Safe AI Data Resilience

Evan Kirstel (0:12)

Hey, everybody, fascinating topic today. Everyone is switching on Copilot and AI agents across Microsoft 365, and most companies can't tell you what data those agents actually reach. Great chat today with Rick from Veeam on what that means for backup and recovery and what to expect at Veeam's global launch, Veeam On on October 15th. Rick, how are you?

Rick (0:34)

Hey, Evan, I'm doing great.

Happy to reconnect with you and the audience here again. Yeah, we're all, all effort going to our big event, our Veeam On franchise event coming up in October.

Evan Kirstel (0:47)

Exciting times, you're deep in preparation. Before that, how do you describe Veeam these days for those who may not be familiar?

Rick (0:54)

Well, I love this question because I have it real simple. I mean, we keep good data safe from bad things. Officially, though, we go by enabling safe AI at scale.

We do that through products that are in the resilience. A lot of people know us from backup, data protection, the resilience domain, as well as security, compliance, governance and privacy capabilities that came from a recent acquisition, which is really the heart of one of the big things that we're doing at this event, which is really starting to have some of those different technologies really converge and become one offering in the market. So it's a wild time to be green.

Evan Kirstel (1:33)

Hey, it is. I love that. So let's dive right in when customers take a first look at their real AI stack, the harness, the data set up, the whole nine yards.

What surprises them the most?

Rick (1:46)

Well, there's simply, Evan, there's an untenable amount of data. And so I think the market has to really put some intentionality behind understanding what they have. And there's going to be massive questions like, where is sensitive data? What agents have access to what?

Is the security even configured correctly? And these are very dangerous areas to go with bad answers on the edge of more and more AI projects. And then, of course, we have to be ready for things like hardware failures, operator accidents, fire, flood and blood, ransomware. All these things are additive. None of them have really gone away, but I think there's a new era of risk that we're helping people remove some of that risk.

Evan Kirstel (2:36)

Fantastic. So everyone's in a hurry to roll out Copilot and AI agents. I've got two running in the background. As we speak, I hope they don't crash the machine here. But beyond practitioners like myself, what mistakes do you see them making in these deployments?

Rick (2:51)

Well, there's a lot of what we call toxic combinations, which would be, let's say, for example, obsolete data that shouldn't really be used, and then maybe sensitive data or even worse, sensitive obsolete data fed into anything. It could be as simple as an agent that an individual makes or could be as powerful as an AI factory, okay? So at any level, we have these potential toxic combinations of bad data being ingested to machine speed solutions, and that's going to put AI project success at risk, it's going to put the organization at risk, and then wouldn't it be great if we had a solution in the market that could maybe archive that storage off of that data, off of primary tier one type storage?

And that's where really the Veeam that a lot of people have known over the years can come in, and we have capabilities to archive data off so that when folks go into AI initiatives, we're not feeding it obsolete data, we're not feeding it sensitive data to create some of those toxic combinations that could potentially happen.

Evan Kirstel (4:02)

Yeah, that's really interesting. There's a new buzzword on the scene, Rot, redundant, obsolete or trivial data. I'm not sure if you guys coined that, but I'm going to steal that. And on one hand, Microsoft 365 is dominating the enterprise. Just been on a tear. Congratulations to them.

But this problem of Rot, of all this data, is catching up with people. Tell us, why does it matter as AI pilots are rolled out, as AI agents are rolled out?

Rick (4:32)

Well, the first thing I want to, first of all, challenge the market is that everybody has Rot data.

Bold claim, but we've got data points to prove it. And when I say everybody has it, everybody has some amount of it. So I think it's a really good place to start, to kind of understand some of the new complexities that Veeam is out there to solve. And obsolete data is really tough because it works backwards to a phenomena that we've had for years, which is someone, okay, agreeing with everyone, let's delete something. That's kind of hard to do, right? It's kind of hard to get stakeholders to agree to delete, because it's like, well, what if we need it? And we live in a world now where we can have these archive class storages that provide the absolute lowest cost for that just in case type of moment. But it really protects things from being obsolete data, things from being fed into models. And I almost akin it to throw in a piece of gravel into a machine that's doing something important, like bottling water.