What's Up with Tech?

How Agentic AI Turns Fleets Of Macs Into Enterprise Muscle

• Evan Kirstel

Share episode

Interested in being a guest? Email us at admin@evankirstel.com

What if your next high-performing hire never sleeps, knows your stack, and ships secure environments in hours? We dive deep with Chris Chapman CTO of MacStadium to map the shift from brittle automation to agentic AI running on fleets of Macs, where outcomes replace step-by-step scripts and digital employees take on repeatable, high-value work. Chris opens the hood on Orca, their software-defined Mac platform, and shares how customers virtualize thousands of machines to train agents, spin up secure developer workspaces, and deploy specialized small language models that thrive on Apple silicon.

We explore why early movers gain an edge: building agents around their policies and workflows instead of bending to generic tools. That means using standards like MCP and A2A for reliable tool use, embracing hybrid architectures that keep sensitive data sovereign, and bringing compute to the data rather than hauling data to the cloud. The result is practical speed: fleets that scale on demand, agents that verify progress at checkpoints, and environments that remain auditable and compliant. Add the efficiency of Apple silicon—high performance per watt, unified memory, and strong on-device inference—and you get meaningful cost control without sacrificing capability.

Security stays front and center. We unpack a security-by-design approach where agents inherit guardrails, act on alerts, remediate drift before rollout, and document every change. We also talk shop about the road ahead: M5-based clusters for training and inference, deeper graphics and AI integration in Orca, and Apple’s unique edge-to-cloud continuity that unlocks multimodal workflows from Vision Pro and iPhone to Mac mini clusters. If you’re weighing how agentic AI will reshape roles, budgets, and delivery, this conversation offers a clear blueprint: encode goals, codify guardrails, and let agents do the heavy lifting while your team moves upstream to architecture and strategy.

If this resonated, follow the show, share it with a colleague, and leave a quick review to help others find it.

Support the show

More at https://linktr.ee/EvanKirstel

SPEAKER_00:

Hey everybody, fascinating topic today as we talk about the next wave of enterprise agentic AI with an innovator in this space. Chris from Mac Stadium. Chris, how are you? Doing well. Great to be here. Thanks. Well, thanks for being here as a Mac fan. I'm very familiar with your product. But for those who may not be familiar, who is Mac Stadium? And talk about your role in the enterprise tech world to start.

SPEAKER_01:

Sure. Yeah. So Mac Stadium is an enterprise class Mac solutions company. We have data centers around the world. We are fully based on Mac, no other type of compute. We've been around for about 14 years. We have a flagship solution called Orca, which is a software that enables software-defined Mac at scale. So we have north of 20,000 Macs in our data centers globally. And then Orca allows you to really turn it into a true cloud and deliver it for software development, remote workspaces, or for now, obviously AI as it comes into the picture. And I've been here about seven plus years. I came in through acquisition to really help accelerate and define the software capability of Mac Stadium, who had traditionally been more data center and then obviously was wanting to move more cloud and now pure play software, whether it's our data center, AWS, or even on customers' prem, helping them build hybrid clouds where they can combine Mac resources wherever they are and really connect them to do what they need to do for enterprise scale solutions.

SPEAKER_00:

Great. Well, as a Mac fan and user, uh it's thrilled to see uh Max eating the world uh these days. Um but switching topics to agentic AI, um you've you've seen this evolution, you know, pre-hype, now we're in peak hype. But what do you think about hundreds of millions of digital agents entering the enterprise? Do you think that's a real prediction? And how are you uh anticipating this rolling out?

SPEAKER_01:

Yeah, I I do think it is real. I obviously there's there's what we can do right now and what we hope to do in the future. I I do agree with you that there's the the hype is still very hypey, but um but we're starting to see uh actual utility out of it and and emerging standards. MCP obviously is a big one um that that the folks with Claude and everything have put out, and then A to A on the Google side. But this really starts to give us the ability to intelligently define interfaces that were traditionally static. So I, you know, the way I think of it is we've had automation forever and it's gotten robust and complex, but that involves a big team behind it. So you've got skilled workers like developers and IT admins and things like that defining these interfaces and sort of making static decisions, and then they're forever condemned to keep up with that sort of evolution. And every time a system upgrades or changes, they're working on both sides. And then we got AI, and everybody got really excited. But the initial burst for AI was really data analytics almost. It could look at things and go, hey, based on what I'm seeing, you might want to notice this trend. But you again, still the human going to do the effort. I think the promise with agentic AI, and for us in the Mac space, we see it as this evolution to what we call MacOps, which is really the ability to quit focusing on the individual task of the desktop because MDM will let you manage the state of a machine. But what if I want 300 machines to create a solution for developers that have multiple tools, multiple security solutions, talk to several systems, and I just want to state the goal. I think agentic AI is the promise of that, where it can, on my behalf, be that worker be, and I can start leveling up and saying, this is what the business needs to do, this is how we need to do it, this is what we need to package with it. And then I have a fleet of those workers who can tirelessly interface, automate, you know, kind of check the data, get the feedback. And it really sort of boosts productivity significantly, I think.

SPEAKER_00:

Interesting. So, from your point of view, what's driving enterprises to kind of really explore deploying it to ethnic AI versus just waiting for the tech to mature and deciding, hey, when the person next door does it, we'll do it.

SPEAKER_01:

Yeah, you know, well, obviously, I think AI is in early arms race phase now. So everybody was sort of mandated by by their boards, by their business, by their go figure out this AI stuff. So I think that's part of it. But the reality is I I think it's in a phase now where it's not enough to just AI something. AIs are gonna make your business intelligent and effectively you're going to start creating digital employees for your company. And there's a certain amount of off the shelf that you can buy that can do. But I think the earlier you adopt and the quicker you jump on that train, the more you're gonna be able to put your fingerprints, your business, and your technology into place and really build it in a way that becomes that digital employee. I would much rather dictate how my business operates with intelligence than have the intelligence dictate my business. So I think that's the reason to go early.

SPEAKER_00:

Yeah, it makes sense. And are there any you know, specific anecdotes or stories on how your customers are using agentic in practice, not just uh the business, but utility.

SPEAKER_01:

Right. So we we've had we've had a lot of folks come to us for AI from just pure industry side, but also, believe it or not, from the creators of AI, the big clouds that are building the frontier models and things like that. So relative to agentic AI, we've had those folks basically take fleets of Mac, virtualize them with Orca, and then train and develop what the agent can do by creating massive fleets of virtualized Macs at scale, doing tasks over and over and over and over again so that they can create things like you know what you're seeing now, which is intelligent Mac-based browsers that can go book a flight and do a thing and all that kind of stuff. So they're actually creating their agentic product with us. And then there's there's other companies that are more in line with their business use that are just again trying to create sort of fleets of specific masks to say, again, remote workspace for developers that are offshore is a good example. We have companies that want to spin up contracting teams overseas and they want front-to-back security. They want all the tools, they want remote access, they want this sort of all as a package. And using a gentic AI can rapidly scale up and scale down a fleet in hours versus a human cobbling together individual Macs and shipping them. And you know, you've got weeks and months and logistics, and it's just really accelerating the time to delivery and the control of the environment when they do it that way.

SPEAKER_00:

Yeah, interesting. So speaking of control, you know, companies are still trying to balance cloud scalability and all the goodness there with data control and uh control of individual devices. How does how do you think about that tension and how do you balance those two?

SPEAKER_01:

Yeah, I it it is a it is an interesting problem. And with AI, where your data is and what it's doing for you becomes an even greater problem because it's sort of automates itself and spreads out where it needs to go. I I think it's super important at that point, and it's a stance we took at Mac Stadium to not just be the cloud, but to really help our software enable stuff wherever it sits. So hybrid. I I think data sovereignty, I think the ability to manage the transport and security of your data where it needs to be and still connect to it is important more than ever. And I think we're gonna see providers start helping you get to your data versus having you pull your data to them. I I think that's gonna be a mode that's critically important in AI. You want the data to stay as tightly coupled to what your boundaries are as possible. And so uh again, I would I would urge people as they develop systems to see how they can keep the data contained and bring the resource to the data versus the data to the resource.

SPEAKER_00:

Oh, well said. And so, as a fan of Apple, uh their silicon continues to blow my mind. Uh uh M5 just out, of course, the price performance is just continues to uh be a rocket ship. I I assume that gives you an edge in the enterprise as well.

SPEAKER_01:

Yeah, we're we're we're really seeing it. I, you know, obviously AI is amazing, but it also brings the fear of the most power-hungry thing we've ever developed as humans, and more power needed than than devices. But Apple's got a really powerful position there in that even in our own data centers, we effectively overpay for power because the performance of a Mac and the power consumption of a Mac is so far below other types of compute that it's hyper-efficient. I think as people become sensitive to that and to green initiatives, Mac has a really powerful position there. And then, as you said, from a technology platform, they really are all in on making it the most powerful AI compute as in an individual device as possible. And people are starting to see as these models get into new performance, new heights, they really are starting to be able to shrink down uh small language models and SLMs instead of LLMs, are really becoming interestingly effective. They use a little bit of a different thought technology behind them, but those things can run on devices. And Apple's got you know neural processing and and uh really high shared memory and things like that that make it a really amazing compute device. So what we're seeing is, you know, initially the wave was kind of just inference only, Macs can be used for inference, but um, we really are starting to see them in specialized cases for training. There's there's technologies um that are starting to develop around that that are helping Apple scale. And uh we've been able to do small-scale clusters for training and inference as well. So it's really kind of a bright future for Apple, in our opinion, because it's not only cloud enabled, but it's very, very powerful at the edge.

SPEAKER_00:

Fantastic. And speaking of cost, uh there's a big concern in the enterprise about the cost, not just of uh, you know, the applications that are pretty costly, but the cost of compute and power and any data centers, cloud sprawl, you know, paying more for cloud usage costs are rising. How do you think agentic can help reduce costs, uh, not just add to them?

SPEAKER_01:

Yeah, I you know, I think agentic is is again kind of what we talked about early, very specific to your business goals and your needs. I I do think humans are gonna have to become sort of leveled up and be the shepherds of good guardrails and good architecture, and that's gonna dictate the energy use and the consumption and how things move. But um, assuming we do that right, agents can can make sure that tasks are specific and quantified and that they're not just blasting out for these broad generalized answers, but they're going to perform a specific goal-oriented thing. And in that goal, they're going to come back at the appropriate times and check for is this where we want to go? Is this what we want to do? Um, that coupled with things like Mac, where instead of it being a generalized platform of just massive compute that has to schedule with everybody else and flight for resources, you can have dedicated, private, clustered, scaled stuff with super low power consumption. You can really build something that's that's really effective and really powerful.

SPEAKER_00:

Amazing. Well, I remember a day when you couldn't bring your Mac into work, into the enterprise. And if you did, you had to buy it yourself. And it was all kinds of exceptions. And uh man, that was tough. And we've come a long way in terms of security and compliance, and even employers giving Macs to employees. But what's next with security and compliance when you get into this agentic world? Uh I imagine there's some big barriers there.

SPEAKER_01:

Yeah, yeah, yeah. It it can get it can get spooky from that perspective. You have to, uh you know, but I think it's true of how security should be applied to software development in general. It should be security by design, not security after we built it. So I think if you're building these agents and you're integrating them up front with best practices and standards and compliance in the agent, it inherently becomes aware of the security context that it has to maintain and uphold. So it starts to not only alert you like current securities tools do when something goes wrong, but it can act on the alert itself and go, hey, this is out of spec, this is out of scope, we're out of bounds here, I need to adjust this, I need to make a change here. And it's all again because of how you built it, how you trained it, how you how you locked it into your tooling. But it can also, kind of different from me, from the way we do things now, can go out and touch those systems and go, hey, this device was way out of spec. Can we go ahead and fix this before I put the next layer of software on? Um, so I think the opportunity is greater than the risk in some cases if you do it correctly.

SPEAKER_00:

Amazing. Um and in terms of um the human role in all of this, um roles are reshaping, changing, whether it's DevOps or system admins or support. But how do you how do you see these roles changing even more with Identic AI?

SPEAKER_01:

Yeah, I I you know, honestly, I'm I'm excited for it because I think it offers us the ability to move upstream and focus on strategy, innovation, sort of solution building. And you know, instead of being the guy that has to hammer each nail, I can be the architect now. I can be the solution builder of the whole thing. Um that does take an adjustment, but um I think that's the skill set that it that it brings to bear. It's just like the industrial revolution. I mean, we we stopped hand knitting everything and then we've got large machines that can kind of do it. This is the same sort of thing, but digitally speaking. So I, you know, that that's where I see this going. Um and what I hope it empowers again, like like our concept with bringing DevOps to Mac in a in a Mac ops way is it starts to let you think about not the individual machine and all the software on it and how you have to put it together and what you but like what am I trying to deliver? Is it a remote access solution? Is it a developer tool? Is it a creative suite for Adobe? You know, what is the person trying to do with the thing that I'm delivering? And then let the agent and the AI help me create that world to give to them super fast and consistently.

SPEAKER_00:

Fantastic. Yes, consistency is key. So looking ahead, what are you excited about? Any big releases or news or events on the horizon that uh you could share at this point?

SPEAKER_01:

Um, for for us, we're we're rounding out the year with another release of Orca. Again, that's our flagship software that virtualizes and containerizes Apple at scale. So, you know, I'm always excited when we get that stuff out because we're integrating more into the graphics layer of Mac Compute and the AI layer, and that gives us the opportunity to serve our customers with regard to this technology a lot better. Um, Apple's M5 stuff that you mentioned already coming out in an iPad and and uh laptop form. If they follow cadence and model types, that means hopefully early next year we're getting a mini or a or a an uh a desktop form factor, which for us selfishly is the type of Mac we like because we treat those as servers. So um we see those as clusterable servers, so I can put 40 together and create an AI cluster out of that. And if it's M5, it's gonna be significantly more performant. And then broader, you know, just around this use of tools, I think Apple specifically is one of the few ecosystems from that's got consistent hardware and software from edge to cloud. And I think when you get into agentic AI and how you can apply that to a business solution, you could have a Vision Pro or an iPhone or a watch taking telemetry at the edge, helping you act uh connected to it in an intelligent way, streaming it up to a cloud of Mac minis doing work and then spitting it back down. So to me, as they evolve those platforms and the models on those platforms, it's gonna unlock uh really, really powerful multimodal AI capabilities in the business. So I think that's that's super cool.

SPEAKER_00:

Amazing. Do you get any special attention, love, support from Apple in your endeavors or Tim Cook, or do you have to go to the Apple store like the rest of us with the Red Card and buy your it's a it's a bit of both.

SPEAKER_01:

No, I would say over the past, you know, we've we've always had an a very in interesting perspective for from uh we've always had a big interest from Apple from their perspective because of what we do with Mac. And even back to in the 2018, they said these guys are doing the definition of mission critical with Mac because it was kind of an interesting use case. They love to tell stories. But I would tell you, in the past year, we've we've really, really leaned in hard with our relationship to Apple. We we are an enterprise partner. We talk at the highest levels to the the field and the strategy teams that do enterprise because we see Mac differently, to your point, instead of it being that other machine you had to pay for to bring in the office now, especially with the the the M4 Air, it's really become uh a choice for CTOs. It's got best TCO, best power, and people are kind of walking in who have grown up on these things and said, This is my productive laptop. Help me use it to make your business productive. So that kind of drive has gotten us a lot of attention there. And then from an engineering perspective, we're definitely integrated with their OS teams and the virtualization teams because we're we're kind of constantly trying to make the Mac do a little bit more than maybe it was intended to do from a consumer perspective. So um, yeah, so so we definitely have a lot of a lot of interesting conversations and uh maybe not not so much special privileges. Apple's pretty good at keeping everybody on the same playing field, but definitely a lot of insight and discussions and opportunity from the enterprise perspective.

SPEAKER_00:

Yeah, it's amazing how far they come and what a great story the side of Apple and uh we don't see that often in the press or the media. Congratulations on all the success. Thanks. We we appreciate it. Thanks. Uh Chris, thanks everyone for listening, watching. Check out our companion TV show, uh techimpact.tv on Fox Business in Bloomberg. Thanks, Chris. Thanks, everyone. Thanks.