Vinod Muthukrishnan, Cisco | The AI ROI in Contact Center Summit

Organizations are investing heavily in AI for customer experience. But the conversation is quickly shifting from experimentation to measurable business value. Enterprises want to know whether AI can resolve more customer issues, improve agent productivity, and deliver better experiences without creating new security, governance, or operational risks. Now, Cisco has been evolving WebEx contact center around this opportunity, introducing autonomous AI agents, real-time assistance for human agents, AI powered workforce engagement, and new capabilities for managing a blended workforce of people and AI. Welcome to the CX Summit. I'm Bob La Liberte, principal analyst, The Cube Research, and I'm joined by my good friend and co-host, Zeas Caravala, founder and principal analyst at ZK Research. Welcome Zeus. >> Hey Bob, good to be back. >> Absolutely. And joining us to discuss Cisco's direction and what it means for customers is Venode Muto Krishnan, VP and GM Webex customer experience at Cisco. Venode, welcome to the summit >> folks. Thanks for having me. >> Yeah, so we're going to have a lot of fun today in this session. Um, I want to hit you with the first question. Cisco has described this as the beginning of the agentic era of customer experience. So what does Agentic CX mean to Cisco and how is it changing the role WebEx contact center plays in the enterprise? >> So obviously that's a mey question. I'll try to be succinct here. Um I think one of the things we've seen with AI especially in CX is you know it's gone from being a toy to a tool to actually a digital coworker. And what it means uh in terms of implication is what it's been able to do 3 years ago is very radically different from what it did yesterday to what it'll do tomorrow. And and I honestly feel all of us have bandied the word agentic around a little too liberally. Uh an AI agent that can fulfill slot fill or fulfill a transaction is not agentic. Agentic is much more of a framework of tools and products and processes. So for us the big thing around agentic is going away from just voice agents that can answer a question, fulfill a transaction to orchestrating a journey to now evolving all the way to be able to orchestrate a relationship because when you move away from just what is the intent of the caller, what is the transaction to be fulfilled and you think about the fact that why is Z calling me uh and be cognizant of everything about Z's relationship with me that matters then I will treat that conversation dramatically differently from he wants a refund, can we give the refund, can we not give a refund? Um, so for us, agentic truly involves thinking through not just the immediate conversation, but putting the relationship lens on, bringing together tools, systems, and processes that allow us to govern and deliver the kind of relationship the customer wants and then follow some very very uh deterministic processes, but using AI to demonstrate the kind of autonomy that a human would. For example, like when you ask for a refund, it's fairly deterministic. Like I can't just say, "I like the your tone of voice, so I'm going to give you the refund." That's not how it works. How you process so many things are very deterministic because there's policies and procedures around how you treat your customers. Even discretion is codified, which is if someone's a platinum customer, they're super angry. What can I give them immediately as a customer service agent? Even that discretion is codified in some way. The true arrival of agentic is when you're able to take all of these and give a very human-like experience both on an AI interface and a human interface wherein the agentic system is powering this what I call um humanlike customer experience no matter what the interface is. So that's what we mean uh by agentic. Again, not an agent that fulfills a transaction, but an agent that can orchestrate the entire relationship itself. And obviously, as we go along the conversation, we'll talk a lot about what are all the pieces that help make that possible. >> Yeah. Now, Venote, you know, the the vision you articulated there isn't dissimilar to what we're going to hear from every other company. I know though more and more Cisco has been going to market with their one Cisco vision right which incorporates not only you know your stack but also networking security. So how does that um you know vision create something that's unique to Cisco in this space? >> Fabulous. Um we've said this in many places and thank you for for uh highlighting that that point. It's never in our lifetimes been easier to build an AI agent. I think all of us can build an agent through this call if we wanted. It's never been harder to make it enterprise grade. And the reason I say that is again if you go back to the analogy of being a toy and a tool and a digital worker or handling one conversation to a journey to a relationship when it was one conversation, the threat surface was limited. You're asking for, hey, refund my my ticket. I validate who you are. I can give you the ticket. or not. So there's only that much that can happen in the conversation unless it's a very poorly guardrailed um agent. Now when you're orchestrating a journey, I'm checking your entitlement. I'm doing this. Suddenly there's more surface to the conversation. Once it comes to the relationship, you realize that there are so many places where this agent could go horribly wrong. And remember these are openloop transactions which is these are transactions that you know you're accessing company information customer information you're acting autonomously and you're talking to end customers right so the threat surface is potentially infinite there and so as you go from building really nice cute sounding AI agents to actual systems of AI the role of observability and security and governance is critical now from an observability perspective we need to observe at machine scale which is You need to look at every touch point internal, external, where you're either accessing information uh uh inferring it, reasoning with it or doing something with it. Being able to respond at machine speed, which is how fast can you move when a million agents are having simultaneous conversations uh between first and third parties. Then obviously securing it wherein you have to think through security not as an add-on but essentially if you have complex multi-turn conversations, how do you secure them and ultimately have a governance framework which is ethical. That's what we mean by responsible AI. >> All of these are not separate things. So you can't say hang on, I've got this now I'm going to do some work on how how do I treat your data or security there are six third party products. I'm just going to plug them in. Observability is something I do in terms of I'll tell you the five metrics that I can track. That's not observability. So for us the big thing and just to to tee up the answer is how do we bring the fact that we have a multi-billion dollar observability business and a security business and vertically integrate that in what we call the AI agent 360 itself. So you build AI agents as easily as you're supposed to be able to do it. But you're also able to embed all of these natively into the platform. So security, observability and building this AI agent are not conversations you have in different rooms with different vendors. You have it in one single place in one design environment with one vendor. >> All right. Now um so you know I get that with the with your integrated uh platform there but on the topic of the conversation itself you've in recently introduced the concept of an AI concierge and that's designed to maintain that context that you're describing across channels agents business system and human human handoffs and so um from a a practicality standpoint how do you does Cisco deliver that as one continuous conversation. >> Okay, thank you. This is literally the holy grail. Um, so you know >> indeed. >> So this um so we always call it the intelligent front door but this morning we workshop with a customer I shall not name them and they presented their vision uh to us and of course we've been working with them on it and they said something very interesting. They said there is no wrong door which is no matter which touch point, what channel, where in the enterprise you knock, the door opens and miraculously it's almost it's the same person on the other side. It's the context is continuous and you're picking up the conversation from where it left out. Which means in every touch point across all customers, you almost have like a personal relationship manager who just knows everything about you. So the big evolution that's happening I feel is twofold. One is contact centers are evolving into what we call context centers because this is where most of the conversation gets funneled. The next is the absolute breaking down of barriers between contact center. Let's let's say you're a retailer, right? The storefront, your mid or back office and your call center. They're all touch points. If you find a phone number in page 17 of your website and call, it's still your experience with the brand. You call the store, it's the same experience. And if you call the contact center, it's the same. The calls may also get bounced around from one to the other. Now for you it might be hey you're the retail team Zeus you're the finance department or inventory team I'm the contact center that's our chart for the customer they're all the brand so no matter which touch point I call I call IM message or you call me or you message me for me it's part of my continuous conversation with the brand the reason we've not been able to stitch it together one is infrastructural you're using some telephone system you're using some phone endpoints you've got a different contact center but we we have UKcast and CPASS together we are able to bring these together then the lack of AI to understand the context of the conversation we're now able to record or analyze these calls even without recording them transcribe them topic analyze them and understand what's happening in millions of these conversations on real-time basis and then our biggest area of investment obviously has been the agentic context engine where we are now building a real time knowledge graph of each customer's taste choice preferences so it's not that you got the same person, no matter who picks up the phone or opens the storefront, human or non-human, starts off the conversation with an absolute understanding of the context with which you're calling. And for you, which means no matter which touch point I interface with, it feels like part of one continuous conversation. So for us, that's why it's a holy grail. It cannot be solved by just a context engine. You first need the infrastructure and plumbing which we have with UKcast, CCAS, and CPASS. You then need to have a common record, common store. You need to have a context engine that is contextualizing all these engagements or different touch points. And then your application layer should be able to deliver this insight to the human or to the AI agent. So when the customer calls, it's intuitive, it's instinctive, it's literally like having the same conversation with the same person. So I know it's a lot, but if you do all of these things, you'll be able to offer this experience. If you do just one layer of this, it'll be very hard to deliver upon this promise. >> Yeah. And so Venode, you're talking about when you're talking about this context in bringing in the other systems into play, I think it's in in your terms, you go from the intelligent front door to the intelligent engagement with with the client. I'm just curious how much work, how much customization is required to be able to get to that level, right? If you're trying to get into Salesforce and Service Now, others and things like that before an enterprise can put AI concierge into production. >> That's a great point. So I think the incredible part here is you don't need to boil the ocean all at once. >> Yeah. >> The first thing people are there are already huge problem statements that don't even need a concier agent. For example, are you available 24/7? Um when I call a store, one of two things happen, right? You know, there's no winning. If if if I'm if I'm one of the two folks in a store and I pick up your phone, I'm probably not serving a customer who should be served, right? If I don't pick up the phone, I probably lose a chance to sell you something. So whether you pick up the phone or not, the org is going through some sort of a problem. So being available 24/7, dduping calls in queue, calls going answered or unanswered in the storefront or where have you, these are all like absolute uh tier zero level of problems that need to be solved. Now the reason I say that is we don't need to start with a concept of a concier that answers everything. That's all knowing. You start by solving the problems that need to be done knowing that the infrastructure and plumbing, the knowledge layer, the application layer and the AI I mean AI is the application layer are all on one platform which means you can build towards your northstar goal right so as more touch points start solving point problems that matter know that you're not solving them using point solution you're solving them using a platform and as you get a certain degree of sophistication and a volume of conversation in you can then start threading them in saying hang on I can now do these 10 or 12 types of conversations with you. The concage knows it all. It doesn't matter whether the phone rings on Bob's desk or we know in the contact center or Z is in the procurement team in the back office. When the call rings the context is preserved and I know what I know about you. So the way we look at the concage is you can start with simple conversation you can start to orchestrate some sorts of journeys. The fact that you have one platform underneath means that even though you're offering a point uh sort of solution, it's not on a point platform. You're actually doing it in one common platform and it's your ability to build towards the concierge agent becomes that much better. Also remember the ability to access backend information, orchestrate workflows with backend systems, all that needs to be in place for some of these high order values to get done. But you don't have to wait for everything to be ready to get started. There are hundreds of problems costing you millions of dollars that need to be solved today. So long as your platform underneath is the same, it's okay to go point by point till you can stitch these together. >> Absolutely. That that makes a lot of sense and we hear that a lot. Just get started, right? Don't wait for everything. It doesn't have to be perfect. Get started and you can evolve from there. And clearly, as you mentioned, AI agents are getting a lot of attention right now. Everyone's interested in them. But you guys are also doing a lot of other work as well to help the human agents things like real-time assist right some other capabilities. How's AI changing that agent experience and where are customers seeing the greatest productivity gains and returns with that technology? >> So no thank you for asking that because we feel this whole absolute fixation with AI replacing humans. Maybe it happens, maybe it doesn't. I I wish I had a crystal ball. I don't. >> Yeah. What I do know is something very interesting. Uh we did a survey of thousands of customers in a in a healthcare segment. And the funny part was this. Of course, you know, some organizations just some areas just lend themselves to poor customer experiences. Almost all the great experiences they listed had one thing inconsistent, a name. They mentioned a human agent by name. So, so, so I don't know about bad experiences, but great superlative experiences are delivered often by humans. So the way we look at it is whether you have 100% AI 100% humans or you increase the number of humans or have it we don't know our belief is that we need to think about the end customer which is how do you make the end customer experience better which means the underlying system of AI is the same whether you use text to speech to power an AI agent or you use real-time assist to help a human agent the common the context layer the learning the memory should be the same. So know that the engine that helps the human agent get better is the same one that helps the AI agent also get better. It learns from both touch points and it powers both touch points. From a human agent perspective, the things that come in the way and it's I'm hardly the first one to be breaking news on this front is a simple fact that they have a lot to do. They have calls in queue. They need to wrap up work. They need to do this. They need to click on systems. So if you can transcribe, summarize, automate follow on actions, load the right pills in front so they're able to go take someone through a conversation and allow humans to do what they do best, which is emote, empathize, connect, respond, you will have lower agent attrition, higher agent productivity. So for us really the entire AI um assistant uh product has all of these pieces. It has auto cesat. It has everything for the supervisor like topic analytics. And what it's done essentially is twofold. I've always felt that a single metric-driven CX action is always bound to fail. For example, if I tell you lower average handling time, we've all been around contact as a long you tell someone reduce your average handling time, they will reduce the average handling time. Okay? They may not increase the seat. >> [laughter] >> So, so you got to look at AHT with FCR with seesat which is our customers getting what they want right are they getting it in with less customer effort than the past and are they net happy with the experience that is FCR AHT and seesat or what have you working together and our aim with our customers have been to help agents deliver exactly that which is there is a average handling time goal that we have which we want to reduce obviously sleep. We want to make sure the FCR percentage goes up and the auto sees or manually gained seesat is at worst neutral at best better because customers just got what they wanted in less amount of time. So that's where we've seen the maximum amount of uh uh benefits from this. But the other part of this also has been topic analytics which is what conversation should I automate and not. As you know we are huge um uh disbelievers in terms like containment and deflection which I think no customer wants to be either contained or deflected but preemptive automation productive automation that is great and so topic analytics tells us all the call types that can be automated. Then you put a overlay on for moral, ethical, brand or other reasons what call types we don't want to automate and it allows uh the human agents to be used for the kind of conversations they add most value in. And that's how we're seeing all our customers go about it. >> Yeah. But no, you went through quite a bit there and um Bob and I actually talked about this when we we opened the the session up and historically the contact center industry's lived on some of the holy grails, certain metrics, handle time, first call resolution, things like that. You threw out a bunch of alternative um things you could measure, seesat, you know, things like that and uh even on boarding time. Um [clears throat] what have you found that customers have coalesed around uh a set of new metrics to measure or is this still work to be done here? >> I think it's it's evolving. >> Uh the simple stuff, right? For example, if you your call center was shut on Saturdays and Sundays and 10:00 p.m. to 6:00 a.m. and now you offer an agent that answers some questions at 2:00 a.m. It's its conversational skill doesn't need to be better than the human, right? Cuz you're offering something earlier that the customers never had the ability to to reach out at 2 a.m. because they are flying through some countries or whatever, right? So I think the reason I say it's evolving is customer expectation I believe is growing at a dramatic pace. Any treatment on FCR or uh AHT without linking it to seesat I think is bound to fail. And the great part with the analytic suites available today is you can actually see them together. So you can't just you don't just have to say something like what's my call containment rate? Like again I'm fan of that at all. So it's like okay if I got 10% calls contained if I may use the word what happened in those calls what is the seesat on those calls how many of them got contained because someone actually the last words were never calling you again like containment could well be me telling you I'm never calling you again that is not good containment so the big thing I'm saying is people want to understand what happened inside this don't just tell me the metric tell me what happened inside that metric what is the quality of the conversation what words were spoken, what was the sentiment of the customer. So a much greater um investigation if you may into what's the story behind the qualitative story behind the metric, right? As opposed to just the metric because first wave just was 10% calls need to be contained so I can save x amount of money. I think customers have moved very very very fast through that. Uh so it's not net new metrics but the willingness and the the ability to see these metrics in tandem which ultimately tells you when Z is called did he have a good experience or not and that's the ultimate metric. Yeah, I want to shift gears a little bit here to a topic that um few years ago nobody really talked about. Now everyone's talking about that's workforce management, right? Work workforce engagement managements, WFM, WM, that whole suite. Uh you recently announced your own AI workforce management and the the obvious use case is I got to manage people. I got to manage AI manage AI agents. if I do them separately that uh problems are going to happen. And so talk about the new operational challenges that uh CX organizations might face with this blended workforce and then how your integrated workforce management actually addresses that. >> Fabulous. So I think the one great part is we put the customer hat on which is when a customer calls whether they speak to AI or to a human what is the kind of experience you want to give right what kind of handle time what kind of seesat what's the customer sentiment we desire to deliver now you deliver that using a workforce that is human and AI as we said very um openly the few the the AI has now become a digital coworker it may be in the early stages of being deployed such but when you architect for the next decade you architect this way. So the end outcomes will be the same for the customer irrespective of AI or human. But how you train, orient, schedule this worker is going to change depending on whether it's AI or human. So you may do the same QAQM on both because quality is about what the customer's experience is or what experience you delivered. But how you then apply the learning of that coaching plan will change. For a human agent, you apply a coaching plan differently. For an AI agent, you'll go to the supervisor who may ch make changes to the prompting or what have you as a as a human safeguard on a recursive learning loop. You could theoretically put a recursive learning loop and the AIQM agent could continue to optimize the uh the AI agent. But today we find most of our customers want one human checkpoint before the in that recursive learning loop itself. So you will see self-arning agents happen. So obviously that's different from how you typically coach and train a human agent. Same with workforce the inday management um the scheduling itself there are certain constraints of capacity and capability that apply to human agents that also apply to AI in in the case of capacity it is is it budget which is what what how many dollars have you allocated for AI is it if it's open-ended no problem if it's $10,000 a day so be the number so the the um inday management needs to account for what call types by policy and by training I have agents to answer which ones I don't and for certain call types how many how many tokens can I burn on automation because for example if you had a thousand agents and you just infinitely left the tap open for for AI agent which is fantastic but if you have 200 agents sitting without calls now you're wondering what you're doing and then if I were to further compound that the uh average seesat on the AI agent call was worse than the human now you took an expensive resource kept them idle and then you you you you directed calls elsewhere. So I think what we're seeing is a lot of experimentation going around what can and should happen. Uh you can theoretically quote unquote automate all the calls but then there's a seat cost to pay. >> Yeah. >> So the workforce tool is supposed to allow you to dynamically run these tests and really see what is the best customer journey with the best outcome, FCR, AHD at the best cost basis. So if you have a ven of these three, you will see businesses on a real-time basis tweak this. That's why you can't have an AI agent management platform and a human agent management platform and then one third exotic sort of dashboard which brings these together. You need to be able to look at the forecast, manage the flow, run your schedules and do the intraday management in one place and use the quality management loop to optimize these workers, human and digital off of one place. And that's why we announced the AI native workforce platform. >> Yeah. So I I find this fascinating and it was interesting to hear the the conversation between the different AI agents and human agents as you're looking at doing quality management and so forth. Should the AI agents and human agents be evaluated against the same quality and customer outcome standards especially when you know humans are going to be handling the more complex issues for now anyways, right? Versus the other. How do how do you look at that and what do you recommend that for for customers when they're using this? I wish I had all the answers, but I'll I'll take a first stab at it. Um, again, as I said, we need to put the customer hat on. I I can't tell you, hey, you're going to you're going to have a terrible experience because this is not human, right? Or you so so the customer is asking for a certain number of things. They're asking for you to do something. So the some metrics like FCR, AHT, and seesat are customer metrics, right? How much time did I and customer effort score, right? How much time did I spend? Did I get what I wanted? How did I feel through the conversation? That's been consistent today 10 years ago, 20 years ago, 30 years ago. Right? So, we got to judge human and AI interfaces with the same lens. There's a an internal mistake which is cost to serve. If I can with minimal degradation in service offer you the same experience at 1/4 the cost, why would I not do it? >> Yeah. >> Right. Ultimately, you have to optimize the operations. So, that is important. Now the point is if I do it at like 50% of the seesat now we got a problem. So, so this is what we essentially seeing brands do which is I think where all of us are in certain degree of experimentation on this front and the the holy grail essentially is how do we take internal metrics and optimize those in a way that the external metric that the customer cares about it is at worst neutral >> and if you could find the holy grail of making it better again as I said if I could not call you at 2 a.m. and I can call you at 2 a.m. I'm okay with a metallic voice telling me, "Okay, here's their account balance." I'm just making the issue up, right? >> Yeah. >> If I'm now competing for airspace wherein I tried to speak to ZK and he's not available and I'm getting his agent and that agent sounds terrible, that's a problem cuz you're trying to tell me that instead of speaking to me, please speak to my agent, right? And that seesat has to be at the very least as good as my interface with him. So that's what we are seeing happen. And ultimately as you spoke about the difference between the two, I think you will end of the day the customer recognizes that they're asking for something non-standard and it's a complex one. So it's okay if the human agent is doing the real complex ones. It takes 5 minutes and 10 minutes and lots of pings and pongs whereas the agent is doing the easy ones. So I don't think it'll be apples to apples but so long as we are doing it by customer uh issue type and what resolution we're offering and we're measuring the right thing for the right customer journey we should be fine. That's my guess. >> Got it. Yeah. No, I think I think that makes um that makes a lot of sense. I think um I think one of the the other questions I wanted to ask you, you know, we've seen a lot in the news about AI going rogue, right? people concerned about what's happening as AI agents gain access to customer data right they start completing transactions and so forth you can look at that autonomy is also increasing business risk so how's Cisco helping organizations govern secure and observe those AI agents before and after they enter production >> no again as I said this is the question of our times never been easier to build an AI agent's never been harder to make it enterprisegrade And there's a line we often use. We say um uh autonomy without accountability is a liability. >> So if you're having a a build my AI agent conversation in one room and how to secure, manage, observe your AI agents in another room in 2026, you're either in the wrong year or you're in the wrong room. And so for us, I looked at what we announced at RSA, which is not the contact center of business, which was the rest of Cisco saying protect your agents from the world. protect world from agents and observe at machine scale and respond at machine speed. This is what we've embedded into our AI agent builder itself. Because when you think of AI agents, we need to start with how are we building? You're connecting with an MCP server. What's on the other side of the MCP server? You're connecting with the knowledge base. What's inside the knowledge base? Is there a spirious link there? Then you're building. So this is how you're building. Then you're trying to build and deploy. you need to actively red team to see how will this agent theoretically uh respond uh in runtime. Then you go to runtime itself. In runtime you've got prompt injection. You got 100 things that are happening. How will you observe and monitor this in runtime? Sometimes it may well meaningly hallucinate. And so what semantic chunk do you go down to? How did it pick $20 instead of 30? Oh, now I understand what the reasoning mistake was. let me apply a patch in a recursive way to make sure it doesn't make the same mistake again across the thousands of other conversations going on and then ultimately how do you calibrate this agent which is all the telemetry we said around how the performance is what metrics it's delivering the ability to do all of these together in one place is what I think is the clincher otherwise you'll always be playing whack mole with what my issues are so bringing this entire observability suite bringing our AI agent observability platform and our AI a defense platform vertically integrated into our AI agent builder which is AI agent 360 to us is the holy grail. You should not have to think of these as separate products. You should not have to think about these as separate vendors. You should not have to stitch together a bunch of these solutions. And that's why the fact that we're a full-time observability company and a full-time security company for me is incredible because I get to stand on the shoulders of giants and build the best AI agent experience knowing that there are thousands of other engineers working on the security and the observability that I just get to plug in. >> Uh and I consider that my greatest privilege as I run the CX business at Cisco. >> Yeah. So Venon on that topic you bring up an interesting point because so that makes it easy to integrate those agents across the Cisco stack. Of course, no vendor is going to deliver and CX on its own, right? And one of the struggles I suppose for this industry historically has always been walled gardens, right? We we've got our big application vendors, the great throne walled gardens. I know Cisco though has been very aggressive and vocal about its support for standards such as ADA and model context protocol uh which allows WebEx agents to not only work within the Cisco [clears throat] stack but also with third party agents and other enterprise system and so just talk about how why openness is so important for Cisco strategy how customers should be thinking about that as part of their evaluation platform and just with your industry hat on are you getting the same kind of reception from your peers you know and other vendors that indeed this we may see a new era here uh of openness in this industry >> no thank you for saying that look uh you know Ju Patel has been fairly vocal about this and I I'll I'll add my own I think [laughter] I think openness for us at Cisco is both a cultural and a technological uh trait for us so what I mean by that is are you willing to even be open, interoperate and integrate with systems that theoretically conflict with you. Short answer for us is yes. Um and there are two three reasons for that. See our aim is to be the one platform you innovate on, not the one product you innovate on. Your in CX for example, what is a CX plane? Earlier it was contact center and CRM. That's what someone called one box. What is the box now? The box is open wide up. um it includes CRM, it includes your inventory management system, it includes some other system. So every system that has that potentially interfaces with customers andor has context that is relevant to the customer interaction is part of the CX plane. So obviously you don't own all of them. So our so the the line we had was you know we don't believe in integrations. We now believe in interoperability. The systems need to be able to talk to each other, communicate with each other with no fear around, oh my god, I also have a comparable module. Should I integrate? Short answer is you integrate because you want to offer the platform on which customers can build their innovation on. We even have some things like bring your own AI. Use first party AI or third party AI. Whatever does the best job for you, you should be able to load it on top of a Cisco platform and use it in a way that your business goals are fulfilled. So again as I said this openness and interoperability not just integratability for us is technical and cultural and I mean you said it you kind of also gave the answer there. I don't I think the days of wall gardens are long over um you have to like if you're not embraced these standards you're probably already too late and in our case you'll see that wherever we have first party products of choice we openly um have on our catalog third party solutions. So when you choose Cisco, you're not essentially signing away the next 10 years of your life saying I can't use any other product and service. On this platform, you can load anything you want that helps you achieve your business goal without being locked into, as you said, the wall garden. So that's really where we stand on this front. >> I'm not convinced the air the wall gardens over. I hope you're right, though, because it make things life a lot easier for the customers. >> The walls are getting shorter. [laughter] >> They're coming down slowly. >> Yeah. Uh, hey Veno, this has been awesome. And as as we wrap up, I wanted to ask you one last question. For the CX leaders that are watching this discussion, they're thinking about getting out of their AI pilot into production. What are the two or three concrete actions they should take over the next, you know, four to six months? >> I said this in a, this isly random thoughts. I hope I structure them in the right way. But I've always said like, >> you know, savings or what have you is the entry point, not the exit point. So it's very important. One of the things we've worked with all of our customers and partners is let's find the empirical formula that we can scale and the empirical formula is take a use case take a journey take a knowledge source think through the security observability and learn something at an empirical level that you can dramatically scale afterwards. So we always we may show the northstar which is blue sky but never start with deploying blue sky step one being able to put points on we literally have a document around AI readiness assessment which says put some points on the board and so our first job is to help people put points on the board. When you put points on the board how does your organization respond to it? Simple stuff. If you put AI assistance for human agents are they receptive to it? How do they treat it? So unless you learn it, if you have 10,000 agents, you want to start with 100 and see how they respond and then scale it to 10,000. So put some points on the board. Definitely have a very very ambitious notar because I think the blue sky is still not ambitious enough. That's how much this can do. But then next build a framework wherein you've understood the risk, the mitigation of the risk, the governance of this platform, the human and AI interface. You've and cost cost benefit. You've kind of done this at an empirical level. Once you build that empirical block, scaling becomes a whole lot easier. That's the second one. Third is no matter who your vendor is, if they're a vendor, you already have a problem. They need to be a partner. You must insist on openness and interoperability. And I put my hand up and and I commit to that. We have bring your own AI. All our AI assets are there. You can use first or third party. We have things like campaign management. You can use third party. Workforce, you can use third party. Cuz we feel all of these companies exist for a reason. Our aim is not to exclude anybody. So if openness is not um a foundational cultural and technical attribute, you're betting on the wrong partner because you're assuming for the next decade, this one company is going to own the innovation road map for the universe, be on the vanguard of AI, software, infrastructure, and if there is one such company, I'd be very happy to hear about it. But I doubt if that's going to be the case. So the most open platforms will help you scale because what we know now may not be true 6 months from now. But if you're open, you are able to leverage the opportunity of where the world looks like in 6 months. So those are my sort of very high level thoughts. Have a very aggressive northstar. have a very measurable empirical formula that is replicable and scalable and always invest in openness because when you are on open platforms you can evolve as the times go as opposed to looking at a contract that has 10 more years on it that you can't negotiate your way out of. Yeah, absolutely. I think I think that makes a lot of sense and I think that's great advice. Venode, thank you very much for joining us. >> Thanks so much for the time folks. This was a fun conversation. >> Yeah, thanks. >> Absolutely. So, the key takeaway is the next phase of contact center transformation uh will not be defined solely by how many interactions AI can automate. It'll be determined by how effectively organizations can combine AI and human expertise to improve resolution, consistency, customer outcomes. Right? A lot of those things we talked about the new metrics that you need to look at. Cisco's direction reflects that broader transition. Autonomous AI agents, real-time assistance for employees, unified management of a blended workforce, and the security and governance required to move from experimentation into production while leveraging that open platform. So, Zeus, thank you for co-hosting and thanks to everyone for joining us for this segment of the CX Summit.