Zendesk CEO: AI Customer Service Agents Are Ready For Primetime
Channel: Alex Kantrowitz
Published at: 2025-03-26
YouTube video id: -araPrJQodA
Source: https://www.youtube.com/watch?v=-araPrJQodA
If you're like me, you've been wondering how much of the conversation around AI agents is hype versus actual reality. Today, we cut through the noise with Zenesk CEO Tom Egimire as the company launches its AI powered Zenesk resolution platform in a conversation today brought to you by Zenesk. Tom, great to see you. Welcome to the show. Great to be seeing you as well, Alex. All right, so you are working on AI agents. Uh it's a term that we've been hearing all over the place from almost every AI company and every nonAI company. Uh everybody's been talking about agents. Um the one thing we're trying to figure out on this channel is what's real and what's not and it seems like there's a lot that's not. Um but rarely do we get to speak with someone who's actually developing the technology ground up and that's you. So just give us a sense as to what the lay of the land is right now. um where we stand with AI agents, what's real, and what is hype. So, I think AI, you know, I'm old enough to remember when 20 years ago there's just going to be this AI revolution with machine learning and predictive analytics. I hate to say it, but old school AI. And I think there was a lot of hype there, but there wasn't much delivery. I will tell you overall, I think large language models are actually changing the game. You know I I believe in the old adage uh that you know technology uh impacts in the short term we overestimate in the long term we underestimate. I think that's what's going to happen with large language models for customer service in particular. U we are seeing AI agents really transform our particularly our business to consumer customers. We're seeing some people get 60 70 80% what we call automated resolution. So a consumer comes into their business, they can actually solve the problem with an AI agent 60 70 80% of the time. We see B2B businesses lower like 20 30 40%. Um what the hype is probably is I see a lot of people go out to companies and say we can solve 80 90% of your interactions with your customers. Not solve them but I mean solve your customers problems in a matter of we'll implement in a day. we'll have we'll be have you up to 80% in a matter of days. It's all going to be good and you're going to have no mistakes. And so I think that is the hype. I think there's still some hard work behind uh getting people to 60 70 80% even if they're a business to consumer uh company and 20 30 40% if they're a businessto business. But um I think um the hype is real overall. AI agents can solve consumers problems. I think we're going to get into there's a little bit of a hype right now about you having your own personal bot uh contacting a company. I don't think that's working really really well yet, but it's going to come. And so I think the hype is actually justified and we're seeing uh the impacts with our customers right now. Okay. So let me talk to you about this for a minute because we've been hearing about automation. There's been robotic process automation before there was AI agents. Um there have been automated customer service chat bots for a long time. So I I'd love to just pause here for a minute to for you to talk about what the difference is between let's say 3 or four years ago and now and then even more than that what are the type of ramping up uh capabilities you've seen from AI generative AI companies recently that's allowed you to be able to do more than you have been able to do in the past. Yeah. So I think I think a couple things are different first um agentic AI buzzword. Okay, I recognize that. But I think bots are actually able to reason now where in the past 3 or four years ago, you would have an AI bot, but behind the covers it was like 95% rules-based. If X do Y, decision trees, things like that. And so people would, you know, triumph it. Hey, we got a really great AI agent. It really or AI bot. It really was rules-based you know was 90 95% of the underlying technology with Agentic AI and with the latest large language models you are really able to have take a input into information uh take the large language model do some post training and it can get the right answer just an incredible amount of the time and it can reason and so I think that's one of the differences. The second difference is you're able to personalize them so easily right now. Um I can't use the uh pop star but think of a major pop star uh and she was uh uh she was releasing an album uh and um she wanted to have an experience that um the u the uh bot uh the AI agent was going to be a little more in her tone uh when she was interacting with her fans and we were able to personalize that you know in a way that the fans had just a truly fantastic experience like that they were talking to the artist and I don't think You could have done that personalization really easily three or four years ago and I don't think you could get to the personalization of a company that they wanted to have really a tone with the brand. I know we have an options three like default options. Do you want a professional? Do you want casual? Or do you want some kind of hybrid and and and do you want to go in different languages with the customer? There's just a lot of personalization that's easy to turn on. Finally, Alex, I'd say it's easy to get up to speed. In the past, you would do a lot of customization and there are six, nine, 12month pro projects because you'd have to take all this data in. You'd have to go put all these decision trees, these rules, you'd have the bot on top of this. You know, we can get a bot or an AI agent up and um running in a matter of days um with a lot of personalization. So, I think it's a lot easier right now to go implement um and to get a high resolution rate. So, for all those reasons, there is some hype. Don't don't don't don't don't uh don't misunderstand me and I don't want to be too uh on top of the hype cycle, but we're just seeing like really really strong um uh automation rates and strong impacts on our customers. And we are going from I think that you mentioned we're going from uh if then statements basically to a broader range of possibilities and you're going to see more powerful stuff come out because of that. And we just had a conversation with the people that built the new Alexa Plus here on the channel and they are also saying yeah we we basically I mean my reporting said it was a a you know if entry they want to talk about the new technology they're now moving to something that leaves uh a much more broad space of possibility open to their new assistant. Now there's an interesting component to that which is you go from uh deterministic technology where it says you know if this happens then this should happen. If I say, you know, uh, a word, and I won't say it because it will, that, not that a word, the a word that summons this assistant, you know, turn on the, um, the lights, it knows, okay, I'll turn on the lights and the lights go on. But when you're living in the world of large language models, it's there's so much more that's left open to chance or probability. Uh, you're in more, as they call it, probabilistic space. So, I don't know. We we still don't we still haven't seen Alexa Plus. Maybe it'll be come out by the time this this interview hits. Uh but I don't know how you do that in business. Uh because you really in a business uh you're kind of risking a lot if you start to leave that open. So how does that work? Yeah. So we we take um you know some off-the-shelf large language models and we actually have um 18 billion interactions that are um that are anonymized uh in a database with people rating customer uh interactions a positive or negative. Real simple thumbs up, thumbs down. And we do a lot of post-training on our models, the basic models after the fact to try to get their probability to your point, win rate, their resolution rate as high as possible. And what's interesting um you know when a when a bot makes a pro an error they do make errors it's a hallucination when a human makes an error you know it's a mistake and what we find right now is on our next generation agentic AI bot um we have a lower error rate than a human being uh from a you know contact center support center answering like for like inquiries and so when we talk to our customers we say look there are going to be some errors 100% it is going to happen. Um, just like a human being, no matter how much you train them, how much you work with them are going to have some errors. But what we're finding generally is you can serve your customers at a a much lower cost. you can get their your customer satisfaction up because if you have a personalized bot that can instantaneously respond versus waiting online for 5 minutes or 10 minutes for someone to respond to your email, your message, your chat, your web form, your phone call, you're generally more satisfied and um the error rate is actually lower. Okay? And so there's some things that you can uh protect yourself that you could hardcode in still like hey I don't want to share any personally or take in any personally identifiable information to the bot. You could have a like a hardcode rule to protect some things particularly for regulated industries. But overall bots are quicker, they cost less uh and they have better customer satisfaction. So you know I think it's a win-winwin if you implement them correctly. Yeah. Yeah. And there were some early examples that were floating around the internet of I don't I don't think it was Zenesk, but like companies setting up C bot customer service and them like giving them 50% discounts on a Mercedes. I'm making this up, but it's directionally along the lines of what it was. But what you're saying is now you can actually set some rules in to make sure that stuff doesn't happen. Yeah, I think you can you can lower the error rate. I just just like if you train your human agents really really well, you could still have errors. And it's kind of funny some of the things uh it's a competitor. So, I hate to give them a pass, but I will. There was a, you know, airline company that had an issue with a bot to one of our competitors. When they went into the details and figured it out, it sounds like what it was actually was a documentation error on the company that the bot took information in. And so, again, this is why documentation, knowledge, we call something a knowledge graph that we have, you know, inputting information. It's really really important to have accurate information and to scrub that information because um you know the bot does reason the or the AI agent reasons the AI agent gets personalized the AI agent uh takes in information but the the bot is only as good as the information it's taken in. Okay. Well, justice for the bot. This was clearly a human error where they uploaded the wrong stuff. And I want to publicly apologize for defaming that bot and hopefully uh it doesn't take legal action against me. So I I hope so too, Alex. And and again, I think it's gonna be cool with stuff like Alexa, and I'm worried I have Alexa right next to me that it's going to turn on here, but Alexa talking to uh an AI agent that's, you know, powered by Zenesk. I really think that's going to come over the next six or 12 months, and we're going to have some really, really cool use cases where the consumer is there using their own bot to interact with a company's AI agent and it's just going to really uh take a lot of friction out of the system. And so, I'm kind of excited when we see that. We've got, you know, we've got examples of that, but on scale, I think it's coming with, you know, Amazon and Apple and Google and everyone else coming out with these personal assistants. I'm definitely looking forward to bottobot communication. And I do want them to develop their own language just to be be booping the information back and forth and ditch natural language. Uh, maybe we Whenever I hear that, I think of Star Wars, by the way. I think of Star Wars, the same kind of thing. C3PO was on to something, right? Everyone made fun of him for beeping, but actually turns out we're the inefficient ones. Exactly. Was it C or R2-D2? Anyway, my knowledge is getting rusty. Um, so but but this is also coming at a at a moment where you have big news around this topic, which is that you're launching the Zenesk resolution platform. This is brand new and there's lots of news in there uh in terms of how you're going to uh enable people to build agents and serve your customers with agents. And I think um like again as we started out with the beginning there's a lot of people philosophizing about Agentic AI buzzword it might be but again as you said you're doing it so tell us a little bit and it's great to have you here as you announce it. Tell us what is coming today. Sure. So we're uh announcing um at our customer event uh March 25th to 27th uh in Las Vegas our global relate event that we're uh launching what we call in our resolution platform. Um we have a couple kinds of points of view here. One that AI for service is unique. It must h handle to our earlier part conversation unpredictability. It's high stakes a lot of times and a lot of times it's a customer that's interacting that's having an issue. So you really need to think about when an AI agent handles something when a human agent handles something to solve the customer's problem. And we've built the resolution platform platform to solve problems not just manage tickets. So it's integrating AI automation and h human expertise for real outcomes. Um I'll talk a little bit uh at relate about how one of our competitors is charging by per interaction. And companies and customers don't care if they have an interaction, they care if they get their problems solved. And so we're really pushing this resolution platform and the whole premise behind it and how we've built it is how do we go help companies solve their customers problems or their employees problems or their uh their business problems. And so that's what we're doing. So um other people are looking at isolated AI agents. We're delivering an intelligent coordinated AI and human network that's wellrained. It's like a you know a well-trained search and rescue team. It's not going to replace human agents. We think we have got a point of view that there's always going to be human agents. It's going to enhance them and it's going to take off a lot of the low um low difficult task off their table that um the the resolution platform can go solve. All right. So, can you just walk us through because we'd love to hear the practical examples. How would a customer basically use this platform? Yeah, a customer would use this platform say um we what we do is we come in uh for our larger customers. Uh for our smaller customers um we will go real quick analyze their ticket data or their customer interaction data and we say we think there's an opportunity to automate 40% of those interactions. Okay. And we think we can go do that with you in two or three months or two or three weeks or two or three days depending on the complexity. And um and so we're pushing some things in products to let them know what we think's going to happen. Like we think we can go automate 40% of resolutions. We think this is going to lower your cost to serve by X. And we think it's going to take your customer satisfaction up from depending on if you're doing seesat uh top box from a 4 to 4.4, if you're doing MPS from a 60 to 67. And so we talk about that with them. And then what they do is we go in through um you know an implementation with them in an adoption phase to go get that. We usually do an AB testing to show them um on the um on the on the platform that we are getting what we promised okay through an AB test and they are getting the customer satisfaction. One of the biggest worries we see with our customers is I believe you can go automate X percentage of my interactions. I'm I'm worried that customer satisfaction is going to go down and um companies love the loyalty loop. They love making sure that they have happy customers and that's the biggest worry. So a lot of times we go in and show through AB testing that customer satisfaction for the similar kind of interactions is actually going to go up and once we able to do that the the roll out the adoption usually goes pretty quick. That's on the AI agent side. On the human side, uh we've got a co-pilot solution that integrates with the AI agent solution. And so for those kind of interactions you want to go take to your humans, it's a really uh easy implementation. It's part of the Zenesk platform. You turn it on. Your human agents are getting assistant. They're getting suggestions. Hey, this is the ticket came in. This is the reply we're recommending for you. Would you like to accept it? Would you like to change the tone of voice uh that you're replying? Would you like to add something? would you like to say this is wrong and we need to go suggest a different reply and so that's what we go in and talk to our customers about to really really generally for everyone an easy implementation but the hardest thing is getting people over the hump that this is going to not degragate um uh um customer satisfaction. You said in your example that you can go in and say there's 40% of customer interactions that can be automated. uh where did that number come from and do you think that this is going to be like an average experience with the company? Yeah, so what we do is we get permissions from our uh customers and we take a look at their interaction data. So we use an algorithm uh that looks at all their interaction uh data. So uh we do about we process about five billion do five billion tickets or five five billion interactions a year almost. And so a company will have a subset of that. We look at the data and we say based upon what we know, we have over 10,000 AI customers right now. And those 10,000 customers, we think this password reset, we think this return, we think these 17 use cases can be automated. And we tell them this is 3% of your interactions with your customers. This is four, this is seven. And there's going to be some of those that are going to be really complex that are ultimately going to humans. We break all that down and we say these are the interaction types. these are the percentage of those interactions or use cases that we think we can automate and we give them that the data and then we say hey we've got 10,000 customers around the world using our uh resolution platform and you look like two or 30 hund of those customers and these are kind of resolution rates that they're getting from doing this automation. So it's not a um it's not a you know wet your thumb put it in the air and kind of guess this is based on uh old school uh a little bit uh algorithms machine learning to predict what kind of resolution rates our customers are going to get and again it gets into vertical the size of the company are they B2B B TOC what region are they in you know it's going to be different in Germany and even within Germany it's going to be different between Bavaria and u maybe um the Berlin you know region And so we get all of them. Okay, it's coming in as for this type of uh request, let's give them a human or it's coming in from this type of request, let's give them robots. And then uh I imagine the customers are able to if they need uh get to a human if the robot's not able to solve their issue. I mean for me I just type agent agent agent when I need a person and eventually that works 50% of the times I would say. So how's that going to work? Yeah. So we we try to design uh some some companies just accept hey we want to go automate all this. Others will say we want to automate this unless it's Alex. Alex is one of our VIP customers. So you can do customer. Oh that's interesting. Yeah. And Alex no matter what even if uh we could do an instantaneous password reset for Alex. We want to give him VIP service. So we connect Alex with a human being you know immediately. Okay. So you can put into those kind of um you can put into how you c you configure the system those kind of um you know segmentation of customers you can put in um if the you know if it goes too long you can go to a human. We again recommend we don't think we think a 100% of interactions are going to be AI related meaning uh it's either an AI agent or it's a human agent using AI but we think about 80% will be automated within the next 3 to 5 years leaving 20% of those interactions still to go to the human being and so we think it's important to have that whole resolution platform that looks at automation in the human experience other people are doing um you know disconnected things so if you've got this great AI agent that's really not tied to your human, you're going to get frustrated, Alex, even before you go agent, agent, agent, agent. You might have a conversation with the AI agent. You don't get total resolution. You want to have that all that data, that rich u conversation passed on to the human agent so they're not asking the same 15 questions again. You want them to start where the AI agent left off. And we think that's one of the advantages of the of the Zenesk resolution platform. We realize it's not going to be all automated and you've got to have these tight links. And even just as importantly, we figure out, we have something called quality assurance um that figures out looks at 100% of interactions, whether it's an AI agent or human agent. And we find out what interactions the humans are doing better. And then we try to train the AI agents with that information. We actually find out what the AI agents are doing better. And for those, you know, you still want to go to the humans for some of those VIPs. And we try to change train the humans to do better as well. So you have to have this kind of virtuous loop between human and AI agents and our resolution platform. Tom, first of all, if you're able to get the human that picks up the line to be queued into what I've spoken with the agent about, uh, you've done us all humanity a big service because all too often in customer service, you go through the same steps again. You put the same account in. So if you're able to make that more seamless, uh, not just myself, but I'm sure half, you know, half the planet, if not more, will be will be thrilled about this. And then you've brought up the the co-pilot a couple of times. So what is that going to look like in a customer service agents dashboard? And then we're going to move on to jobs because the questions about what will happen people jobs are starting to percolate. But I want to know what it looks like in the dashboard. Old school human agent would get a um you know information from a customer whether it came in in a message, a WhatsApp, uh Apple business chat, an email, a web form, you name it. Okay? And then they would craft their own reply and they'd have some canned replies on likely interactions um that they could, you know, like cut and paste basically. Okay, that's kind of old school human agent uh you know resolutions. What happens now is if you have even on a voice bot, but I'll we'll put voice the side any kind of digital interaction the platform should be suggesting a response for you. So it's not you pulling as a human agent, okay, data or response into a response. It's actually here's the suggested response. We looked at 17 similar customer interactions. This is the response that got the best uh response. Do again, do you want to personalize it anymore? Do you want to change the tone? Do you want to accept it? Do you want to edit it? And so it's kind of changing the human agent experience uh experience from a lot of times creating things on their own or you know pulling information in to more of an editor where they're deciding hey this is the right response. this maybe it's not the right response because um the um the co-pilot does not res realize how upset you are Alex because you had to type in agent 55 times to get to the human agent and you've got a really big issue and so then they can take over and edit and you know profusely apologize for that experience where maybe the suggested response wasn't there. So, we really think it's changing the human agent to higher level tasks to more of an editor to more uh less of a creator uh and able to respond more quickly and more accurately to customers. Yeah, look, I promise I'm not that guy who's uh getting mad and slamming agent. But anyway, I always have a lot I always have a lot of empathy for agents and I say, "Hey, I'm I'm especially on the phone. If I'm on the phone, I'm not frustrated with you. I realize you're doing your job. It's one of the most difficult jobs in the world because you don't get people to call and say you're doing a great job. Uh but I disagree with your policy or I disagree with the outcome that I'm getting on, you know, this particular service. Yeah. No, definitely it is uh without a doubt among I would say top 10 hardest jobs uh in the world to be able to sit and sometimes have to take take it and you know deal with with customers that are not happy. Um they call people call customer support lines often because they need an issue resolved and um and yeah I mean that's if you're doing that all day long that is that is tough tough work. Yeah. Most of those jobs have 50 to 100% turnover a year in a lot of companies and so really really high turnover because to your point it's a really difficult job. It's stressful. You need to learn a lot and it doesn't pay that uh well in a lot of situations. And so we think like co-pilot AI is going to help people have a you know better job that's higher paying hopefully and uh it's going to allow them to have better job satisfaction because they're able to you know answer customers uh interactions better. But will they have a job because you mentioned that you want to what did you say 80% of these interactions automated. So I I guess like companies aren't thrilled necessarily with the way they're doing customer support and customer experience because people are just bogged down handling handling basic queries and they don't take get a chance to take a lot of time to spend with their customers. So this might free the existing customer service reps to spend more time with people. However, there's going to be companies out there that will be like we can probably get the basics done with robots and I anticipate there will be layoffs. So what does the job outlook look like from your point of view here? So a couple things. One is uh we've seen the amount of interactions and we do some surveys on this double the last two to three years. Okay. And this is for a couple reasons and we think it's going to continue. U number one depending on what survey you look at uh there's between about 13 and 18% of the world economy's digital economy. and you still have, you know, 82 to 88% that is bricks and mortar. That's going to still that's going to continue to move more e-commerce. The more e-commerce happens, the more online or voice interactions that happen um just over time. And so we think uh that's one trend. Second trend is we've seen with other technology uh changes when you lower the barriers to entry, when you lower friction between customers and companies, interactions spike. So we have a point of view that interactions are going to go up between 3 and 5x the next 3 to 5 years. And so what we think is going to happen are interactions are going to go massively up. We're going to help companies automate 80% of them. And you're going to have about the same amount of human beings that you had before in the human agent role because of this, you know, uh bricks and morted e-commerce shift, lower uh friction, more um more um more uh instantaneous responses. And what we've seen with a lot of our companies is they want to get even better service. And so they realize when it goes to that 20%. So those that uh companies that have got to 80% or plus automation rates, they've actually usually kept their customer service flattish because that last 20% is usually really a really iate customer or a customer that has um a really really complex or really really important interaction. And if you nail that, a lot of times you're going to have a customer for life with a customer for the next 5 to 10 years. And so what we've seen so far is people flattened out hiring just to be clear on customer service uh human beings or slightly up. But what they're doing uh is um really making sure that those people are editors more highly skilled and really really about the customer experience and try to drive customer experience through the roof. And you might have higher retention because instead of people waiting on the phone for 10 minutes and you being pressured to solve a result, solve a case in 30 seconds, maybe the companies will say, you know what, take five minutes, talk it through. They'll be happier. It'll be a better interaction for the customer. Alex, you're spot on. There's um in the, you know, phone contact center world, there's a the term called average handle time. And there was always how can you get AHT down? We're starting to see a difference. Exactly what you're saying is hey let's talk about the resolution and do we have a positive resolution for the customer. Okay that's why again why we're launching the resolution platform less about AHT and more about because you have a little more capacity if you do the automation right more about customer satisfaction more about the loyalty loop. Yeah. And you've written a little bit about Sebastian Simonowsk's uh plan to purge customer service at um at CLA and you said that's uh not exactly the right way to go about things. We've had them on and I think they did have to roll back a little bit in terms of their all allin on AI moment. Yeah. You know, I I love when people are like um I don't know Sebastian personally, but like I really respect Clara, you know, and what he's done. um you know with the afterpay market just like created a market himself. I think absolutely fantastic what uh Sebastian Clark have done. you know, he made some bold predictions about um what you could do uh from an AI perspective and um you know, he he said he did it a lot on his own. And what I tell our customers, if you want to go spend, you know, tens of millions of dollars upfront and you know, tens of millions or five plus million a year on perfecting that. You want to take some engineers and not every company has engineers like Clara has engineers, software engineers to go design this. You could probably get a decent customer service solution. I I don't think it's going to keep up because we've got thousands of engineers making sure that we're delivering the best for our companies, but you could do it. Um, but I think what he rolled back a little bit was um there is still the human element at the end of the day where automation can go solve a lot of things, but it can't solve everything. And so, you know, what we look one of our, you know, missions um is that we want to democratize AI. Like I said, there's a really small percentage of companies can do what CLA did because you have to have the the money, the technical knowledge, the software engineers, you know, etc. to go do this on yourself working with Open AI and uh you know, on a really really or you know are um anthropic or someone else on a close basis for the 99% of customers that aren't able to do that, Zenesk wants to democratize this so they can get a lot of use out of AI on the resolution platform. will automate a ton, but we've got a strong point of view that humans beings are the customers or the employees at the end of the day. And you know, you might start out like we talked about before, Alex, your personal bot talking to uh a company's AI agent. Uh if you don't get resolution, your bot might go to a human agent. Ultimately, it's going to be back to the human going to the human on a really, really difficult problem. And we always say the customer is always human. And uh we think um you know CLA got a lot of things right but we think what they got wrong is uh you know they for a little bit they thought they could automate everything and where you know maybe forgetting a little bit that the customer is always human. One interesting thing that Sebastian told me in our interview was that he uh actually was able to do as much automation as he was because his customer service wasn't built in the right way and there were a lot of phone trees that were built the wrong way and therefore it was easier to automate it with bots and maybe if you just build in the right way in the first place you won't have to resort to this and and I think that's one of the like technology technical debt for the last 20 years um people buil built these logic Tre's, you know, phone and IVR, you know, interactive voice response system that that just became spaghetti code. And so I think there's an opportunity where people have got to refresh. And what I always recommend to customers is if you're just going to have the same process, the same people and do it the same way, a technology upgrade is not going to really help you. You really got to go reimagine processes, people, your customers, how you want to serve them better. And so I think a lot of places are in a similar situation like Clara. grew up really fast um and they had kind of, you know, these um old school decision trees and old school workflows. It's really really easy to have a best-in-class um customer service platform right now. And we think Zenesk can help customers do that because uh you know we like to think that we're easy to implement, easy to use, and easy to value. And we're democratizing AI uh for the thousands and tens of thousands and hundreds of thousands and millions of customers that um you know don't have the knowhow to do it themselves. Yeah. On those phone trees, I'm just hitting zero. So I think I'm bad. I think I'm just concluding that I'm the most annoying person. Alex, I'm I'm old enough to remember where I thought those phone trees were a little cool where uh the first before the web you could call and get your bank balance. And I thought going you'd have to go to the bank. That was really cool. But those those IVRs, phone trees very quickly outlive their usefulness. I'm with you. All right, we are about out of time. I would just love to hear before we leave state of Zenesk because you've had a very interesting recent history and just give us a view of uh what the company's up to and where you're heading. Yeah, we're we're pretty excited. Um we just passed our we have about 100,000 customers overall and we just passed uh having 10,000 customers on AI. Um we're excited that you know the first act of Zenesk was revolutioning um uh revolutionizing customer service help desk uh you know kind of CRM customer relationship management for service and we think we're in this second phase now of disrupting the marketplace again uh and uh disrupting both employee service and customer service and uh as you just mentioned there's a lot of legacy out there and um if you want a beautifully simple really powerful solution. Um, we think Zenesk is the right way to go. We've really fundamentally shifted the company the last two years. Like I said, we've gone from zero to 10,000 AI customers. We think we're going to have over 20,000 AI customers by the end of the year. Um, we've done some really interesting acquisitions. Um, M&A has been a big organic growth, but M&A has been a really, really key part of that growth strategy. And, um, we're really excited about being on the forefront of this agentic AI revolution. There might be some hype along the way, but um you know in our last quarter we saw u I'll just give you a range between 20 and 30% of our bookings being AI related uh and you know versus zero two years ago. So kind of getting to back to your first point. It's not hype. It's uh and I'm using bookings as a proxy for uh customer satisfaction. Our customers wouldn't be buying from us unless they really adopted these great solutions and they got a lot of value from it. Well, the Zenesk resolution platform is out today. If people are interested in learning more, where can they find more information? Uh, they can go to zenesk.com um or uh, you know, don't hesitate to uh go get on our website. Um, you know, we've just rolled out. We were one of those companies that automate it too much as well. And so now, if you need to, you can get click to call uh you can get a human being uh after you go through our AI agent, which we think is going to solve your problem. Go to zenes.com and we'd love to go interact with you. Good stuff. Well, Tom, great to see you and thanks so much for joining the show. Thanks a lot, Alex. Really appreciate your time. Thank you so much, Tom. And thanks to everybody for watching. will be back on the channel with another interview