[00:00:21] Speaker A: Hello listeners, and thanks for tuning in to this edition of Metro G's metrocast podcast. I am Beth Schultz, Vice president of research and principal analyst at metrogy, and I'm excited to introduce my guest for today, Craig Walker. Craig is the founder and CEO of DialPad, which he launched a dozen plus years ago or so and built from the ground up as an AI first communications platform for businesses. So he came to Dialpad having been on the forefront of voice technology dating back to the early 2000s or so when he launched Grand Central Communications VoIP provider, later acquired by Google and which eventually became Google Voice. Now Craig is applying his visionary chops to agentic AI. And today we'll be talking about his company, how. How his company rather is moving beyond simple chatbots to create true autonomous AI agents and what that means for the future, both for CX and employee communications. So Craig, welcome to MetroSight and great.
[00:01:20] Speaker B: To be here, Beth.
[00:01:21] Speaker A: So, so I touched on this in the intro, but throughout your career you've shown a real entrepreneurial spirit and that's taken you down quite a few paths. I didn't go into all of them, but ultimately it led you to found Dialpad, but you found a Dialpad, what, a dozen or roughly a dozen years ago, right?
[00:01:39] Speaker B: 2011. Yeah, 14, 15 years.
[00:01:42] Speaker A: 15 years.
[00:01:42] Speaker B: Actually we left Google at the end of 2010. So technically 15 years ago, literally November 1st.
[00:01:51] Speaker A: Okay, so that was a while. That was a while ago. So what kind of feed this entrepreneurial drive and spirit that you have, clearly, that you've exhibited throughout your career, you've been at Dialpad now for a long time. So what keeps you going there?
[00:02:08] Speaker B: You know, it's one of those things where when you see products or you see solutions that you don't think are great and you know, you could build a better one or you think you can, it's really exciting to try to go do that. And I look back to when I was at running Google Voice. After Grand Central got bought, my co founder and I, Vincent Paquet, we shared this tiny little office at Google and we're running Google Voice. We had tens and tens of millions of users on our platform. I still had like a Nortel desk phone on my desk. And I realized, look, you know, like there's another mountain to climb. And it's, it's really this whole enterprise communications world, which at the time was not cloud based, was, was not innovative, was basically the same it had been for decades. And we realized, okay, if we could come and build a new brand new station and unified Platform from the start. We could do some really interesting things and go take on the folks who had been dominating that space beforehand. So that was the impetus to do it. And it's this massive market and there's this massive opportunity and along the way it just got more and more interesting. So we started with a conferencing product and we added a UCAS or the phone system product. Then we added a contact center product. Then in 2018 we bought an AI company and built it into the core, that unified platform. And then I get excited today, I think through all those things and now it's gotten us to be in a really, really great position to go be very successful in the Agentic market. And Agentic is probably better than all the things I mentioned beforehand. So this is like, it's almost like you work 15 years to get to the starting line, but you're at the starting line in pretty good sh.
[00:03:50] Speaker A: Yeah, it's kind of one of those, I don't know, it's a little scary in a way to think about that as agentic being, you know, more, more, more than everything you just talked about. Right, because.
[00:04:00] Speaker B: Right.
[00:04:04] Speaker A: There's so much promise, but there's so much that could go wrong, you know, on, on the way to realizing that promise. But anyways, so speaking of agentic, last month you introduced your Agentic roadmap for dial pad.
[00:04:17] Speaker B: Right.
[00:04:17] Speaker A: And when you did, you had stated that CX as we know it is dead or it should be dead or will be dead soon. So let's walk through that a little bit. First of all, what do you mean when you say cx? Is that a statement about the level of automation and AI that we're able to achieve today or is it broader than that?
[00:04:36] Speaker B: Yeah, I think like the way I've always seen, in a way everyone's kind of attacked the customer support. And let's just go to CX for this conversation. Let's just call it, you know, customer support functions for businesses worldwide. Generally they attack it by, okay, I hire a bunch of people, train them. Generally I hire a bunch of people in a low cost geography, whether, you know, in the US or, or even to, you know, places around the globe looking for low cost labor that you're going to try to train to know everything about your product that are then going to handle these issues that come up when people call in to get their problem solved. And then like the first steps of automation over the last decade were all about how do I have some automation system that keeps that from getting these agents So I can just get rid of a bunch of things. It was all about almost like just avoidance, like cost containment and let's try to just not make this an excellent experience. Let's just try to lower our costs and not have so much stuff go to humans. And it just was not, it's just not a great experience. You see that through anytime you call into a thing and there's a hold queue, or anytime you go to a chatbot and it doesn't answer your question or doesn't do anything for you, just like give you links to help center articles or anytime that you're not, you know, the voice one is the worst. The voice automation is the worst. And it's not can't do anything for you, tries to give you some information similar to a chatbot, but now voice and how many times do you hear people just screaming agent, operator, representative, press zero, a thousand times. Like it happens all the time. So that reactive kind of negative experience of what automation has been or if you didn't try to do a bunch of automation, it's like how do I get even more people to handle peaks? There's not long hold times. That entire messy CX world should be dead and will be dead soon. And we're going to try to help it accelerate that debt. And it should be more, it should be more of an outcome based approach of hey, whether you come in on a chat or you come in on a phone call, the agentic, the intelligent agentic agent should be the first thing you interact with. And if it's a simple request or if it's something that the, the agent can handle, scheduling, onboarding, you know, like screening, interview type things, like just a whole host of things. It can handle those 24, 7 answers on the first call or responds immediately on a chatbot and solves the problem. It doesn't just give you information, actually solves the problem. And if it can't, it can then seamlessly hand it over to a human who's now not been doing 6,000 calls of how do I change my password? It's now hands off all that information with a bunch of, you know, all the information about what you've talked about already. So then the human can intelligently take over and solve the problem that the agent couldn't. That's an outcome based approach. And we've seen, you know, just in our own eap, we haven't made it generally available yet, but we're seeing like 60 to 70% deflection rates already just with our earliest customers and we're seeing CSAT scores actually go up. So you have this incredible, this incredible, you know, virtuous cycle of customers are getting their questions answered accurately or having action taken for them immediately.
Humans are now dealing with less of a mundane and solving the more complicated problems. It saves money because you don't need as many agents and your CSAT scores go up like, it's like kind of the win, win, win, win. So that's, that's what I mean when I say those things.
[00:08:13] Speaker A: Okay. And our CX research definitely shows those same trends that you're seeing with csat, et cetera. But what I'm kind of wondering about, we're seeing early users, early adoption, I would say at this point. How quickly is this going to change? I mean, a year ago, Agentic was barely spoken and now it's the only thing that's talked about or one of the only things. Right. It's so hyped up right now. So at what pace do you see all of this changing?
[00:08:39] Speaker B: You know, it's funny because there's, you see the promise and people have been talking about it like over the last year, like, oh my God, this is exactly kind of like, you know, what, what the future will look like. And I think that the early results have been a little bit disappointing. Right. Customers who have deployed it have heard a ton of use cases of wow, you know, like, it's a lot of work. It didn't really provide solve those things we wanted to solve and it's not there yet. So I think like any new technology, we're not at the point where it's there or it's getting there. And that's why we're really focused on very specific skills and very specific use cases to make it bulletproof and make sure your first experience with it is going to be fantastic and you're going to get these great benefits. And so I think we're, you know, like it's, it's, it's moving very fast. The, the results we're getting from our, our early Adapter Pro program users are super positive.
So I think now's the time and I think it's going to move really.
[00:09:40] Speaker A: Quickly now as long as they, because you kind of reference. So a lot, a lot of things can go wrong. Like there's a lot of work that needs to go into within a company to get the organization ready to get the infrastructure, the knowledge infrastructure in particular, ready for any AI. But Agentic AI in particular, I would.
[00:09:56] Speaker B: Say, by the way, if you get it wrong, I mean the last thing you want to do is set up an automated machine that's perpetually doing the wrong thing. Right. So you got to have a ton of safeguards and stuff. And we built capabilities where you can literally take your prior conversations that you've had with customers and see how the human solved them. So you're not risking it on actual live data, but this is stuff that had happened before.
And then you can run it through, we call it our proving ground, but you run it through our proving ground and you can keep tweaking your prompts and we'll coach you or it'll automatically coach you on, on suggested changes until you have very, very high confidence that this is going to be as good or better than the humans who did it before. And then you put it out in the wild. And those are the kind of guardrails and the kind of, the confidence inspiring things that customers need in order to really trust, you know, this automated agent to actually take action. So you do have to do things like that.
[00:10:50] Speaker A: Yeah, the assurance. And we've seen a lot of interest in AI assurance products now.
[00:10:54] Speaker B: Right.
[00:10:55] Speaker A: Because of that or strategies, processes.
[00:10:58] Speaker B: Right.
[00:10:59] Speaker A: So do you think of that as a differentiator, as a, you know, there's so much, there's so much conversation about agentic AI, so many products being pitched out there. So why dial pad versus any other CX or communications platform provider in general?
[00:11:14] Speaker B: Yeah, great question. I think, I think it goes back to the, you know, just kind of like our founding philosophies and principles of we are going to build this microservices fully modern Google type engineered platform that's going to take on all these legacy folks who weren't built that way. And by doing that we've been able to easily add in capabilities and acquire skills and talent to that platform that others would just try to bolt on, but they don't have a foundation that makes it easy to bolt on. So when you think of legacy providers, it's really tough to take a, you know, and kind of earlier architected platform and try to just bolt things onto it. So number one, I think we have a platform advantage. Number two, that platform was always built to do everything. And then number three, and most importantly is we made a very, you know, strategic acquisition back in 2018 when we acquired a real time artificial intelligence company whose entire mission was to understand business conversations and to be able to do great real time coaching and AIC sat and sentiment and QA tracking and all these things based on live conversations. And with that came a team of PhDs in AI. And that was again, this was seven and a half years ago. So that team has grown and grown over the years. We now, you know, we filed or we have been awarded 21 AI patents. We have a ton of people working on just AI.
So we now have not only the team and the platform and the know how, but we also have the experience and the experience of understanding of like okay, if you've, when these types of business conversations happen, these are the next best things. So we actually have a ton of history of how to understand what these conversations mean. So I think we're uniquely positioned to have all of that versus you know, when gender of AI came out three and a half, four years ago. Every single one of our competitors is like, you know, they changed their websites instantly. They said they, they were AI first and then they were like literally scrambling around trying to get ChatGPT and others to allow them to use or you know, even give them the capacity to get access to their AI engines. So having our own models, our own engines and being able to supplement those with large language models and have the know how, how to do it well and to scale it is a huge head start versus everyone who jumped in the game recently and everyone who therefore if you jumped in recently, you're relying solely on these generic, expensive large language models of other, other parties. So it just gives you, you know, just a better solution.
[00:13:53] Speaker A: So Craig, you mentioned a couple of times the idea of skills and Dialpad's strategy. It's a Gentek AI strategy is very skills based and I think ultimately you're going to have this portfolio that comprises a lot of very granular skills based agents. So explain why you think some more about why you think this is approach to take.
What does that portfolio look like and how is it going to manifest itself over the next couple of years.
[00:14:19] Speaker B: To our earlier conversation about you want it to be highly accurate. You want it to prove to the customers that this is really, really useful and really, really valuable. And to do that our view is you make really precise skills and you make a lot of them, but each precise skill does that precise thing very, very well. So things like scheduling, interviewing, general support, tickets, you think of like in the healthcare space.
It's interesting because we're leaning in their heart is the biggest consumers of healthcare services are people later in life 65 years or older. If you're 65 years older, they saw a study that only 7% of them are comfortable doing things online, particularly around healthcare. They'll pick up the phone to do everything. So the amount of conversations that people have on scheduling, on, you know, do I need to fast before my appointment or my procedure? And what's the follow up? All these types of things. You've been in a doctor's office, that's phone ringing all the time, the person's checking in, they're trying to do this, they're faxing things around, they're dealing with insurance. It's a madhouse. If that could just happen. 24. The precise scheduling skill, precise new patient onboarding skill, a precise surgery follow up skill. Those are wonderful things. And again, you start with the things that are going to be the lowest hanging fruit and then you continue to add, add, add, add, add, add, add. And then once you have all these super accurate skills, you then start having them work together. So yeah, okay, this one does that. But then the conversation turns another direction, they ask for something else. Oh, do we have a skill for that? Yes, we do. And then you take it. So I think it's the right approach. It's like how things one bite at a time. So I think that's the right way to do it. And if you try to do too much with one and there's just going to be some generic agent that can do everything, I think that's going to be a much harder road.
[00:16:12] Speaker A: How do you think about this in context of that human agent? Do we, maybe we don't need to hire as many human agents. Maybe we start replacing some human agents, I don't know. But how do you think about empowering the human agent of the future In a world of agentic AI?
[00:16:30] Speaker B: The human in the loop is really, really important.
And think of it this way. We think of it as like the agent or the AI agent. They're going to handle the things that are repetitive, they're going to handle things that are easier and they'll get, as we mentioned, they'll get smarter and better over time and there'll be more advanced skills. But the things that require empathy and judgment and more complicated things or conversations that people just frankly want to speak to a human about, again, healthcare comes to mind on that too. It's going to make the agents so much better because one, there will already have been kind of like this pre conversation. So the agent's going to get this handed off to them with all that information available to them, so they're not starting cold and asking you again for your account number and all that type of stuff. Secondly though, the human's going to get coached by the AI as well. So Even though like maybe that particular skill didn't exist, they handle it autonomously.
The AI still has a lot of wisdom and understands everything in your entire knowledge base and even some proprietary things you want to train it on. So it's, they're coaching the agent in real time as well. And then frankly the nice thing about having this in one unified platform is let's say there's some wrap up procedure or scheduling at the end of the thing, the agent or the human can actually send it right back to the agent to go handle that wrap up stuff too. So we really see this as a, like working together, making the humans better. But like, you know, the reality is will you need as many human agents if a lot of the low hanging non skill, really easier things to handle with an agent get taken care of by agents? Yeah, you're not going to. And that's, that's good. Who wants to do that? All day long of resetting passwords and updating billing addresses, you need upskilled agents.
[00:18:14] Speaker A: Right, right. Human agents. And it's a vice versa thing too. Right. You said that the AI, that the human agents will learn from the AI. Well, the AI will learn from the human agents attacks as well.
[00:18:25] Speaker B: Attacks, right, yeah, that's gonna, that's what this really cool process. If you have, let's say the same type of topic or action comes up the agent can't handle, then it can understand and learn from how the human handled it. And then, you know, we can service that data to a company and say, hey, would you like to build a skill for this thing? This is handling, you know, this much of stuff. And, and then the agent itself, the agentic stuff will get even smarter. And if we can get to the point where that becomes, you know, automatic, that'd be great. And think of it. When something unforeseen happens, like let's say there's, you know, a customer has, or a company has some kind of product problem, then all of a sudden, you know, let's say Netflix stops working or something. You know, like all of a sudden everyone in the world is calling and the agentic thing's not trained on it because this is an unexpected thing and didn't know it might happen. But if the agent were able to learn from, oh my God, that there's a million people calling on this, they're saying the exact same thing. I can learn what the answer is now. I can just handle those. And then if I can just handle those, that queue goes from 4 hours to 2 seconds and it kind of Goes away. And so that's a, that's really exciting, a really exciting endpoint that I think we could get to.
[00:19:42] Speaker A: And to your point earlier, CSAT goes up, right?
[00:19:45] Speaker B: Yeah, absolutely.
[00:19:49] Speaker A: So I wanted to touch on, and I know this is a, a bigger conversation, but how do you approach pricing and monetization for AI agents and our enterprises ready to really deal with the kind of the fully agentic environment?
[00:20:04] Speaker B: Two good questions on the pricing and monetization. Obviously you know, traditionally it's been the seek based pricing for, you know, for human agents. The agentic one is going to be much more consumption based. And there's models out there where some folks are doing, we get paid by resolved cases, others are doing it by, we're going to do it by every conversation. We're playing with both of those models now during the cap process and determining which one we think will be best for us. But yeah, I think that's, that's obvious that's going to happen and our enterprise was ready for it. The benefits that enterprise gains by being ready for it are so great that, that you see the, it's not, it's not us pushing it on them, it's them pulling from us. Literally. Our wait list for when we announced our eip, our wait list just exploded when we had our last customer advisory board, two customer advisory boards ago, no one was asking for, for Agentix Solutions last May, everyone was asking for it. Like they're pulling, they're pulling. And the interesting thing is the larger the company, the larger the enterprise, the more they have to gain by doing this. And not only the more they can gain, they also have the team, you know, the IT infrastructure and the team and the people who understand or are tasked with understanding this and evaluating vendors and then making sure that these safe things to put out in the wild. So it's a real interesting thing where customers are demanding it and the biggest customers are demanding it the most whether they're ready or not. You know there's going to be this whole process of going through that but there's certainly, they're certainly banging the table for it.
[00:21:44] Speaker A: Yeah. And a little bit different than we've seen with some other new technologies come into play. More at the small company size can be more flexible and more. But not in this case. It is the larger companies driving it.
[00:21:56] Speaker B: Yeah, yeah, totally. It's interesting. Yeah. Because you know, I've been in the Silicon Valley my whole life. Like it's always been SMB. They're willing to try anything.
They're the first to Go. And then like it gets the credibility and maybe they add some more kind of enterprise features or mid market features. So then bigger companies and then, you know, you start getting a couple enterprise wins, you cross the chasm and then you have, you have to build out all the things to support the largest companies. This is entirely opposite and it's, I've never seen another, another technology like that.
[00:22:27] Speaker A: I love talking about what's going on in the world of customer experience, customer support, but I don't want to overlook the UCAAS side of your business. So just give me kind of a sense. Where does Agentic AI lead the UC industry over the next five years? And how does Agentic AI impact the integration of UCaaS and CCAS?
[00:22:45] Speaker B: Yeah, it's interesting because, you know, a lot of CCAAs companies are CCAAs only they don't have a UCAAS solution.
And I think on the UCAAS side you're going to see a blurring. I mean, the UCAS CCAS thing is going to be a blur. And if you had all of your, you know, just ucas, let's call it just, you know, business conversations, messaging, meetings, things like that, and being able to have the same capabilities of understanding everything that's going in there. And you see it already, there's recaps of calls and there's action items pulled out and there's, you know, what was the purpose determined by it? It's going to blur to the point where like, I think, you know, like calling a company, let's say a company puts in a UCAS solution.
It'd be really cool in the future where you just call a company. And it's not like I'm not calling their support, I'm not calling their sales, I'm not calling their hr, I'm not calling their finance, I'm not calling the receptionist, I'm just calling the company. It's like, hello, this is the iPad, how can I help you? I said, oh, I'm a, I'm a user, have a question about, you know, which phones are compatible or something. It's like, oh yeah, okay, let me do which ones you look at and just solve that right there. Or it's like, hey, I'm a potential user. I want to know if it could do this or that, here's how big I am and where I'm located. It's like, oh great, let me, let me set you up. Or it's like, hey, you know, I want to know what jobs are available and I'd like to apply for them. Okay, great. Let's, let's have a conversation. Like just having that, that unification is going to be so cool and that's a unified future. That is really attractive to me. And I think for, for the UCAAS only companies, it's going to be a bit of a challenge because they're missing out on all that others, all those other conversations.
[00:24:24] Speaker A: Yeah, I mean, we, we definitely, in our research, we definitely see that increasing demand for UCAAS CCAAS integration, particularly with the same, from the same vendor, within the same.
[00:24:35] Speaker B: Yeah, if you think about it, you want all these analytics unified, you want the age under this AI, you know, capabilities to learn from all of it. You can't, you don't want to go from one AI thing to another AI thing to a third AI thing to someone else. You want it all in one place. And you want one admin panel. You want one. Yeah, I think the analytics is the key, but enterprises care a lot about access control and things like that. You just need it all in one. Otherwise it becomes a big headache to have to go do it in multiple places.
[00:25:04] Speaker A: Well, Craig, I think that was all the questions I have for you today and it was really fun talking to you. I really appreciate you joining us and I really do look forward to hearing more about how Dialpad and its customers move ahead on a gentleman. So we will be watching.
And then listeners, thank you for joining us as well. Be sure to Visit
[email protected] to check out our AI and our CX and our UC research. And as always, we're happy to hear from you. Feel free to reach out to me via email@bethro g.com and that is all for now. On behalf of the Metro G team, Goodbye till next time and take care.
