figure this out for me?

I don't know yet. I think this has got legs because I love the idea that Ansible is an automation agent and an AI agent, and I would actually challenge most people to tell me what the difference is. I would say that the AI agent will look at the Ansible playbook, determine that the Ansible playbook or any other playbook is out of date, and write another playbook that is more relevant and useful.

Maybe. Yeah. But how frequent is that going to happen, do you think?

Well, once every 12 seconds, probably. Well, there's another piece to this. Computer speed, right?

Yeah. Machine speed. Jack, what do you think?

I think there's another piece to this that Red Hat's trying to play up, and that is one of isolation. If I can insulate you or isolate you from something gone bad because you just wrote a new agent or you just had a new playbook, that's a security issue, right? That's a resilience issue.

That's a performance issue. " And if they can pull that off, if they can do that, then that becomes really important for a lot of IT shops. Well, how likely are they to be able to pull that off, though?

Well, how likely is anybody able to pull that off? Right? But they're probably in a better position than many.

Okay. Because they're really- VMware, as well, but VMware, we don't even hear from anymore. But if you can build an isolated VM somewhere and run the agent in that, at least they can't do damage outside of that isolated environment.

Do you think that's sufficient for building trust, though? Because trust is a kind of different word there. Yeah.

But it'll get more fun than that because the management of the applications and the management of the infrastructure, along with the security, is going to become more integrated as we go forward. And I might have an AI agent, or maybe one or two of them, that are all part of what some of the folks in the AI world will call a harness, and they are collectively managing that thing together. And this distinction we've had between infrastructure and application management for years will just get blurrier and blurrier.

What do you think, Andy? Yeah, look, I think that barrier's got to go anyway. We've already gotten a long way past that step by outsourcing our infrastructure entirely to cloud services, and we don't even see a server.

Half the time these days, we don't even see an operating system. All we see is the container. So I absolutely think that that's where it's got to go, and I think that someone like Ansible is in a great place to own that, at least within their boundary lines.

Because they do cross boundary lines. By the way, I've done my own research on that. Crossing boundary lines is super efficient.

Humans are really bad at it. Computers don't care. And so I think if Ansible can actually cross that boundary line effectively...

And by the way, we trust Ansible. This comes back again to this idea. Ansible hits my prod systems.

You mentioned Harness. That's another automation tooling. As a brand name or as a concept, either way, we're hitting the production instance with these tools.

By the way, Veeam as well. We're backing up our production data, and maybe restoring into production when we need to. SAP, obviously, production systems all around.

This is about, for me, Ansible is someone I trust to hit prod. Right? All of the agents we're talking about, a lot of it's just back office stuff.

It's do my email for me. It's come up with a strategy for me, write some code for me. Then I'll review it and see if it's worth putting in front of my board, my CEO, worth checking into my repo and putting into prod.

But all of these things we're talking about so far, they're already in prod. So if I already have that relationship, that trust, if I've already ticked that box that says, "Do not show me this again," maybe that's the right place for those systems to be. And could Ansible gain more across boundary lines?

Absolutely. Automation tools always expand across boundary lines, especially when they're orchestration tools, and I draw a distinction there, and I think Ansible is that, too. So John, I'm going to give you the last word on this, but I always feel like there's a bit of a paradox here, and it may be just me.

But we all have trust issues, some of us more than others. Probably me more than others. Two.

But at the same time, when someone comes and starts poking at my trust issues to tell me that I should buy something for them, that just hits me the wrong way. I don't trust that person. If they're provoking me about issues of trust, they're projecting what they want me to think, and I trust them less.

And when I think about that, there are a couple companies that come to mind from different spectrums. Like on one point, when Apple says it, I kind of agree with them because that's part of their culture. But when I hear it from somebody like Meta or, back in the day, Facebook, immediately Spidey sense goes off.

Yeah. It's like, I don't trust this company. And I'm starting to get that feeling to various degrees, and I'm sure all of us are, about some of these companies that we've been talking about.

And sometimes it almost seems like there's a sense of doubt or desperation, and they're trying to appeal to fear, because fear sells. Our politicians have proven that, I believe these companies- Fear has always sold very well. Fear sells well.

Not just in politics, but in tech. Especially in tech. But- Well, I'm going to end this group therapy session right here on Techstrong.

I think we're- I don't trust you, Mike. But hey, if you are having trust issues, at least you can have some company now because you're probably not the only one. Hey, I want to thank all our panelists for sharing their insights.

As always, guys, thanks for spending some time with us. " As always, it's our pleasure. And please stay tuned for the rest of the replay of the Techstrong TV lineup because, well, there's lots of content in there, and I'm almost certain that a lot of it has something to do with trust.

We'll see you next time.