Cribl on Why 96% Want Agentic AI But Only 23% Are Ready For it

[00:00:00] - [Speaker 0]
If you are running a business right now, you may have noticed there's a quiet shift happening, one that most people are still underestimating. And that is your company doesn't live inside your network anymore. It lives inside the browser. That's where your SaaS apps sit. That's where your data moves.

[00:00:19] - [Speaker 0]
And increasingly, that's where attackers are focusing their attention. So NordLayer has just launched its new business browser, and it's designed specifically for small and medium sized companies that need visibility and control without the overhead of enterprise security tools. What I like here is the balance. You get advanced protection, better compliance, and full visibility into how your team is working online, but without slowing anyone down or forcing them to learn anything new. Please head over to nordlayer.com/browser and check it out, and let me know your thoughts.

[00:00:58] - [Speaker 0]
But now on with today's show. Every business leader now wants to move faster, cut costs, and improve decisions. But what happens when the infrastructure underneath it all can no longer keep up? Well, today, I'm joined by Abby Strong. She's the chief marketing officer and chief customer officer at Cribble.

[00:01:22] - [Speaker 0]
And together, we're gonna talk about the hidden cost of Agentic AI. Because as AI agents create more queries and more pressure on already stretched systems, many companies are discovering that the real challenge is not ambition. It's readiness. So our conversation today will get into the data layer behind AI and understand why observability now matter more than ever And why that path from pilot to production would depend on your infrastructure, and how that infrastructure can support the scale, the cost, and the trust demands of enterprise AI. And we'll do it all in a language everyone can understand.

[00:02:06] - [Speaker 0]
So enough scene setting for me. Let me introduce you to Abby right now. So thank you for joining me on the show today, Abby. Can you tell everyone listening a little about who you are and what you do?

[00:02:18] - [Speaker 1]
Sure. I'm Abby Strong. I am the chief marketing officer and chief, customer officer at Cribble. I've been here nearly six years, since we were sub $5,000,000 in revenue and just passing our recent milestones, you know, crossing over 300,000,000. And, and watching the company grow from just less than 30, what we call Cribble GOATs, all the way through, we're over a thousand employees now across the globe.

[00:02:45] - [Speaker 0]
And for people listening and hearing about Cribl for the very first time, how would how would you describe everything that you do, and what makes you different from other, solutions in the market there?

[00:02:55] - [Speaker 1]
Yeah. Cribl is the, the pioneer in what's known as the observability or telemetry pipeline. This is really connecting unknowns, any source to any destination, any type of, telemetry data, logs, metrics, traces, dealing with that, you know, the massive volumes, the the variety, and, of course, the unknown value where a lot of this data is kept, just just in case. But in the recent years, what we've really been focusing on is how do how do we build the AI platform for telemetry and give users the ability to really customize the experience that they have in interacting with this data, deal with that fundamental tension of data growth being at 30% and and budgets, you know, maybe sub 10 if they're if we're lucky and a lot of times contracting. And so how do we make sure that folks can deal with that those volume, variety, and value challenges facing them even in this world of AI where these it's, you know, dramatically increasing above that 30% marker already.

[00:03:56] - [Speaker 0]
And I've been following your work for some time. I love what you're doing here. And there is at the moment, I think, broadly, there's a growing sense that many organizations have rushed into generative AI experimentation over the last two, three years with without fully understanding the infrastructure implications. And I'm curious, from Cribble's perspective here, what what is that biggest gap between executive AI ambition and the operational reality that you're seeing right now?

[00:04:22] - [Speaker 1]
Sure. I mean, this is one of the biggest challenges that are facing our customers today is that, you know, we are seeing customers tell us mentioned that 30% marker. That's an IDC value saying that, you know, data, telemetry data has been growing at 30% year over year, and that's pretty conservative. What we're hearing from our customers is that, over 76% of them are saying that their telemetry has potentially doubled or tripled or more, since introducing AgenTex solutions into their, into their environments. And so if you think about it, a lot of times, the value proposition of these solutions is, hey.

[00:04:58] - [Speaker 1]
We're gonna make your you know, we're gonna offset the cost of your tier one SOC, or we're gonna help you troubleshoot faster. But all of that relies on having access to telemetry data, and that it's not just the cost of the systems themselves that are they're struggling to keep up with, but it's the infrastructure required to deal with double or triple the amount of volume that they were dealing with previously. And so infrastructure costs are coming in significantly higher than expected, and these solutions are only capable if they have access to the data. And so we have many customers who have tried some of those new, SOC agents or SRE agents, and they turn them on, and they very quickly turn them off because they've made their, their telemetry infrastructure inaccessible to all of the regular users and workflows that they already had running in their environment.

[00:05:52] - [Speaker 0]
And I also wanted to highlight today the Harvard Business Review Research, which it recently found a massive stat. I think it was 96% of leaders say Agentic AI is strategically important. Maybe not too too much of a surprise there, especially of everything I've heard at so many different tech conferences this year. But here's the kicker. Only 23% believe they actually have the infrastructure strategy required to support it.

[00:06:16] - [Speaker 0]
So why is there such a dramatic disconnect here between the vision and the sales pitch that we see everywhere and the the actual readiness?

[00:06:24] - [Speaker 1]
That this is such a great question. Thank you so much. Like and that that I mean, like and the the challenge is mostly because the the cost profile is so out of whack for the value that they're getting from it. Like, if I say I can offset half of the cost of your tier one stock, but you have to quadruple or quintuple the the cost of your infrastructure, the the math is just so far off in the cost that they're not actually saving any money. They're actually increasing their costs fairly dramatically.

[00:06:53] - [Speaker 1]
And so the how do I make this successful and how do I get the data into a place that these agents can actually act on it, that they'll get a consistent answer? I mean, keep in mind, this is security and IT data. So, like, you have to be able to stand behind the answers that you're getting. But AgenTic systems, they're not exactly known for their consistency and determinism yet. Right?

[00:07:13] - [Speaker 1]
Like, they're still learning in these different environments. And so the the the differential between that cost, that is where Cribl has really focused is how do we help you collect, transport, transform, store, and then analyze that data and make it useful, not only to humans, but also to agents.

[00:07:33] - [Speaker 0]
And one of the most interesting findings is around telemetry growth there, which I know is a subtopic very close to your heart. And many organizations already struggle with observability costs, and AI systems are dramatically increasing data volumes yet again. So do you think companies are underestimating the the overall operational cost of AI itself?

[00:07:55] - [Speaker 1]
Oh, for sure. I mean, like, we're like like I mentioned earlier, that the conservative estimate was 30% year over year. Nobody's budget I've been all over the world asking folks, how many of you have a budget that's growing at 30% year over year? I've never had anybody raise their hand. Yeah.

[00:08:09] - [Speaker 1]
And and that was the before value. When we talk to the some of the analyst firms, they they come back and they say, we haven't really figured out how to even, put a put a number on the amount of telemetry growth that we're gonna see right now and and from from these AgenTic systems. And so they haven't even we we call it a levendy billion because we think it's funny, but there isn't really, like, an actual quotable value for how much how much growth they're seeing. What we do know is, like I said earlier, they're they're talking about doubling or tripling the volume that they already had.

[00:08:42] - [Speaker 0]
And we constantly hear about AI models, copilots, agents, but much less about the infrastructure under underneath them there. So have you seen observability and telemetry quite quite become one of the most important layers in the enterprise AI adoption? What what have you seen here?

[00:09:00] - [Speaker 1]
This is the basis upon which these run. Like I said, a couple minutes ago, these solutions are only as good as the data upon which they're operating. And so if you need to be able to stand behind the answer of saying, like, this is a is an incident. It's not an incident. How do you know that this agent is actually getting the right data, has the right context, and knows?

[00:09:23] - [Speaker 1]
I know a great example, it's not with telemetry data, but it's one that I had to live just yesterday, is I went into our, our business our business intelligence system, and I said, I wanna know how much ARR is associated with this product. Well, it turns out we have six different fields that say ARR in our in in in our BI system right now because, you know, somebody wanted a derived field and somebody wanted a field that meant, this is specific to the product, whereas it's overall. The agent doesn't have the semantic understanding to know which of those ARR fields is correct. So I asked the question, one of my colleagues asked the question, we all got different answers from this system because we didn't say this, you know, this is the field that matters when you are asking a question like this, which is how you have to structure. If you think about the the massive volume of of telemetry data compared to BI, daily BI on a good day, you might have, like, a a terabyte of landed data.

[00:10:17] - [Speaker 1]
Maybe maybe a petabyte would be one of their larger, installations. Our customers are doing that in a day. And so when you think about the volume that's facing users, and then now they have to create this semantic understanding around these are the fields that matter. This is what you know, I need to know if something happened here. This is what I need to look at.

[00:10:36] - [Speaker 1]
And then teaching all of the agents that that may be talking to other agents to get that information is, it's a pretty massive challenge facing customers. And like I said, that's that's really where we've tried to focus our time and attention is how do you get value from these systems? Like, the the promise of them is brilliant. The infrastructure is not ready for that load and the many hundreds of times, magnitudes greater queries that these systems or or these solutions are going to, to put on the infrastructure.

[00:11:07] - [Speaker 0]
And bear in mind, everything that we've seen and heard over the last few years, maybe it's unsurprising that ROI was identified as the number one reason AgenTek AI projects stall in pilot phases. But what is it that organizations are missing when it comes to measuring AI effectiveness, quality, governance, and and business outcomes at scale? Because as the old belts and braces mantra in IT goes, you can only improve what you measure. Right?

[00:11:33] - [Speaker 1]
Exactly. You're a 100% right. And it is it's that disconnect between, well, if I'm offsetting let's say I'm offsetting 10 people, but I have to increase my spend on the infrastructure to be prepared for this magnitude greater of queries for, creating making sure that it's all structured in a way that there's, an understanding, there's context as to which fields matter and why they matter, that the cost differential in having to prepare all of that data is significant if you're not using a solution like like Cribble. And so it's, it's really interesting when they're trying to prove out, oh, I can, you know, I can save on 10 bodies, but I'm going to you know, I need, you know, many millions more in infrastructure costs. It just, it it has been a struggle.

[00:12:18] - [Speaker 1]
And I think, like I said, the biggest the biggest challenge is the the load, in addition to the understanding of these agents is the the load that it puts on the infrastructure. If you think about it as a human, a really productive human, I might issue ten, twenty queries in an hour. Right? Like, what about this? And let me let me let me get an answer to this, and I'll dig in.

[00:12:39] - [Speaker 1]
A robot's gonna do that at a 100 times that. Right? Like, because it doesn't cost them anything. It doesn't it's not time. It's not just gonna keep banging at it until it comes back with an answer.

[00:12:49] - [Speaker 1]
And that is so significant that that the the volume of just the sheer number of, like, queries and things that it might issue into the back end is so much higher that it makes it inaccessible to any of the other things that we're running on top of that infrastructure. And that's that's really where they're struggling at the moment is is trying to find that that balance of how much they have to spend in order to make these the the solution actually work at the speed it's promised.

[00:13:15] - [Speaker 0]
And before you join me on the podcast today, I was reading how Cribble recently introduced integrations with OpenAI and Anthropic Claude compliance APIs. So why is visibility into AI activity? Why is this becoming such a a major issue for security and observability teams? Because I think this is something some, might have misstepped or or or not even noticed.

[00:13:37] - [Speaker 1]
Yeah. Well, I mean, I think everybody notices. They just don't know what to do about it right now. Yeah. Now we're talking about significantly greater volumes.

[00:13:44] - [Speaker 1]
We're talking about, you know, agents talking to MCPs that may be connected using, you know, either human credentials or potentially even, like, nonhuman identities, like, all connected to each other. And the question becomes, like, what do you monitor? How do you know what good is? What is what is normal behavior? Because there is no norm I mean, these these solutions didn't even exist six months ago.

[00:14:04] - [Speaker 1]
And so as you're trying to think about what does this, what does this look like and what do what what are folks, you know, how do they compare it against baseline? Well, there's no baseline anymore. And so it's a really exciting time. I think we're in one of the most exciting market transitions that I will ever get to experience in my in my lifetime, and I think many of us, you know, it's it's it's even bigger, I think, than the transition to mobile or potentially the, you know, the introduction of the Internet itself. It's a this this is changing how everybody everybody does work.

[00:14:38] - [Speaker 1]
And so but now you come back to most folks are using these frontier models, and we have to make sure that they have a way to start monitoring, you know, how are you using these? What is the sentiment analysis? What is you know, what kind of prompts are being issued? How how often are they doing it? How many for I know for me, one of my biggest things is how many tokens are we using a day?

[00:15:01] - [Speaker 1]
Like, what does my budget look like? You know? And so you have to so these integrations give our customers the ability to start monitoring all of the telemetry data that's interacting with these with these models. And that'll be the important first step. I think there's gonna be a lot of exciting solutions that pop up to help them deal with all these known unknowns that are coming out.

[00:15:20] - [Speaker 0]
And Cribl's growth numbers also suggest companies are prioritizing data management and observability despite the broader economic caution that we're seeing in our newsfeed. So what do you think or or what are customers telling you about the pressures that they're facing as AI adoption moves from those experimentation phases into production environments and and scaling from there?

[00:15:44] - [Speaker 1]
You know, I think what we're seeing a lot is that we we are going to need to help help customers. I think that the biggest thing that they run into at the moment is more compliance and governance regulations, which is again one of the reasons, like, we're doing all the integrations with the frontier model providers because folks are going to need to show that they are giving it their best effort to start monitoring all of this increased data and how these interactions are taking place. But the the biggest pressure I think that they're facing is that they're asking they're they're being told go faster. Don't get in the way. This is a market transition as I was talking about.

[00:16:22] - [Speaker 1]
And then but but make sure there's no breach. Make sure that there's no incidents. Help me troubleshoot. You have to do it faster. You have to do it more.

[00:16:28] - [Speaker 1]
And I think there's this tension right now between how do we enable folks to embrace this market transition, make sure they're not left behind. But then also, hey. I'm responsible for securing this. It's my butt on the line if, you know, something, you know, something goes awry. Or same thing, like, I need everybody to have a great experience, so I can't let one agent go and take out the infrastructure for everybody else, or or take down this application.

[00:16:53] - [Speaker 1]
And so I think a lot of the the challenges facing our customers right now is just like, how do I do this quickly and also make sure that I'm doing it in the safest and and most secure way possible.

[00:17:06] - [Speaker 0]
So a special thank you to Denodo for supporting the Tech Talks Network and helping us keep these conversations going because moving beyond AI pilots all starts with connecting your models to trusted enterprise data. So if you're ready to move beyond AI pilots, Denodo can help you connect your AI models to trusted enterprise data in real time. So you can scale faster and reduce risk. So if you're interested in turning AI into business value, simply visit denodo.com. And as we look ahead into the future, I'm curious.

[00:17:44] - [Speaker 0]
Do you think enterprise infrastructure will eventually be redesigned around those AI first operations? Are we still heading towards that period where companies continue to bolt AI onto systems that really were never designed for this level of automation and autonomous decision making?

[00:18:01] - [Speaker 1]
I don't think it's an either or. I think it's I think the answer to your question is yes. I think there's going to have to be both happening in in enterprises. I think we'll see the new whatever we're gonna call it now, the new digital transformation, which will be like the AI transformation, and we're all gonna be sick that term in in, you know, another six months. But but it's going to have to happen, in many of these.

[00:18:23] - [Speaker 1]
I don't know because it's but and you're going to see a lot of folks go through, like, a a, I'll call it, a maturity journey with AI where the, you know, there a lot of folks when they first embrace it, they're like, they do the bolt on. Let's make what I'm doing faster. Let but they're not fundamentally changing how they were operating. But as they go up that curve and it's like, okay. Now we've done we've added the AIs, bolt on, but why do we even do this anymore?

[00:18:48] - [Speaker 1]
Like, we these systems could talk to each other or this code could be generated and and pushed to production automatically and then tested against itself. Like, there could be all kinds of just complete workflow changes that, that are gonna come out of this transition. I think folks need to do a bit of the meet me where I'm at and do the bolt on solutions while they're rethinking how work happens, in their environments and how applications are created and and secured. So meeting them where they're at is our fundamental goal here at Cribble, which is how do we make sure that we can all of our customers get the value that they want and the choice control and flexibility for the choices that they're making today as well as the ones in the future.

[00:19:29] - [Speaker 0]
Well, I think so much of what you've shared today will resonate with business and tech leaders around the world. And for anyone listening that wants to carry on that conversation, dig a little bit deeper on some of the stats in the report and also some of the big announcements that could be coming out this year from Cribble. Where would you like me to point everyone listening?

[00:19:49] - [Speaker 1]
Oh, come see us at at cribble.io. It always has the latest and greatest, and we also have a community Slack that you can access if you wanna come in and interact with everybody from our founders to folks across company and and the Cribble community.

[00:20:03] - [Speaker 0]
Well, as we went through the the HBR report earlier, 96% of survey respondents said AgenTiKi is very important to their strategy. 23% said they've got a a strategy and the infrastructure to support it. And another stat from that report, 80% already expect their infrastructure will need to change because of AgenTik AI. There's so much great work you're doing about this in creating solutions for it, and I would encourage people to continue this conversation that we've started today and get involved on that Slack channel. Read more about it and keep it going because I think we're all in the same boat here.

[00:20:39] - [Speaker 0]
But more than anything, just thank you for sitting down with me, taking the time to start it, and, leave everyone listening with some valuable takeaways. Really appreciate your time today.

[00:20:47] - [Speaker 1]
Thank you, Neil. I've really enjoyed the conversation. Thanks for having us.

[00:20:51] - [Speaker 0]
I think one of the many things I'll be taking away after speaking with Abby today was that gap between AI ambition and operational reality. Because the the promise of Agenic AI is undoubtedly powerful. But as Abby made clear, the infrastructure cost can quickly overwhelm the savings companies ever hope to achieve. So if agents are querying systems at many times the rate of humans, costs will increase and teams are left trying to govern the behavior that has no clear baseline. And I think this episode is a reminder that AI success depends on far more than just choosing the right model or or building the right assistant.

[00:21:30] - [Speaker 0]
It does depend on data governance and visibility of what's sitting beneath it. Loved taking a look under the hood on this one. And for business and tech leaders listening, I think the message is simple. AI will, yes, change how your work gets done. But without the right infrastructure strategy.

[00:21:49] - [Speaker 0]
That pilot phase may be where too many good ideas get stuck. But what are you doing to avoid that? What are you gonna take away from today's interview? Please pop by techtalksnetwork.com. 4,000 interviews, eight podcasts.

[00:22:04] - [Speaker 0]
I'd love to hear from you. But that's it for today. So thank you as always, and I'll speak with you all again tomorrow. Bye for now.