Securing Retail AI & Agentic Workflows with Clyde Williamson & Jessica Hammond (Protegrity)

Scott Mann

Welcome to PodMagic, real conversations about solving real IT challenges. I’m your host, Scott Mann, SVP Global Sales at StorMagic. Here at StorMagic, we’re always exploring how simple, reliable technology can benefit you and the people you serve. Whether you’re running branch offices, retail stores, or supporting customers on the front lines, our goal is always to bring interesting guests, deliver some value, and have some fun along the way. 

Today we’re digging into a tension every retailer leader is feeling right now. How do you put customer and operational data to work with AI without putting that same data or your customer’s trust at risk? And joining me today are two guests from Protegrity. We have Clyde Williamson, who is a Senior Product Security Architect, where he helps organizations strengthen their security posture and build smarter data protection practices. And Jessica Hammond, who is Senior Director of Product Management for Gen AI, leading the charge on building secure, scalable AI systems across the enterprise. Clyde, Jessica, welcome to PodMagic. Great to have you both.

Jessica Hammond

Thank you. It’s good to be here.

Clyde Williamson

Hi Scott.

Scott Mann

Pleasure to have you. And let’s get into it. Retailers we know today are leaning more and more on AI. There’s personalization, dynamic pricing, fraud protection, customer service, store operations, and it all runs on the same fuel, which is customer and operational data. So to start things off, let’s start with where we’re currently at. I’ll start with you, Jessica. What are the most impactful AI use cases you’re seeing right now in the retail space?

Jessica Hammond

Well, a couple have been around for a while, and that’s personalization and recommendations. And then there’s the second one to that, which is demand forecasting and inventory. But the newest of the bunch is actually agentic customer service and operations, really taking what used to be manual or human-led processes and agentifying them or making them run more autonomously. And then of course supporting users with agentic responses or agentic customer support. And I think Clyde might actually have a little bit more to say on some of those things.

Clyde Williamson

I spent about 10 years doing IT security for retail organizations. And the personalization and recommendations that working with the customer data, fine tuning exactly what works for different customer groups, was a huge part of what marketing needed to do and it was constantly an area that IT security had to keep an eye on. AI makes that both more powerful for the marketing team and a little more concerning for the security team.

Scott Mann

I completely agree and you said a lot of things there that we want to dig into more, but I don’t want to lead the witness too much on some of these questions. I think what’s interesting though is the marketing side of it is the very unique part with retail. It is so customer focused and it’s all about that customer experience and ultimately, how do you get them to buy more and engage more. But now there’s more evolution to that as well on the operational side and everything. We’ll get into all of that, I’m sure. That sets the precedence for like a lot of this conversation. 

We’re going to be talking about security, and we all know that everyone says that data is the fuel for AI. What is the biggest challenge that retailers face as they try to use AI more broadly? And I think we touched on that a little bit, but let’s dive a little deeper on that with the security side, especially. Jessica, you want to kick things off, go for it.

Jessica Hammond

Security is actually, I think, one of the biggest data challenges that retailers are running into, but that’s also a result of a couple of different things. One of which is, do they have the tools that provide the appropriate level of security or data protection in these new systems that can be reliable to five nines, if not greater. But also, we’ve had this historic experience of people and humans doing a bunch of different things with data and that has historically and currently still does create a visibility problem where we don’t actually know where all of the data lives. All of that sensitive data exists and if we can’t see it, we can’t protect it. If we don’t know that it exists, there’s nothing that we can really do about it. And if a human can find it, an agent can probably find it too. So that’s also in addition to fragmentation where you have data all over the place. I’ll pass it over to Clyde because I know Clyde has something to say on that one.

Clyde Williamson

We actually have run into this time and time again. It’s the most common challenge our customers have, even before Gen AI showed up. Just in the decade plus that I’ve been working with Protegrity, we’ve been helping customers protect data. Almost always when you walk in and we talk about how do we protect this data? Where’s the data? We don’t know where the data is. That has been one of the most common responses from customers, consistently. And when I was actually in retail organizations, PCI, the payment card industry’s standard, had just come out. And they put me in charge of figuring out how to implement PCI for the organization. We spent six months figuring out where the credit card data was across all the platforms, because nobody had ever bothered to figure that out. PII data is even worse. Credit card, at least some people understood that maybe don’t put that everywhere. But customer names, addresses, phone numbers, that’s public information. Who cares? It can be anywhere. 

Another organization we worked with, which was a government organization that was actually trying to deal with making sure that people were taxed appropriately, they had a major problem because John Smith versus Jay Smith versus another variation of that, those are all the same people. But they will change the way they write their names slightly to hopefully avoid getting hit with the same taxes. There was another case where they had massive amounts of information, massive amounts of PII data about their own citizens, and they had a really hard time figuring out from the fragmentation, who is this? Is this the same as this person? That’s something that we’ve seen across retail customers as well.

Scott Mann

It’s almost like there’s an abundance of data, but how do you control the data, where it’s coming from? Where can you access the data and what do you do with the data? Many organizations, now that they have more data, everything is producing data, it used to be human inputs, now it’s environmental inputs. There’s whatever they say, a thousand times more data that’s being generated, but why are why are so many organizations still struggling to turn that into business value? They have the data. We know they have the data. How do they translate it into the business outcomes that they’re looking for in the retail space specifically? But you’ve got a lot of knowledge elsewhere, feel free to expand on that.

Jessica Hammond

Data readiness and data cleanliness has always been an ML problem. Now it seems as though maybe it’s not as bad because AI can help you clean it up a little bit, but that’s not completely true. Then additionally, that means that you’re entrusting whatever you actually do have in that data set to the model that you’re working with, which is kind of the problem. If the model can’t necessarily be trusted fully or the data leakage is a concern, or that’s all on the prep side. And actually, that’s been the historic problem. Now we’re moving into the agentic problem, which is the autonomy of agents and leveraging that data, whether in a messy state or a clean state, on its own and running different tasks, running on behalf of other users, running with their own identities. And it really boils down to a data trust and data governance problem where, if you don’t know where your data is, you don’t know what condition it’s in, you don’t know how it’s protected, you don’t know what sensitive data lives inside of there, there’s no way that you can actually manage it or govern it. There’s no way that you could properly protect it. Whether you want that agent in the end to have full access in the clear or not. Maybe you’re in a position where the model can receive that data in the clear, and that’s fine. But more than likely there are points in time when that’s actually just not true. There’s more times where that’s not true than there are times when that is true. To remediate the initial problem, that’s going to be how you can alleviate some of the secondary problem. To be able to have the coverage, have the visibility, govern it properly, secure it properly, and then make sure that that security extends into the usage of that data. No matter where it’s going, whether it’s going into an ETL pipeline or into an agentic process.

Clyde Williamson

I think this challenge is different for different organizations. Some organizations we work with are very concerned about credit card data, social security numbers, sensitive business information. But PII data, not that big of a deal to them. Whereas some of our customers in Europe are so terrified of the privacy legislation that one meeting we were in, this customer was talking about the amount of data they’re collecting from their cars. They manufacture cars and sell them, and those cars send massive amounts of data back to their organization, which they think would be really amazing to hand to AI so they can understand more about which parts are failing and did they come from the same manufacturer and things like that. But they’re too afraid to touch any of it. They’re just collecting all of that data, storing it, and locking it because they don’t know what to do with it in a responsible way. So, there’s two sides of what’s going on in retail right now. We have some retail organizations that are responsibly collecting the data and don’t want to have risk associated with using it. And then we’ve seen other customers where the data is everywhere and they’re just applying AI to it, trying to make it work. But AI on poorly managed data is just going to be poor.

Scott Mann

I imagine with a lot of these we’re talking about, why are they not getting business value out of it? It’s really tough because not everybody has all this experience with AI. Some of these projects might be brand new, never been done before, just testing the waters. I do a lot of things on the back end trying to figure out how we can automate the marketing and the sales side, use AI in to make our team more efficient, more productive. I imagine a lot of these customers dealing with sensitive data becomes an even bigger problem. That’s where Protegrity comes into play there, where you guys do this on a daily basis. It’s not your first time doing it, you’ve probably heard everything that every customer has to say at this point. A slight pitch for you guys without expecting it. 

The one thing that we talked about is you say Gen AI and agentic AI, the game has significantly changed as talking about agentic AI now. And that’s where it gets a little bit scary around all of the security. But how has that conversation around data protection changed since it went from gentic AI and evolved to agentic AI? Did I say those names right? I think I did.

Jessica Hammond

There’s a couple of different lanes to this conversation. To your point, Scott, there’s talent and understanding of what agentic workflows can do for a business and then the repercussions of that is still growing and the knowledge base is still extending itself across different sectors, some more quickly, some slowly. Security teams have to be on top of this really, really quickly. We encourage security teams to become the AI experts in-house because they will be the blocker or the enabler for their entire org in advancing in this day and age. Also, security should be the first thought, not the last thought. It’s not the gate that says no to everybody. It’s the thing that enables everybody. And if security teams can take that repositioning for themselves, totally different ballgame for them.

That being said, with the increase of agentic and autonomous conversations and the usage of data in them, which has been true for Protegrity for a very long time, you’re not talking about perimeter security, you’re talking about data centric security. And if you bring the security to the data, then who’s getting in the house almost doesn’t matter because the valuables in the house are locked down. A lot of people say, if we just lock the door, then nothing will happen to our stuff inside. People get inside the door all the time. If you’re locking down the actual valuables that are inside of the house, then you have a much better chance that your valuables stay with you, or that they stay protected, even if they’re moving around. 

The other piece that we’ve run into on this is the ‘will this leak’ conversation and who authorized this. The audit trails and such, we’re very familiar with that too on the security side of the house, making sure that we understand who, what, when, where, and how, and making sure that policy and data-centric protection applies at the data level, no matter where it is. I mentioned that earlier too, whether it’s an AI workflow or an ETL pipeline, the security has to stay with the data.

Another big deal that we see coming up a lot, which is hard to control, but equally, if not more important, is flat out shadow AI and enterprise contributors using whatever tool that they can to solve a problem that they have. Even yourself, you’re saying, I’m trying to build this agent or that agent to solve my marketing problems. Everybody across the org is trying to do this. Some with more experience and some with less. The security team can be an enabler in that, let me help you do this in a way that you are solving your pain point, but also, we are making sure that the data stays protected. 

So those are the kind of the top three things that I’m seeing. Clyde, I’m sure you have something to add to that.

Clyde Williamson

In information security right now, there are three ways that most of the people that I’ve talked to seem to be trying to approach the agent issue. The initial one that everybody leans to is well, the user is using the agent, the agent is a tool, therefore the agent has the user’s permissions, and that’s all we need to say. Then as they start thinking about that a little bit more, they realize, the agent is doing things on its own, not simply being a tool, meaning that if I tell the agent, “Hey, go and do some analysis and figure something out”, it’s going to do it however it has decided to do it. Which could mean running a thousand queries that do table scans across my database. So even before we get to the question of how does this agent handle the sensitive data? How does the agent get the sensitive data? And is the way the agent’s going to try and get that sensitive data going to pull up my environment? Then we get to how does it get to the sensitive data, should I give it access to the sensitive data in the same way that the user has, or should I trust it less than I trust the user? Now the idea of the agent actually having its own identity and needing to be vetted is starting to get some traction. 

It’s not just, “Oh I have Claude, therefore everything is fine”, or “I’ve approved Claude”. Well, what is Claude’s prompt? What are the tools that Claude has? What are the skills that I’ve downloaded for Claude? I downloaded a skill for personal use two days ago, an architecture review skill, that let Claude review architectures. Before I use any skill, I’m paranoid, so I run scans on it. Turns out this skill uploads all of your code to somebody’s server so that it can do that analysis.

Now maybe they’re really nice people and they’ve got this hosted server for free that’s doing analysis on my code. Or maybe that’s not what’s happening at all. I don’t know, but I didn’t use that skill. I actually reported it back up to the site that maybe there’s something happening there. How many people inside an organization are doing that? It’s not just, “okay, well, we can trust this agent”. What tools does the agent have? We have to trust those too. What skills does the agent have? We have to trust those too.

This is a really big area of a whole lot of moving pieces. For data protection, we can say, “lock down the data, don’t give them access to the data”, and we have less worries. But as soon as we start giving access to the data, all of these additional things, additional add-ons that we see happening in the agentic world suddenly come into play.

Jessica Hammond

The solution is not just to say, “no data for you”. That doesn’t solve the problem. It’s not going to bring your organization the value that it could have if you allowed some representation of that data to exist. And that representation can look like synthetic data that’s generated in some mirrored fashion, to the original data set that you had. It could look like anonymized data. It could look like semantically similar tokenization. Like there’s a bunch of different ways to actually use real data in a protected manner with AI and have AI still infer against that data and produce sound results for you. That’s where we want to get to, and that’s what we do every day. But I don’t know that everybody’s considering that across the board. There’s varying degrees of ignorance.

Scott Mann

All three questions that we’ve asked, how do we get the business value out of this and how’s agentic AI changing things? This all ties back to each other. It’s so new. As much as we’ve all talked about AI for what feels like forever at this point. It is so new and the knowledge is probably the gap for a lot of people and the experience that they have with it. I’m learning things. This is probably the most interesting topic that I can talk about on any of these podcasts, no offense to any of the other ones that I’ve done. I love hearing your perspective on it all and the concerns that we’re all thinking about in the back of our head is there as well. 

The next question we’ve really touched on quite a bit, if there’s anything else that you guys want to add to it. As retailers are moving from the pilots, they have these ideas, they’re trying to learn themselves, they’re moving from pilot to production. What challenges emerge around data security governance, and the big thing is trust.

Jessica Hammond

I would concur, yes. It’s the last mile of governance and actually considering governance as the last mile rather than as the starting point is where a market shift and a thinking shift needs to happen, especially as it relates to AI and how much enablement you’d like to achieve out of it.

There’s a bunch of different ways to skin a cat. There’s a bunch of different ways to solve the problem. It depends on your use case, and it depends on what you’re trying to achieve as an outcome as to what tool you use to solve your challenge. As a contributor or as somebody that’s building in this space or somebody who wants to build in this space or wants to solve my own problem, I would start with that perspective, the use case as to what I’m trying to achieve, what I want the outcome to be, and then working backwards from there. Because there might be ways for me to achieve that, with a little bit of research and maybe a conversation or two with some folks who do this a little bit more regularly. I would be available for that conversation as a side note, but my point being that you might be able to achieve that without having to stall it if you thought about it earlier in the process. It could be the guy sitting right next to you, he’s got he’s got the answer for you and we just didn’t have that conversation.

Clyde Williamson

Part of this is the speed at which we can move now. The speed at which we can make a prototype, get amazing looking results, and quickly roll that out is powerful. It’s intoxicating. Vibe coding and agentic engineering are wild. But it does mean that sometimes we are designing while we’re writing the code. We are implementing while we’re still building and because of that, there are classic information security gaps that get missed. I’m thinking about the resume AI, was it McDonald’s or Wendy’s, one of the two, where because it was exposed to the public, the public were putting their resumes in and they were talking to the AI and they were getting interviewed by the AI.

Nobody thought that now hackers also have access to the AI and hackers can do things with this AI or at least try to get it to do things. Nobody had considered that. They were so quick to get it out to the market, they didn’t really think about what’s the impact to the data that’s being collected. All of these resumes are being collected by this AI. Why did the AI have access to spit all of those resumes back out? Classic information security problem, but it didn’t get addressed because everybody was moving so quickly. It doesn’t mean that we have to slow things down. It means that information security now is going to have to move that quickly as well.

Scott Mann

Everybody’s trying to use AI to get an edge on their competition. Do something to create more efficiencies, productivity, whatever that looks like. That’s where the snowball effect happens. If you don’t have that early foundation governance and everything that we’re talking about, once that snowball starts reaching a mass that you can’t stop, it just keeps going and people want to get it out there as fast as possible.

Jessica Hammond

Everybody wants to do something AI, and it becomes like AI has zero meaning. Some of our team members have been to AI conferences recently and you cannot distinguish one vendor from another because everybody’s AI. Everybody says AI this or that and well, how are you AI different? How are you distinct from the guy next to you? It’s becoming a wash of hypersimilarity. Maybe think about that too as you’re designing these AI products. You still need to get the attention of the consumer and whatnot, and you want to talk about AI, but it becomes like they’re all the same. Everything’s the same.

Scott Mann

You get to the point where it’s almost like everyone’s the same commodity, but that’s where you guys come into play too. You guys have the conversations, the vendors vetted out. You’re the consultant for the customers. Again, I didn’t mean to give a pitch, but there’s another for you. 

A topic that’s very interesting for us, obviously on the StorMagic side infrastructure. 

It’s around where that data resides and how retailers should think about protecting that sensitive data as it moves between on-prem, cloud, applications, AI models and everything. Tell me a little bit about that. Let’s dig into how we protect that data as it’s moving across. And if you want to dive into where you think it should live and how it should live, what that architecture should look like.

Clyde Williamson

I can start with this one since it’s a data architecture question. Generally speaking, so historically, for twenty-some years, when Protegrity has gone into organizations to help them protect their data, the first thing that they recommend is that you protect the data at the moment you collect the data. If you are a retail sales organization, your point-of-sale data should be getting encrypted at the point of sale. It should flow encrypted through your entire organization and it should only ever be decrypted when I need that piece of information. 

Then tokenization came along several years later, we were actually able to shrink the percentage of use cases where they needed the real data. It turns out if marketing is trying to figure out based on the credit card number, whether across the six different brands they support, they have the same customer. They want to be able to do analytics and see “how often does this customer come in and shop”, et cetera. They can do that with tokenized data because tokens are consistent and they’re linkable. The same credit card always generates the same token, meaning that for most of my use cases, suddenly I can work with this protected data. 

AI doesn’t change that. Keeping the data protected from the moment you grab it and only protecting it at the moment that you really need that information is critical. In an agentic system, that might mean that the data at rest is always protected, that when the data gets picked up, it’s protected. And when the data gets displayed to the user, it’s unprotected at that point, in a report, perhaps. 

Or there might be a specific use case where the agent actually needs access to that data. There might be a process available for the agent to get access to the PII data. But again, it should be very specific. Always the data is protected, and for a very good reason, we can create a spot where the data could be unprotected, maybe at the application layer or the presentation layer. But when you work with it like that, it means that the security can be distorted so that I can now focus on these three points where the data might be accessible versus everywhere that that protected data flows.

Scott Mann

That’s a complex question that you gave a real simplified answer to. Kudos to you, Clyde. 

We talked about how they want to get to market with these things fast and then there’s the protection side of the data. How can retail retailers balance that personalization with the responsibility to protect customer information and maintain the trust?

Jessica Hammond

It’s kind of wild and Clyde and I were chatting about this a little bit earlier too. There’s what customers or consumers say and then what consumers do. I was looking at some stats where 75% of consumers say they wouldn’t buy from a brand that they don’t trust with their data, and fewer than four in ten forgive a company after a data loss.

For me personally, and I work in security every day, every thirty days or so I get a ‘here’s your free one-year credit monitoring letter, use this code to activate in this credit bureau whatever, whatever’. And the store that might have had the breach, or the credit card, whatever the organization is, I don’t necessarily turn around and walk away from that organization. Yes, of course, when first giving my data to that organization, I had to at some level trust them. But especially in the US, because we have a slightly different conception of privacy than places like the EU, for example. I think our standards as consumers are a little different than they are in the EU. We have a lot more forgiveness than we do holding somebody or holding the business to a standard. Which is not to say, “hey, freeball it guys”. It’s to say maybe we need to do more consumer education. But on the flip side of that, those same consumers are the ones working in your enterprise. It’s a double-edged sword. There’s a much larger discussion that could happen there in general.

At the end of the day, for myself and for the work that I do every day, trust is golden. I cannot operate without it, and the businesses that we work with cannot operate without it. It’s not the little every 30 days, here’s your one-year free credit monitoring insurance that a lot of businesses lean into, that’s not something that the companies that we work with, that’s not an option for them. I’m sure they have, as a result of something really, really terrible happening, but that’s not an acceptable output or outcome. 

Clyde Williamson

I was thinking about the whole concept of trust with our customers and how AI is being used. Most of the customers that have been interviewed seem to think that they would love to have AI help them get personalized coupons or an upsell at the cash register that fits exactly with what they need, or when something they really like goes on sale or their favorite color of something. They would love that. A lot of my friends just geek out over being able to go try on glasses virtually over the computer because the AI brings their face in and they get the glasses on and look at it from all the different angles. That’s great. There’s a lot of normalization of AI usage there. 

But then I watched an interesting documentary a few weeks ago where a consumer rights group had brought an entire category of people in, gave them access to a digital shopping cart solution that’s very popular in the United States today. They went to the same virtual store, they bought exactly the same things, and they had wildly different pricing. That causes consumers to lose trust. Anytime that somebody can show that the AI is suddenly working entirely on behalf of the organization rather than the consumer, the consumer will get suspicious. Now, the advantage is that consumers don’t seem to have a very long memory in the United States, but it’s still a challenge that we’re going to have to deal with over the long term. As people become more aware that AI is doing a lot of things, they’re going to start asking more questions about what exactly is AI doing for me versus for you.

Jessica Hammond

Also, the ramifications, the cost of the consequence. There’s so much, “this is fraud, I can report that”. The protections that banks and credit cards and things of that nature have afforded us in recent years. I’m still concerned about maybe the social security leaking, you can go sign up for stuff on my behalf without my permission, that kind of thing. But until, for example, you have an agent going and circumventing the house deeds in a county autonomously, that’s real, that’s a painful process to try and remediate. It’s archaic in and of itself in so many ways. It’s at that point where you’re going to start having the pain to consumers be so great that there’s a little bit more, “whoa, what’s happening here?”

Scott Mann

Consumer knowledge, but also more regulation around it all. There’s some, but it’s definitely something, an area of focus. 

I got two more questions for you guys. First one, let’s role play it a little bit. I’m a retail leader, what’s one piece of advice about preparing my organization for AI? If there’s one piece of advice, what would that be?

Clyde Williamson

I would give two pieces of advice. The first piece of advice is, you’ve got to make sure that your data is clean, that you know where that it’s at, and that it’s consistent. If you have bad data, it doesn’t matter what AI you do on top of it, it’s going to fail. And then the second thing is, you have to educate everybody in your company as you start bringing AI in. Your information security team needs to understand AI. Your IT department needs to understand AI. The people who are going to use it need to understand it because if they don’t understand the difference between, “well, I’m talking to Mythos, versus I’m talking to Sonnet”, the cost of your use of AI is going to explode. Education is absolutely critical across the board, not just among whatever AI team you’ve got that’s developing a new agent. That’s a very small, very easy part of a much bigger process.

Scott Mann

And the third is, if you’re thinking about AI in retailer, reach out to Protegrity.

Clyde Williamson

Absolutely.

Jessica Hammond

I will second that third option, but I would also raise a fourth. Mostly just to come back to treating every AI agent like an employee, and the registration of that agent, the monitoring of that agent, these finely scoped permissions of that agent. Having a level of trust for that agent that is established, that’s external to just the agent saying, “Hey, here’s my ID, I am who I say I am.”

Scott Mann

That’s a neat concept, I like that. That’s an interesting way to think about it. 

The last question for this, you don’t need to be a retail leader here. AI is evolving really fast. What do you think the next evolution of enterprise AI will look like as organizations move more towards that autonomous data drive system and start adopting everything that we’ve talked about here today?

Jessica Hammond

It’s agent everything, agentic everything, autonomous everything. The businesses of the future are going to look very different. Our roles as humans in relationship to those businesses is going to look different. We’ll still be here. Humans will still be involved, and it will look different. The more we can prepare ourselves, much to the education comment that Clyde just made. The flip side of that is to educate yourself as a human in this world as this shift is happening. It’s really easy, easier now than it’s ever been to educate yourself on new concepts and ideas. I would encourage anybody and everybody to do that. But it’s autonomy everywhere, it’s in places that we didn’t really expect or didn’t imagine it. Coding and engineering is the primary surface right now. That’s the biggest win. It’s the biggest use case that’s fully adopted across a large sector of the industry. But there’s going to be more and more and more on some of those other places that haven’t picked it up in the same way yet.

Clyde Williamson

I think another area that’s going to be really interesting to watch is sovereign AI. This is a card I have right here for my little home machine. AI is expensive when you look at tokens. AI is risky because you’re sending all of your business information and business knowledge off to some third party and crossing your fingers that they aren’t training their models on it or using it later to be a competitor against you. There’s a lot of a questions about that. There’s security issues. I’m at risk. We’ve been talking this whole time, what happens with the PII data? Do I protect it before I send it to the model? A lot of those questions go away if I have an infrastructure in my organization that I host the model on. And models are becoming smaller in some cases. I have models running on laptops here at my house. I have models running on my video cards. A lot of times I can route things to those models and then only route out to the bigger models when I have an issue that really needs that level of reasoning. I think that more and more we will see organizations begin to build their own AI stacks because they want to control their full AI stack.

Scott Mann

That’s great point. I want to dive more into sovereign AI, but just for the sake of time, I think we might have to dive into another one in the future. This was great, I really appreciate it. Both your perspectives on this is amazing. 

To everybody listening in, thank you all for tuning in to PodMagic. Where simple, reliable IT meets real world impact. If you enjoyed the conversation, be sure to subscribe and share. And we’ll see you all next time.