Redpanda CEO on Why Streaming Data Powers the Future of Agentic AI

[00:00:00] - [Speaker 0]
A quick thank you to NordLayer for supporting the podcast and helping me make these daily conversations possible. And if you are listening and you're responsible for security or IT, you will know the reality. The reality that most of your risk now sits inside SaaS apps and browser activity. That gap is exactly what NordLayer is addressing with its new business browser. So instead of bolting security on from the outside, it builds it directly into the browser itself.

[00:00:33] - [Speaker 0]
This means you can control access, monitor activity, enforce policies, and reduce shadow IT all from one single place. And most importantly, it does it without adding deployment headaches or complex onboarding. You get things like browser based data loss prevention, SaaS access control, and zero trust browsing, but delivered in a way that your team can actually use. So if you've been trying to simplify your stack while improving visibility, please check it out at nordlayer.com/browser. Have you ever wondered what actually powers AI agents behind the scenes, especially when they are expected to act in real time rather than simply just respond?

[00:01:25] - [Speaker 0]
Well, today, I wanna go right into the heart of that question with someone who has been building the infrastructure making it possible. He's the CEO of a company called Redpanda, and that company began life as a streaming data platform, and it is now shaping what it calls the agentic data plane. Let me explain why this matters. There's a lot of noise around AI agents right now, but very little clarity on how they access, process, and act on enterprise data, both safely and at speed. And Alex is working on that layer that most people never get to see, where latency, throughput, and governance quietly determine whether AI delivers real business value or ends up being just another pilot that never scales.

[00:02:17] - [Speaker 0]
And when you hear about some of the performance gains like 5.5 times lower latency and over 70% higher throughput when compared with NVIDIA Vera, you start to see how infrastructure choices directly impact what AI can do. So today, we'll learn more about why streaming data is becoming the backbone of agentic systems, exactly what an agentic data plane looks like in practice, and how enterprises can rethink their data architecture if they want AI to move from experimentation to execution. And Alex will also share what he's learning from working with global organizations, and they're working with some of the biggest names in the world right now and helping them process hundreds of terabytes of data every day so that real time access can become the difference between insight and real action. So if you've been hearing a lot about AI agents but wondering what makes them work in the real time, what a kill switch might look like on an agent if it goes rogue, then hopefully, the plan is that today's conversation will help you connect the dots. So how ready is your data infrastructure for a future where software doesn't just respond but acts?

[00:03:37] - [Speaker 0]
Well, before you answer, let me introduce you to my guest today. So thank you for joining me on the podcast today. Alex, can you tell everyone listening a little about who you are and what you do?

[00:03:49] - [Speaker 1]
Sure. Thank you for hosting me. So my name is Alex. I'm the founder and CEO of Redpanda. Currently working on this concept that we invented called the agentic data plane.

[00:04:00] - [Speaker 1]
We really tend to just help global fortune 2,000. So I think, like, gnarly, large scale, complicated companies. That's where we tend to deliver the most value. Yeah. Based in California, I have three young boys.

[00:04:14] - [Speaker 1]
And so between the company and the three boys, I keep busy. Great to be here.

[00:04:19] - [Speaker 0]
It's great to have you with me. And if you've got three boys, then you must be incredibly busy. They must take up even more time than, your company there. But, I mean, there was a lot to unpack in what you said because RedPanda actually began as a streaming data platform. Now you're talking about an agentic data plane, which sounds incredibly cool, but there's gotta be a story there.

[00:04:40] - [Speaker 0]
What changed in the market that made this the right moment to introduce this new layer connecting enterprise data to AI agents. Feels like an exciting moment for you guys, but tell me more about that and the story behind it.

[00:04:53] - [Speaker 1]
So, you know, seven years ago, Red Panda has evolved really organically. And so I wrote the original piece of RedPanda for me. You know, it's like one of the freeing things of being an engineer is you you just type the code. You don't have to ask people for permission. And so I had left Akamai where I was working on large scale distributed systems.

[00:05:19] - [Speaker 1]
And one of the problems I kept wanting to solve was why is the world so complicated? And they're now one of our partners and customers. We help protect you know, they protect whitehouse.gov by way of example, and we helped our security division power their their logs and other strategic things. We also announced a strategic partnership with them and so on. But it was it was a super organic thing.

[00:05:43] - [Speaker 1]
I was there as a principal engineer trying to solve a problem, and so I left, and I just wanted to codify what I thought the world needed at the time for me. And then we launched, and one of the largest social media companies launched with us at Telco in Europe, one of the largest electric car company. It was just from me ranting a little bit on Twitter, and I was like, okay. This feels like a real company. And so that was truly the the origins was never this grand scheme of, you know, trying to launch a hyper growth company.

[00:06:17] - [Speaker 1]
It was more as a builder just trying to scratch an itch that I have of trying to build a system that, frankly, I could understand that process large volumes of data. So that was the true early early beginnings of Redpanda, what is now known as the streaming company. And so I'll actually give you this well, to me, it feels like a cool evolution, but it's one journey of me, the person, and the company kind of all at at the same time. That was the original thesis. And what I was interested in and to me at least, is that, you know, for young companies, you usually don't start with the world's largest companies at first.

[00:06:53] - [Speaker 1]
You usually start out with, like, you know, startups, frontier people, early adopters, and then you move into, like, the mid market. People, they're like, okay. I see a bunch of proof points, and then you tackle the global fortune 2,000. That was not the case for us. When we launched, we just kinda went straight into the g two k.

[00:07:12] - [Speaker 1]
It was it's kinda where where now our customers felt that we deliver the most value, and and maybe some lessons learned from Akamai really helped with with that initial journey. And so you're right. Red Panda started out as this streaming company, but we're now seven years old. And so we've since added bought a couple of companies, three or four companies in between. We've only made news of two big ones where we added connectors as a way to bring in data into the platform, so connectivity, like, okay, a system.

[00:07:49] - [Speaker 1]
So let's say the New York Stock Exchange is a great example. So they're one of our customers, public reference. And so we're helping them move market data, clearly, you know, at, like, Google level scale in in some capacity. And so the streaming engine is here to help people, you know, process basically the the data movement for them. The connectivity is there to help bring in data in and bring data out to other connectors where there's, like, I don't know, the Snowflakes and the Databricks of the world and a bunch of other data warehouses.

[00:08:23] - [Speaker 1]
And then last year, I bought a database, and so it was sort of this, like, natural evolution of what I call tier zero, infrastructure designed for truly mission critical systems that gave us the opportunity to tackle infrastructure in the age of it.

[00:08:41] - [Speaker 0]
I just love the entire journey behind it. You're incredibly humble because what started as a a rant on Twitter, and I think that's where many of us we've all done that, but then we leave it there. We let somebody else go out and fix it. So I love the fact that you've gone out there and created a solution, given the world what it wants. So kudos to you there.

[00:09:00] - [Speaker 0]
And and, obviously, from there, you've built this hugely successful company working with some of the biggest names in the world and now evolving with that to AgenTica AI. Did you ever think it would turn out this way? Because it's phenomenal what you've achieved.

[00:09:16] - [Speaker 1]
You know, from the outside, it feels like the dots connect. But as a as a founder, there's I don't know if there's ever an easy moment. It's just always, like, filled with anxiety and and pressure. It's like, you know, you're never doing enough. It's it's the the feeling that you have at all times.

[00:09:35] - [Speaker 1]
And so that, like, gut wrench is very real for for me and really for most founders. It's like, know, am I doing enough? Am I doing this? And it's only when you take a long time horizon where it feels like you've done a little bit. At least you've made the world a little bit better in in in some ways.

[00:09:53] - [Speaker 1]
And so one of the things that warm my heart my mom called me when we were at the AI is easier to explain to her, but but we ship Red Panda to outer space. Not we. Like, one of our customers put Red Panda at a satellite and and ship to outer space. And when I was a kid, you know, like, we all go to preschool, and they're like, who do you wanna be when you grow up? And I think a lot of STEM folks, they're like, I wanna be an astronaut or whatever, a firefighter, this.

[00:10:22] - [Speaker 1]
And so I wanted to go to outer space. And so I called her. I was like, well, I did never made it to outer space, but the software that I wrote made it to outer space. And it's so there was this, like, moment that but only when when, like, you look at it at a longer time horizon where, you know, it makes you feel proud of the work that you've done. And and, you know, obviously, as the company has grown, it's largely the team that has done this work.

[00:10:50] - [Speaker 1]
You know, I've I've here to kind of help us steer in the right direction and make sure that, you know, we keep money, we can pay people, and our customers are happy. It's kind of my role has shifted away from engineering to being a company leader. But, anyways, to answer your question directly, it only feels that you've made progress when you look at a long time horizon. When you're, like, in it in the day to day, you usually focus on the things that aren't working. And so so so it's only when we get to these moments and we reflect, you're like, oh, yeah.

[00:11:19] - [Speaker 1]
We've done some cool things with some of the world's largest companies.

[00:11:23] - [Speaker 0]
Man, it's it's added even more beauty to the story. I'm now thinking of that young boy staring up at the at the universe there, and now you put software up there. Fantastic. But to bring us back down to earth for a moment, I mean, a lot of business leaders will hear terms like agentic AI, real time systems, understand that, but they will struggle to understand why streaming matters. So in simple terms, why does streaming data from the foundation for AI agents that need to reason and act in the moment?

[00:11:52] - [Speaker 0]
Tell me more about that, what that means.

[00:11:54] - [Speaker 1]
I was talking to a partner at HyperCloud yesterday, and and their reaction as a customer there, like, this just feels like the not the obvious next step for you to do. And as a creator, it doesn't feel that way. You you struggle. So anyways but but the goal was that what as CIOs and regaining remember that the context that I shared earlier is that somehow, and this is one of those things that it feels harder to explain than to, like, live through. We found a great fit for the large, like, global fortune 2,000 that just tends to be the people that come to us, that inbound, that ask me to have a conversation with their CIOs.

[00:12:34] - [Speaker 1]
And so one of the CIOs, a public company yesterday, was like, this is such an obvious evolution of your story because what we are struggling with is governance for agents. And so there's this huge parallel with agents and, basically, workforce equivalents. And so an example would be an accounting agent that keeps your books up to date, a customer success agent, a coding agent, a security agent, a penetration testing agent, and so on. And so I'll use those analogies here to highlight some of the risk. I guess if I were to give you an executive summary, feels like the world on hired a bunch of agents but forgot to onboard them, forgot to give them the right permissions to the right systems, forgot to tell them how to do this, forgot to set the guardrails.

[00:13:29] - [Speaker 1]
Like, don't spend a million dollars on on AI tokens. Don't contact this API. Don't do this. Don't do that. And so this struggle is very real for CIOs around taking this, you know, what is clearly the evolution of enterprises into production.

[00:13:49] - [Speaker 1]
And so for a manufacturing company, they have humans with full time jobs. This is this will blow your mind. It's to basically move data from one system to another system, and it's like these huge bespoke things because, like, the these spindles and and whatever. There's, like, a lot of complexity and regulations and these chip manufacturing companies. It's all nation state secrets.

[00:14:13] - [Speaker 1]
And so, like, you know, you can't share any of the basics. Like, you can even share, like, hey. This database sucks to that. And so what happens is that every manufacturing company sort of, like, reinvents all of the things all on their own. There's very little information sharing in that industry.

[00:14:29] - [Speaker 1]
And so I say this to highlight the point that people were just, I don't know, still are almost gasping for an hour. It's like, how do I give agents how do I accelerate my business without exposing customer data? I know my engineers and my CEO and everyone is asking me to take my business, you know, in an agentic transformation, but I can't do that without risking leaking all of the customer data, all of my plans for the chip manufacturing, all of, like, you know, just all of the intellectual property. What makes a bank a bank in some ways is not that they have checking accounts. It's the customer relationships.

[00:15:08] - [Speaker 1]
Right? And so that is the true data. And so the problem that we're trying to solve, the one liner is we're here to mediate access to private data for agents. And then everything else will kind of it's it's it feels it will feel like a natural evolution of what streaming we were doing, the connectors, the database into that. Does that make sense?

[00:15:30] - [Speaker 0]
Yeah. It does. I mean, you described the agentic data plane as this almost unified, governed access layer for AI agents, but you mentioned a few other problems there. But any other problems that this will solve for enterprises that already have data spread across, let's say, multiple systems, teams, and environments? It'd be great to really drill down for that listener on on the problems that we're solving here.

[00:15:54] - [Speaker 1]
Yeah. So this is like Elizabeth Kidd's poem. Let me count the weights. There are four or five, like, really core core pillars, and I'll define categories, and I'll give insight into the specifics that we do so that we'll do two things. Hopefully, we'll evolve the thinking of the listeners into where the state of the art is for some of the most demanding companies in terms of, like, where they are at their maturity

[00:16:19] - [Speaker 0]
Yeah.

[00:16:19] - [Speaker 1]
How they're thinking about it on the scale. And so we're working with the German bank. They're trying to deploy 300,000 agents to production in the next eighteen months. So massive scale. Right?

[00:16:29] - [Speaker 1]
They have, like I think they have 20,000 employees, and so 300,000 agents is, like, an insane number of agents that they're trying to deploy. For them, it's really working on three things, trust, explainability, and context. So let's let's give you specifics on each one of these three pillars. So on the trust part, it feels like where the developer Mindhive is, where all of their thought leadership in on LinkedIn is, it's like this notion of on behalf of permissions, task based access controls, there's a bunch of standards being developed. And the umbrella of problems here is do the agents have the right access to the right data in the presence of some of these end systems, like databases, systems of records, like HR or, I don't know, ServiceNow or Workday, whatever.

[00:17:26] - [Speaker 1]
There's, like, a bunch of I know some of these companies do have fine grained access controls, but not quite to the level that an agent needs them to. And so on the access control parts, it's like, how do I carry the principal identity? So in other words, how do I carry Neil's HR identity so that when you ask the HR agent, like, what is the equity position of everyone at my company? It gives you the entire tree. But if an individual contributor in your company asks, what is my equity position here?

[00:17:59] - [Speaker 1]
If he or she only gets, like, that individual, but the code didn't change. What changed was the principle that asked for the information. And so carrying that information is a core, core piece. And so just just like access management, both authorization and authentication of agents, that is that is one pillar. As on the access controls.

[00:18:19] - [Speaker 1]
Explainability and context is maybe where we can expand, I think, the view of some of the listeners. So the explainability part, this one is we are phenomenal at this. So so this is probably how the whole piece will will come together and for I took was a computer science major, and I was fortunate or unfortunate. If you've ever taken an art history class, it's so hard. It's, like, harder than computer science in many ways because you have to memorize things.

[00:18:50] - [Speaker 1]
It's, the opposite side of your brain. Where in science, you I don't know. You figure out from probe and you do an inductive steps, then you solve a problem. In the arch, it's a really quite heavy memorization. So long story.

[00:19:01] - [Speaker 1]
There's this architecture concept, which was like the Roman keystone. It was basically a keystone a piece of stone in the middle that had a particular gradient of arch, and that defined how big the arch is. Everything in governance hinges on this concept called agent transcripts. That is the keystone for Red Panda's governance story and something that we've been really good at for many years. So today, we power some of the largest observability companies.

[00:19:32] - [Speaker 1]
And what it is is it is really difficult for people to debug and prove the behavior of what their agents are doing inside their company today. And why is this hard from an enterprise? It's hard because you it is easy to do this in this, like, customer centric, one term based agent. So Neil takes a turn, the agent takes a turn, you take a turn, the agent takes a turn. That's super easy.

[00:20:05] - [Speaker 1]
It's easy to do. But in enterprise, you have this long chain of agents where agent a talks to agent b and agent b talks to agent c. And, you know, there's a long chain and then back into the user response. Now remember that the whole point of AI is that you don't write the business logic. You say, if your name is Alex, you should deny the credit card application.

[00:20:24] - [Speaker 1]
If your name is Neil, you should approve it. Right? Like, you define these high level goals, and you're outsourcing the business logic. You're basically it's you know, an agent is just a set of tools, a prompt, and and some tokens that you purchase. And so you've offloaded your business.

[00:20:38] - [Speaker 1]
You codified your business in, basically, in a set of basic prompts. And then and then you get back a response. You're like, okay. What is this agent doing? And did this one agent and this chain of agents get hacked, or why did the agent get this, like, erosion problems whereby either somebody added a column on a database or they added a data source.

[00:20:59] - [Speaker 1]
If you're a finance agent, and so you're, like, added the Bloomberg terminal data, and all of a sudden, the responses are different, and it doesn't really make sense to you. You're like, where's this agent getting data? And so explaining that is effectively an observability problem when it comes to system levels, and we have been great at that for seven years. This is really the this is the pivotal insight for me on the streaming side that allows me to show up to these companies authentically. We're like, no.

[00:21:30] - [Speaker 1]
No. We've had experience running these kinds of workloads of, you know, tens of gigabytes of of data per second, tens of thousands of agents or topics or partitions or whatever at massive scale for the world's largest businesses. This was this was it. And so this is known as explainability in AI, and it's the idea that you can take an agent and explain every single action that the agent did, let's say, five for years or six years. And so that is one of the most important things that we did.

[00:22:04] - [Speaker 0]
A huge thank you to Donodo for supporting the Tech Talks Network and helping power more than 60 interviews every month. If your business is already invested in data lakehouse, this next message is especially for you. Because if you've already invested in a data lakehouse like Snowflake or Databricks, Denodo will help you get more value from it. Faster AI, smarter analytics, and trusted self-service data access. So turn your lake house investment into real business outcomes with Denodo.

[00:22:39] - [Speaker 0]
So please visit denodo.com. It's incredibly cool. And before you join me on the the podcast today, I was also reading about Red Panda's work with NVIDIA Vera, which some great stats there as well. It showed 5.5 times lower latency and 73% higher throughput, which is phenomenal. But for anybody listening that maybe are more of a business leader outside of the infrastructure world, what what do these gains actually mean in practical business terms for performance, customer experience, or operational speed?

[00:23:13] - [Speaker 0]
Because it's a big deal, isn't it?

[00:23:15] - [Speaker 1]
Yeah. So Lissac, the NVIDIA partnership and then we'll come back because I wanna explain this idea of an agent kill switch, which I don't think the world has, like, explored. Still a business a business level concepts.

[00:23:29] - [Speaker 0]
Yeah.

[00:23:29] - [Speaker 1]
And so on the NVIDIA side, Jensen was on television a few months ago. And they cleared they own the GPU business. Hence, they're the biggest company in the world. They're kind of like, you know, market makers at this point. They they get to say where the market goes.

[00:23:46] - [Speaker 1]
They're that influential, and who gets GPUs is very influential. It's like, who gets to make money? And so the it's it's it's sort of incredible the impact that NVIDIA has in in in the world today. Yeah. But it's not like that'll continue forever for lots of reasons.

[00:24:07] - [Speaker 1]
I think maybe power is gonna be a limited factor. I think today is still still it it's my opinion, not not their opinion. Chip access remains a bottleneck, but I think soon people are also realizing that power. There's not enough power to fit these computers, which is why people are exploring taking these GPUs to outer space so that you have infinite power effectively. And so that's kinda where the world is saying.

[00:24:33] - [Speaker 1]
The evolution to keep these GPUs fed is you need us a new set, basically a new chip that is designed for the true bottlenecks in a lot of modern software today, which is the memory bandwidth between the CPU and the main memory bank. And so Verra, in our experiments, has phenomenal memory throughput performance. Their way it shows up in applications, in customer success agents, in in all of this, in basically user visible workloads. It tends to be in slightly reduced total cost of ownership, TCO, but more importantly, it results in often better user experiences. So one of the cost dominant factor for us so I bought a database last November.

[00:25:27] - [Speaker 1]
We announced it at the New York Stock Exchange. It was it was great. My CTO joined me from Snowflake. His name is Tyler Akidaw. He's worked on the largest databases in the world.

[00:25:37] - [Speaker 1]
And so he joined, we we acquired a database company. And one of the dominant workloads is joining foreign IDs. And so I know it sounds very trivial, but the insight here is you have all these agents doing things. Okay. Great.

[00:25:56] - [Speaker 1]
And you need to make sure that you decommission them in case they misbehave. And so every action also has a different ID. And you have tens of thousands of them, so maybe a 100,000 of them, and they're all acts in a bunch of tools. And so a very cost dominant factor from a CPU perspective is how do you join these IDs? How do you aggregate them in a way that's cost effective, One.

[00:26:18] - [Speaker 1]
But two, it's also real time. It feels like you're interacting with the, you know, with the live system. And so our experiment showed two important use cases to highlight here with the Vera CPU team. One is our database simply performed 10 x better than the state of the art at, like, joining what is known as agent traces with, like, tools and prompts, etcetera, so that we can give insight to people, like, what are these agents doing. And and then secondly, when a large scale provider, it is just a much lower TCO, and so you could do more with less, so that's important.

[00:26:56] - [Speaker 1]
The power consumption is slightly less than comparable architectures, and this higher memory throughput allows us to have this huge visible impact in workflows, like joining agent IDs with with, like, traces and tools, so that we can give insights into what these agents are doing. In case that anyone misbehaves, we can just decommission them. You're like, why is this credit card agent approving all applications? We're gonna go bankrupt. So after you see a couple traces, you can just decommission the agent so that at least your business doesn't go bankrupt.

[00:27:31] - [Speaker 0]
Phenomenal. I mean, you've already mentioned quite a few big names, but, I mean, you're working with so many of them. Are these other companies that I wanna just name drop here as well? Cisco, Moody's, Vodafone, Texas Instruments, and major US banks, but, you don't have to name any names here. But to further bring it to life, any of the use cases that you're seeing today where streaming data and agentic systems are already making a measurable difference?

[00:27:56] - [Speaker 0]
Anything that you can share there that will just hammer home the points that we're making here?

[00:28:01] - [Speaker 1]
Yeah. So let's drive the point of one of the largest German banks and one of the top chip manufacturing companies in the world. So we're working with both of them super, super closely. And for the German bank is to prove to the Federal Cartel Bureau, which is my civilian understanding of it, similar to what the SEC is in The United States. So they have to prove to this organization that their customer success, because it's, like, mission critical and that's basically about as much as I'm allowed to share Yeah.

[00:28:45] - [Speaker 1]
Has, like, end user, you know, like, and I impact in financial decisions. So there is a new agent that they're launching. And we just have to prove to the world that the agent is strictly more helpful based on average statistics across the fleet than the human counterpart. And so it's here because they're trying to improve effectively money exchange and money transfers. It's basically kinda how banks get paid.

[00:29:14] - [Speaker 1]
And so if you can reduce the speed at which information is available to people, then they don't go so let me give you an example. If you're a wealthy individual, you're probably gonna have, I don't know, a few bank accounts. And so you go in and you're like, okay. I'm gonna use this bank. And if it's kind of annoying to get this information, just go to a different bank or you, you know, whatever.

[00:29:33] - [Speaker 1]
Maybe you tell your financial adviser to wire it to someone else. And so there's this this like strategic set of financial information that people need to access super promptly. And so what we help this German bank do is prove, and so record, all of these agentic interactions for two things. One, for evaluations, so that in case a new model drops tomorrow, like in case, I don't know, Chad GPT comes up with GPT next version, then they have a they have a portfolio of recorded actions that give them confidence that they're not gonna ship a terrible user experience. So that part is called evaluations.

[00:30:16] - [Speaker 1]
So that's a core thing. It means their product remains good. But two, they also have to prove to the government that that the agents are doing helpful things. You know, like, you sort of define the parameters, because it it touches, like, real humans, then they have to there's a huge approval process in Germany, more complicated than in The US. And now you you have a place that you can go back, and so should anyone ask you what happened in this, I don't know, Tuesday at 10:30AM, then they could just give you they're like, this is exactly what happened at you know, for this particular customer.

[00:30:49] - [Speaker 1]
And so that's explainability part. That's for a large German bank. On the manufacturing front, it's even more critical. This is where we shine, which is once you've recorded this agentic interactions, this idea of what Anthropic calls agent transcripts, which we've adopted that name internally as well, this end to end execution logs of agents. So you record them all across this enterprise.

[00:31:15] - [Speaker 1]
It is so costly. So effectively, manufacturing companies care about yield. How much does it cost them to produce a working wafer effectively? That's like kind of everything is centered towards productivity of cheaper chips. And the margin, these people aren't floated on software margins or of like 75 to 90% gross margins.

[00:31:38] - [Speaker 1]
Like, the margins in manufacturing are small. You know, the the numbers are large, and so they still make an inordinate amount of money. But but relative margins are are small. And so what we're doing is, like, if an agent misbehaves, the cost of that is actually months on on a wafer production. So if you think about when a wafer starts the the batch production pipeline, if you don't detect the error fast enough, you will ruin about two months of not just machine labor that maybe you can afford, of human labor, of, like, physically moving things.

[00:32:13] - [Speaker 1]
There's still a lot of physical human labor around this cheap manufacturing. So for them, it's not just the explainability, it's the fact that we can do what we call agent kill switch, which is we'll just decommission the agent, like, within, you know, seconds on when we detect a particular action. As soon as we get the signal from the manufacturing plant, we help them decommission these agents. And so those are very real, very, like, revenue impacting agentic evolutions, where streaming was the core piece. Right?

[00:32:45] - [Speaker 1]
Streaming is all of these agents are streaming these logs and executions and transcripts into the original Red Panda storage engine, and then we're able to bootstrap this whole AgenTik governance from that point.

[00:32:58] - [Speaker 0]
It's amazing. I've been to, what, 20 plus tech conferences over the last twelve months. I've heard everything about all things AgenTik AI and thousands of, agents being unleashed into the wild, and I've often felt a little bit cautious and nervous about this. And you're the first person that I've heard say, we about the agent kill switch. So there's gonna be a lot of people interested in that.

[00:33:20] - [Speaker 0]
And, also, people listening gonna wanna find out more about the agentic data plane platform that we've mentioned there, this unifying governed access layer that connects data systems to AI agents. We can only cover so much in our podcast. So anybody listening wanting to dig a little bit deeper, find out more information, immerse themselves in this world, and also keep up to speed with some of the inevitable big announcements that are gonna be coming out in the months ahead. Where where should people go to, stay up to speed with everything?

[00:33:50] - [Speaker 1]
Yeah. The easiest way to some super accessible if you wanna connect with a human. Right now, we're kind of oversubscribed. We we launched this project, and we had to, like, be like, hey, there's a there's a waiting list because we are just kind of trying to keep up with the money. It's a it's a champagne problem, so I'm not complaining.

[00:34:12] - [Speaker 1]
But, you know, if you go on LinkedIn and you send me a DM, I will like connect you to the team and we'll try to see how, you know, we can help you. That's probably the fastest way. The second, and maybe a better way for everyone, is if you go to ai.redpanda.com, you'll you'll get a better, more textual description of what this is with a slightly more technical intuition for those of you that are looking to understand, like, okay, how do the hours and boxes actually connect? And then there's a form there. My team monitors that.

[00:34:43] - [Speaker 1]
There's a bunch of escalations on that, and so if you fill the form, we'll get it. Those are the two best ways for people to to reach out to me. If you want a fast response, I'm pretty responsive. So DM ing me or my CRO or, you know, my head of customer relationships on LinkedIn is probably the fastest way, just given some of the demand. Because you're right, you know, we are I don't know if if like, look, I'm not saying that we're the only company that could do this here, but what we are saying is that we were able to put all these pieces together so it's easy for people to do.

[00:35:18] - [Speaker 1]
And as far as I know, we were the first company that launched this concept called the agentic data plane. Right? It was a thing that didn't exist in the world. And it sort of gave people really easy way to articulate the problem that they're trying to do. Like, your agents are here trying to access private data.

[00:35:32] - [Speaker 1]
We're gonna help you with that. We're gonna do it in a way that's safe. We're gonna do it in a way that helps you kill these things or decommission them if they misbehave. And so yeah. So, anyway, so it's starting to get busy, but we'd love to to help and partner with anyone listening in.

[00:35:50] - [Speaker 0]
Absolutely. Love it. And if everybody listening, if they were to take away one thing today, would be to I would say go out there, look up AgenTic Data Plane, ADP. Immerse yourself in that world. There's some big, big takeaways.

[00:36:02] - [Speaker 0]
You have the attention of the world's biggest names. You're working alongside Ahmed. That speaks volumes alone. So I wish you the best of luck with your champagne problems for the rest of the year, but I would love to stay in touch with you and see how this continuously evolves. I got a feeling in my waters, as they say over here, it's gonna be a massive year for you, and I think it's only gonna snowball.

[00:36:25] - [Speaker 0]
But let's stay in touch. I'd love to get you back on next year, see how things are moving. But more than anything, thank you for sharing your story today. Really appreciate you.

[00:36:32] - [Speaker 1]
Thank you, Neil. Thank you for having me. I appreciate it.

[00:36:37] - [Speaker 0]
So what did we learn from this conversation? For me, it comes back to a simple shift in thinking. AI agents are only as effective as the data systems they are connected to. And if that data is delayed, if it's fragmented or poorly governed, the outcomes will reflect that. And I think Alex offered a very clear perspective on why streaming is moving from a technical preference to a business requirement and why real time data is no longer just a luxury for a few use cases.

[00:37:08] - [Speaker 0]
It's almost becoming a baseline expectation for systems that are expected to reason and act. And at the same time, there's also this growing need for guardrails because giving AI access to enterprise systems without governance introduces new risks alongside those new opportunities. So it seems that there's an interesting tension here. On one hand, companies want faster and more autonomous systems, and on the other, they need control, visibility, and trust. So the idea of an agentic data plane, I think, sits right in the middle of that tension and tries to offer both speed and structure in a way that feels usable at scale.

[00:37:52] - [Speaker 0]
But it also raises a bigger question as more and more organizations invest in AI agents. Will infrastructure quietly become the deciding factor between those who succeed and those that stall? And more importantly, are businesses truly prepared to rethink their data foundations to make that shift happen? As always, techtalksnetwork.com. Keep your thoughts, questions, and anything you wanna share with me, keep those coming over, and I'll return again tomorrow with another guest.

[00:38:27] - [Speaker 0]
Hopefully, I'll get to speak with you all again then. Bye for now.