The Next Phase of Enterprise AI Is About Experience | AppDevANGLE

32% of enterprises take hours to become aware of production problems. And that was before AI agents started running millions of concurrent sessions that no human team could ever watch in real time. My name is Paul Nashawaty, I'm the practicing and principal analyst, and this is the AppDev Angle. Today, I'm joined by Keith from Conviva. Um this is a company that it is a business of real-time experience intelligence long before AI observability became a conference buzzword. Keith, I'm glad to have you on the show. Great to be here. What's What Why don't you tell us a little bit about you and about the company? >> Great, thanks. Great to be here. Thanks for making that for having me. Um so first, me, I've been in technology about 35 years. And uh super exciting because right now I think is one of the most exciting times in my 35 years to be in technology with AI coming on and now we're trending, cuz I really believe it's not an evolution, it's a revolution. So I think the excitement, potential, the promise is massive. And so it's a lot of fun to be a part of it in the marketplace. And that's really where Conviva sees our our where we're going to play. And I think we we play a big strategic part in that is that we started the company 20 years ago really around internet business experience. That if you're going to put up anything out on the internet, you're going to put anything out in the digital world, you need to measure it. Not just measure the systems, but you have to measure is it performing at the end point the way I want to. And at that that time we picked video before the streaming market hit. And we were measuring video quality and showing video publishers that you need to understand the video experience. Not just whether it worked, it came on, but really the things that the human eye can see, resolution, pixelation, you know, buffering issues. So we were talking about experience when it came to video because you have to understand what the viewer experience looks like for you to have a successful video business. And now it's replicating itself as this new agentic future is starting to unfold as companies are looking at using agents to interface with employees or users and using agents interface with consumers. And again, it's the same fundamental promise that you can get the agent to work and it will come on and it will answer questions. But the reality is as a as a business owner, as a product owner, you really have to understand what is the users or the consumers experience with my agent. It's it is not enough to just understand the quality of service. You have to understand the quality of experience. So we're really excited. We built very successful business measuring every the largest streaming apps in the world today. Still big customers of ours. And now we're excited because we're at the forefront of this agentic marketplace and really positioning a very unique and and what we think is a differentiated agentic strategy. >> Okay, there's a lot to talk about here. I mean this this I'm excited to have you on. This is a real a really relevant conversation, but it's also you know, as we've been talking about in our briefings and conversations, there's a lot happening. A lot changes and organizations need to understand what their options are and what they can actually do. So let's set the stage here. When we look at this because the numbers are telling a pretty clear story. And in the last year's research, we saw that 90% of organizations plan to use AI agents by the end of 2025. And 79% of organizations anticipated worldwide world spread AI adoption within 3 years. And that is happening. It's happening real time. It's going now. But that massive wave is moving very fast. It's raising immediate operational question. Are agents doing the work and who's watching the agents, right? And that's something that we have to think about here. But when when we think about this, here's what makes the the the moment interesting from an analyst perspective, right? Enterprise teams are not building most of this themselves. 71% of organizations are sourcing agentic AI capabilities through platform vendors and 68% engage at IT or consulting service providers, while only 32% are building agentic AI capabilities primarily in-house. So, do you get a situation where the you know, majority of enterprises are deploying agent-powered experiences through vendor platforms, but the observability layer and the ability to to actually see what those agents are doing in the experience they're creating this for the end user, it's still a massive open question. Right? So, when we look at Conviva and what we're doing and you're filling the gap, you talked about um you know, tying these these insights. You have a sophisticated multi-agent pipeline in in the in the world. It's a it's a from what I can understand. Um you know, you're doing things that a lot of vendors are not doing. Um when and you're solving problems that people are doing, but when you look at this, what can you know, you you see from the experience of producing in real time, um if you're not seeing it real time, you're really flying blind. And that's not just a philosophical concern, that's a live operational risk that's already showing up in in high-stakes and production environments. So, Keith, question I have for you. When I I want to start with, you know, the current state. Where things are happening. Real-time observability is no longer a future problem. It's happening right now. We see it at World Cup ski, right? Um Yeah, so what are your thoughts here? >> That's a great That is a great starting point because I think the World Cup we're doing the World Cup right now for 20 different broadcasters and 20 different streamers right now. We're almost 30 million daily peak concurrents that we're managing as we speak. So, we understand large-scale deployments of any technology into a business or into the marketplace. And you touched on it. The real-time nature is absolutely critical. Anybody that deploys an agent for any purpose, by the way, the goal should be that you have a real-time connection to how it's doing what you think it's supposed to do. It's not a mature technology. And reality is as you mentioned, it's coming up fast and it's not an if, it's a when. There's no question now. But it's still immature, no matter when you pick your time that you're going to do it. Whether you use someone else's agentic framework, you buy an agent from someone else, or you build your own, it doesn't matter. The key is that that agent is going to be talking to either your employees, your users, or your consumers. And if you don't know what that is doing to those constituents, you could be in real trouble. And and and that's why we said in video, you need real-time analytics. You can't find out tomorrow that you have a video quality problem because people will just leave or people will stop you know using your service. And it's the same with agentic. No matter where or how you're deploying an agent, you have to have a real-time, not sample, but a real-time full census ability to see what's happening. Is it doing what I think it's supposed to do? And then beyond the fact that it's working, are people happy with it? Because you know, measuring an agent outcome is just one small piece of the puzzle. It could frustrate people that get to an outcome, which means they're never coming back. So, it's not sufficient to just say it worked and it got people to an outcome. That's not good anymore. You need to see as it's happening, what is the sentiment of my users with my agent? If it starts to fall, you don't wait for them to leave. You need to intervene immediately and understand what are the things I can do. Whether it's a third-party agent that you're using or or your own, you need to be able to diagnose and make changes as it's happening because that's your business. That's your users. So, regardless of where you got the agent from, the impact is yours. So, you should have the real-time visibility and control to see is it doing what I'm planning it to do or that I anticipated it to do or is it affecting my business in a negative way? >> And and when we look at this, right? This is actually a really important point here. Why are traditional models not working? Like if you're using traditional setups in like, you know, I've been doing this the way it's been working and we're analyzing the data. I know that you can say that automation and the scale is going is an issue, but it's more than that. Would you Would you agree? >> Yes, absolutely. And and by the way, there's a big difference between models and agents. All All the models are the big you know Death They They have all have everything. That's where everything operates. But enterprises, product teams, businesses should not be operating at a model level because now you're talking about a global update, a global intelligence update. Anything you share, everybody gets access to. And in particular, the models update much, much slower. They're great general purpose. They're great general intelligence, general AI capabilities, but the real tuning and the real enterprise value will come from Agentyc. Meaning that I build something that automates my workflow. I build something that talks to my consumers. I build something that calls my data sources separate from the model. So, that's a big distinction here is that anyone who's looking at using AI, and I hear people say, "Oh, AI." Okay, that's like a category. The key is, do you understand how agents work? Do you Do you understand what you need from an Agentyc perspective? Because the world will not operate on models. That's That's not where the majority of the business will be. It will not be where the majority of the automation comes from. A majority of the market and the world will be agents. And that's why Marc Benioff Salesforce said, "The future is about agents. There will be more agents than people in the world in the future." So, that's a that's where we believe you should instill the intelligence. It's where you should have the real-time visibility. And that's really where you should be betting on the right outcomes and building the right experiences to the right outcomes to either your product or your or for your business. >> So, let's let's talk about that a little bit more because I think if we look at this, you talked about future, you talked about the next generation, you talked about the wave of things happening. Um, you know, agents are generating behavioral signals that really existing tools were never designed to capture uh these signals, right? So, I I kind of asked that question of in the in the lines of we know what the heritage tools can do. But that's that's a point in time, right? And that's not the right approach. These signals that are coming out, that Let's talk a little bit about that because what you are doing is interesting and and it adds a lot more value to uh to the delivery, right? It it adds value to the insights of what's happening. And you know, you shared with me this this five-pillar model. I think that is kind of an important thing to kind of go through. I let's let's look at that a little bit. >> Absolutely. Now, so you touched on a word that really really matters here, which is patterns. You know, we we are investing heavily in pattern creation, not funnel steps or journeys, right? Fundamentally, agents are not software. So, that's something people have to understand. Agents are not software. And the biggest distinction between an agent and a piece of software is software is very deterministic. Buttons are put in places that people do things, and when you click that button, something specific happens. And you build your software in very specific pieces, and everything is in front of you. So, there's nothing dynamic. There's nothing that, you know, isn't prescribed or pre-built. In the agentic world, all that goes away. It's now a conversation. So, everything becomes dynamic. You can no longer have a deterministic path. You can't have a deterministic, you know, event. You have to unpack a conversation and convert that to a pattern. I have to understand when when you're having a dialogue with an agent, what was your intent? What were you trying to get done? What were you asking? What was the topics that mattered to you? Those are all semantics layer that you have to unpack from the conversation, and then you have to create a pattern. What is the behavior pattern that I engaged an agent? And all of those things are completely unpredictable and very dynamic. So, that's the first thing is that as as we move into measuring agents or as if you're starting to really understand what you want the agent how you want the agent to work, the first thing is that you can no longer operate from the old tool sets. They're all done. They're all in the past. Those are all deterministic software steps that are built into old capabilities. Now, you have to move into a world of very flexible dynamic semantics unpacking and creating patterns so that I can understand what people were trying to do, not from whether they clicked a button or not, but because they used a specific word or they or they ask for a very specific question. Those are the things that now matter in a gentic and and and that's where we really invest in those things called patterns and that's the difference between how you should look at an agent versus looking at software. It's a completely new design, new architecture, and new fundamental technology requirement that old tool companies cannot do. And so we're excited about this ability to unpack these conversation patterns and really show you how the agent is performing. >> Yeah, and I think it's important to understand that across the you know, if you look across the full agent journey, we see that two-thirds of enterprise AI leaders are already implementing a multi-agent collaboration and live pilot workflows. So, this is not theoretical or hypothetical. This is a problem that we we are we need to have those signals, understand those signals, and then move. But like when we also talked through this, um it's it's about active and growing um now with these agents and and not waiting until, you know, you have a fully baked, fully deployed solution. Knowing what those signals are providing and then acting on that. Um I think I think that's more strategic and versus being tactical or, you know, not being reactive. Would you agree? >> I absolutely. I I that's what I said. I think really understanding first how agents work it's a new it's a new paradigm completely. And then understanding, okay, now what what do I need from my tool sets cuz now old tools don't work anymore. And understanding what's available and and asking yourself entirely new questions. That That again, if you looked at old tool sets and you tried to jam an agent into the old tool sets, first of all, it wouldn't work, but second, you would be missing out on so much intelligence. And that's where we really define it as agent experience because there are so many things, for example, one of the biggest issues hitting a gentic today in AI is token consumption. Everybody knows it's a problem and everybody's seen the value and the benefit of this rapid, you know, deployment of AI, the automation of processes, of workflows, and and reaping benefits, by the way. So, even here at Conviva, we've committed to being an AI-first company. We have massive benefits in our productivity and acceleration of product development. But, then all of a sudden I got the bill. And you realize, "Oh my gosh, in 1 year I became a million-dollar customer with Anthropic, a multi-million-dollar customer with Anthropic, using Claude." So, now everybody's deploying agents, but the problem with that is there's tokens on the back end. And so, now everyone's like, "Wow, I launched this thing and it's and I think it's working and maybe it's not uh you know, optimized." But, the one thing that's definitely not optimized is your token consumption. So, if your agent is wasting time, if it's asking questions multiple times, or it's frustrating consumers, not only are you losing a user or a consumer, you're spending money on tokens. So, experience is actually a way cuz again, QoS means it just happened. It it worked. So, zero token consumption versus token consumption. That's QoS. See, experience is where you can start optimizing consumption. If I can make the experience more efficient, and I can cut down the time it takes an agent to understand the intent, for example. However many back and forth it took, that you can cut that down, not only are you improving the experience, which is by the way an experience measurement, right? Time it took to acquire intent is an experience, not a QoS thing, you're you're creating a better experience, while at the same time reducing your token consumption. So, we're you know, we we we fundamentally believe that once you get past the agent work, which is where the market is today in a very big way, we can show you that, the immediate thing you have to understand is I can start seeing things like how long it took the agent to do something specific. And can I optimize that? And can I make it better? Again, back in the old days, there was no need for that cuz the product was predetermined, it was pre-built. So, there really is no need to say, "Well, can I make it shorter?" Like, it's just can I get consumers to click buttons faster? Now, if you can get your agent to respond to a consumer faster or user faster and start to move through the intent to, you know, outcome more efficiently, you're creating a great experience in which people are going to want to come back and they're going to use your agent, but you're also reducing token consumption at the same time. >> Yeah, but Keith, here's here's where I think it's, you know, we have to kind of double-click down on agent experience is about to become the primary competitive surface for consumer and enterprise brands, right? It's It's It really is about the focus, but yet most companies are shipping customer-facing agents with zero visibility into experience, and those agents are actually that they're actually creating. So, they don't know what's happening, right? Customer support agents, sales agents, booking agents, they're all live right now, but they don't have those those that customer uh or that agent experience that you're talking about. And our data shows that 51% of organizations rely on public AI tools such as ChatGPT and and and Copilot, while 20% report enterprise-wide AI deployments on, you know, they're built on governed frameworks. So, the agent experience, it's it's a competitive advantage if you know what you're looking for, right? >> Yes, and I and I this is where the parallels to the old the streaming market come into play. And you go back and you look at all of the successful streamers built TV quality experiences online. And the head of HBO when we started our partnership said, "If you're going to build a TV business online, you better build a TV quality experience first, because people expect TV quality experience." It's the same with Ajanity is it will be a competitive differentiator for sure. People ask me the other day, "What is Conviva's mission?" And I said, "Our mission is to fulfill the promise of Ajanity to make consumers' lives easier." I I I I would love a world in which I can just talk to my device and it would do things for me. I would walk in my home and and speak to my home and it would just do things for me. If I needed to get something done online, I just tell my computer and it goes and does it for me. That's an efficient world. But that means the experience is good. That means that I'm not clicking and and typing and trying to understand and asking three times and it's coming back with the wrong stuff and I have to re-prompt it. That That That's not fulfilling the promise of Ajanity to make consumers' lives easier. So, we We take it seriously that experience is the competitive differentiator for businesses to build lasting consumer or user relationships. And the fact that their agents so streamlined, it's so efficient, it's so good at getting me to the outcome I wanted, I'm going to come back all the time. Versus if you don't, you're going to fail. And that's where if you go look back at the streaming market, it was the same thing. People would have launching video products. And if you didn't look at experience, you failed because you didn't understand consumers didn't like the product you were pushing out. They may have clicked play, but they left. It's the same thing here. The successful streamers knew the Disney's, the NBC's, the Paramount's of the world knew experience HBO knew experience mattered and they won. It's the same in AgentSync. If you're not thinking experience, you will fail because you're going to lose to those that have a better consumer experience that consumers enjoy dealing with your agent versus frustrated. >> Yeah, Keith, it's incredibly important that you know, you were you're talking about here. Your your heritage and experience on the streaming market is just absolutely applying to the AgentSync AI world. This is this is super important. The audience needs to understand that it's more than just deploying and what you know, what setting up it's one and done. It's actually about that that experience. So So Keith, if people want to learn more, the listeners here want to learn more about what you do and what can be can be as doing, where should they go to learn more about what we're talking about because this is new to a lot of people. >> Yes, no, it's coming on fast and I would encourage everyone to lean in. Learn as much as you possibly can cuz that's why I call it a revolution, not an evolution. And and and those that will win in anything, whether it's your your job, whether it's a company, whether it's your business, those that commit to leveraging AI in AgentSync faster will be the ultimate winners. So it's a learning curve no matter when you start, so start now. You can come to our website canbiva.ai. We have a LinkedIn profile on a LinkedIn profile and then a lot of different social media, YouTube, we have a YouTube channel which we publish a a of videos. Our website we publish a lot of of blogs. But or or or get to me on LinkedIn. I'm happy to have any conversations around anyone who's considering this. And one of the things we've already seen as companies launch these agents into the wild. And what are the the gotchas and the things that you need to be aware of so that you have a successful AI or agentic launch if because not launching is is not an option. >> Well, Keith, thank you for your insights. And this has really been great having you on the podcast today. I think the audience will get a lot out of this discussion. You know, it's definitely something that's top of the ROI on their AI invest. Well, thank you for being on. >> Thanks for having me. This was a lot of fun. I appreciate it. >> Absolutely. And a big thank you to all of you who have tuned in. We do appreciate you being part of the Active Angle community. But for now, that wraps up this episode. And we'll be back next Wednesday with another conversation diving into the tools, trends, and talent shaping the future of application development. So whether you're deploying at the edge, building with AI, or modernizing your cloud stack, we've got you covered. Be sure to follow us on social if you have any thoughts, questions, or just want to connect. Until next time, stay curious and stay building.