Pod Insights Eye On A.I. #258 Brian Peterson: How Dialpad is Building the Future of AI-Powered Communication About this episode What happens when you combine 8 billion minutes of voice data with a full-stack AI engine?
Yes, that's what Dialpad is doing.
In this episode, Brian Peterson, CTO and Co-Founder, breaks down how they've built an AI-powered communications platform from the ground up.
From real-time sales coaching and AI-driven support agents to predictive analytics that can spot churn before it happens, Brian shares why owning the full stack — infrastructure, LLMs, and data is the only way to deliver truly intelligent customer experience.
If you're curious about the future of AI in business communication, this is the episode to watch.
Stay Updated:
Craig Smith on X:https://x.com/craigss
Eye on A. I. on X: https://x.com/EyeOn_AI
(00:00) Brian's Founding Story
(03:56) What Dialpad Actually Does Today
(05:17) Is Voice the Most Valuable Untapped Data Source?
(07:41) Inside DialpadGPT
(10:10) AI Solutions for Sales, Support & Collaboration
(12:24) Owning the Entire Customer Journey with Unified Comms
(14:11) How Dialpad Stays Ahead in the AI Race
(17:50) Real-Time AI Coaching & Playbooks
(22:32) Why Most Enterprises are Behind in AI Adoption
(25:28) Action-Oriented AI Agents
(32:40) What's Next for AI in Customer Communication
Transcript Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.
0:00 Speaker 00 We're an AI-powered customer communications platform and intelligence platform. So we do things like full contact center with omni-channel. If you want to do online chat with your customers, social media chat with your customers, like WhatsApp integration, email. We do the full call center as well, all integrated with one full customer journey and all that. And then we do, on top of that, provide basically messaging, voice, video for everything you need. Full communication platform worldwide. And then we run AI on top of it. So we, I mean, our focus is on customer communication and whatever's going to make that better or more efficient. It's a combination of all of it. I'm not a big believer that it's going to take over the world and you're not going to need a support agent anytime soon. And even if it does, it'll be able to do some really...
0:46 Craig S. Smith? So Brian, tell me, introduce to listeners who you are and where you, what your background is as far as it's relevant to DialPad and what DialPad does.
1:02 Speaker 00 Yeah, for sure. So my background, real quick background, is I've been in communications and tech for over 20 years. I was back at Google. I started at Google when I came out of college in 2002. There was about, I think, 600 people there.
1:17 Craig S. Smith? Wow.
1:17 Speaker 00 So I got to live a lot of the growth of Silicon Valley and got to live it all. So I was there, I was there for eight years, living through like going from 600 to, I don't know, it was like 15,000 by the time I left. And then while I was there, I learned obviously a lot about cloud infrastructure and just Google infrastructure too. And my co-founder from DialPad, he got acquired through his company, which was called Grand Central at the time. And they were a phone system that rings all your different phones. And this is back in 2005. So this is really state-of-the-art stuff back then. They got acquired and then I got brought in and introduced to them because I already knew the technology at Google and the cloud stuff.
2:01 Speaker 00 And they needed help porting their whole system to Google. And so I came in, helped them migrate everything to Google, architecture and infrastructure. And that became Google Voice.
2:12 Craig S. Smith? Okay.
2:12 Speaker 00 So we were the team that built Google Voice. And then we were at Google and we're like, Google Voice was mostly a consumer product. And at the time, we're like, this makes a lot of sense for businesses. The same technology. And we didn't think anything was that great out there. And so we left and that's how Dialpad started. Google funded us and we even have Rich Miner on our board, the co-founder of Android.
2:36 Craig S. Smith? Sure.
2:36 Speaker 00 So we are all in on this and we're doing it now for 14 years.
2:41 Craig S. Smith? Wow.
2:43 Speaker 00 So yeah, that's kind of our quick background, how we got started.
2:45 Craig S. Smith? Yeah, Google Voice still exists.
2:48 Speaker 00 Yeah, it does.
2:49 Craig S. Smith? Yeah. As a consumer product.
2:50 Speaker 00 It does, and it also is a business product. So Google did eventually sort of build it out for their, for their workspaces product.
2:56 Craig S. Smith? Yeah.
2:56 Speaker 00 But it's, it's pretty, it's pretty lightweight. Like we're, we're part of the more advanced version and full contact center and all that, too.
3:03 Craig S. Smith? Yeah, it's interesting, things like Google Voice that they don't seem to promote at all. I mean, I've, I've heard about it through friends. I tried it years ago, but they seem to have these products that are created, pushed out, and then they, you gotta find them.
3:25 Speaker 00 Yep, or they cancel them. Yeah. Yeah, no, they still are going strong on it. We don't really cross over much with them. I think it's still a really good product if you want to get like a second phone number.
3:36 Craig S. Smith? Yeah.
3:37 Speaker 00 So you don't have to give out your personal cell phone number. But I, yeah, we don't really cross over much with them. They seem to be mostly still consumer from what we can tell. But yeah, totally, totally agree.
3:47 Craig S. Smith? Okay, so what does Dialpad do?
3:50 Speaker 00 So we, the, the quick phrase is we're an AI-powered, like customer communications platform and intelligence platform. So we do things like full contact center with Omni Channel. If you want to do, you know, online chat with your customers, social media chat with your customers, like WhatsApp integration, email. And we do the full call center as well, all integrated with one full customer journey and all that. And then we do, on top of that, provide basically messaging, voice, video for everything. Anything you need, full communication platform, worldwide, and then we run AI on top of it. So we, we bought a company about eight years ago called Talk IQ, which was way before, obviously, this big boom in like this version of AI.
4:33 Speaker 00 And they did voice AI. And we knew, we already knew, like, when we started building this, that the future of communications is gonna involve a lot of AI, right? Like, they need to go hand in hand. And so we kind of knew that was happening and we saw this company and we said, hey, this is a perfect fit. They could do live transcription and coaching on a live sales call. And we're like, if we can connect that with our platform, this is just gonna be gold. And that's what we've been doing for the last eight years is building out that team, connecting all the pipes. And now, like, as far as I know, we're one of the few people who can do, like, real live coaching. Things like, you know, if someone asks a question on a live phone call, we can provide the answer. Not a chat, but like an actual live conversation like we're having right now.
5:17 Speaker 00 It can respond and that's obviously super useful for support teams. It's really useful for sales. So like, if someone calls in and they're like, want to ask about your competitor, right? You might be a new sales rep and you might not remember how to answer, how do I go against my competitor? It would detect, it detects the question and it automatically pops up for the agent on their screen, the sales rep. Hey, here's how you should answer this, right? And that's, that's a feature we've had for seven years and it's one of the biggest sellers we have. But we do other things, like all kinds of stuff with this data. And we have all this vast amount of data. We have eight billion minutes of like sales and support conversations running through our platform. And so we do a lot of other things with AI, like customer satisfaction is another big feature of ours. We have automatic AI customer satisfaction.
5:59 Speaker 00 And I know if you've ever called into something and at the end of the call, it says, stay on the line for a survey. Right. Right? And no one ever does it. Or if you stay on the line, it's usually because you're pissed off. Right. And then so it's probably gonna be a bad rating. Well, we built a model that's all our own, that does analyzes the customer satisfaction of every single conversation to 90% accuracy. And so now, instead of like only having 5% of your conversations, like graded and customer satisfaction rated, we're doing, you're doing 100%, right? And so then that can tell you, like, who are your best sales reps, who are your best agents? Who are the agents who, when they talk to you, the customers are the least happy, right? Like, that's super important information. Or what are they complaining about, right? You have, you have millions of minutes on your call center in a, a month. What, what are they talking about?
6:44 Speaker 00 Like, are they complaining? What features are they complaining about? So like, that's our thing is we're unlocking. All that data too. And so like our other slogan is like voice is the last offline data set. It was really, if you think about business, it's the only thing that isn't tracked.
6:58 Craig S. Smith? Yeah.
6:59 Speaker 00 Right? Because it's so, it's hard to do. You have to transcribe it. You have to then analyze it with natural language processing. And the way we've been able to scale is we built all of it ourselves. So all the AI we do, our own transcription, we do our own transcription with high accuracy. We do our own large language model, which we call DialPad GPT, that analyzes all these calls. Um, we have a vast amount of PhDs and AI people at our company. And so we've just been cranking away for the last seven years, building out all these things. And it's that like perfect combination of like AI with communications. So that's, that's our thing.
7:34 Craig S. Smith? The, the model, uh, is that a fine tuned, open source model? Is that built on law more?
7:41 Speaker 00 Yeah. So we took, we took everything that's pretty much open source and we combine it. We fine tune it. We have our own recipes of tweaking these models to get the best out of them. And then we take our like eight plus billion minutes of conversations that are like a hundred percent relevant, right? Cause they're all real sales calls. They're all real support calls. And we use that to fine tune them. And then that in turn gets us super high accuracy, better than chat GPT and better than like Google or Amazon's transcription, all because it's with perfect data. And if you like use chat GPT, it's amazing and it can do a lot of things, but it's built around data of the whole worldwide web. Which is good, but it's also bad. Cause
8:23 Craig S. Smith? There's a lot of junk.
8:24 Speaker 00 Yeah, there's fake news and you know, or like things that just aren't relevant to your business. And with us, like you need it to be so relevant because your conversations are very different than another business's conversations. That's right.
8:37 Craig S. Smith? Yeah. Uh, in the customer, uh, are enterprises or BPOs or where, where are you selling into?
8:47 Speaker 00 Yeah, we're selling to almost every type of company in every vertical. And I think that's one of the best parts about DialPad. And we pitched this to like people who we recruit. Is the coolest thing about our product is literally every business in the world needs our software. It's it's right. It's customer communications, right? It doesn't matter what size you are, one to 10,000. You're going to need it. But we're in we're in about, you know, a third enterprise, a third mid-market, a third SMB. In all kinds of verticals, retail, recruiting, you know, healthcare, everything like Uber is a big customer of ours. You know, Motorola, Randstad is like the largest, one of the largest recruiting firms in the in the world.
9:29 Speaker 00 Los Angeles Chargers is are using us. Netflix is using us. So we have a large and vast range of and we're in over 50 countries. So we've got a pretty big international presence, too.
9:39 Craig S. Smith? Yeah. And and is it primarily customer service or or customer, you know, customer discovery? I mean, when someone and and do you have a voice or a chatbot interface? That that handles most stuff and then hands off to a live agent? If need be, if it's in that space, or is it supporting sales calls where it's? A human calling out and they just need support?
10:10 Speaker 00 Yeah, we do. So we have. We have basically like three, three different things we're selling to. We're selling to salespeople. So we have a product called DialPad Sell, which is for the sales team, both inbound and outbound. So like, you want to put a phone number on your website for sales leads, right? We handle all of that. We also handle outbound. If you're like a BDR calling out people or even messaging, right? We support messaging, too. We have a very large contact center product as well for support teams. So for your support contact center. And then we also have DialPad Connect, which is kind of the knowledge worker. So like, say you you're somebody who just doesn't want to use your personal cellphone number. Or you want you have an assistant who needs access to all your voicemails, all your messages, right? You want them to be able to maybe respond as you.
10:51 Speaker 00 We provide all that software for just basic advanced business phone system and messaging as well. So those are kind of the three things. And then on the on the chatbot. All of those support, especially the cell and the support side, they support full, modern, like rag L L M based chatbot that we've built. We launched last fall or this past year. That does all of that. It sucks in your knowledge base. So you basically connect it to your knowledge base or upload any data you want. We scour the whole thing. And then when someone comes in and says, hey, how do I do this thing with my account? It automatically dips into your knowledge base. And it returns a beautiful, you know, human language version of the answer, right?
11:35 Speaker 00 And uses the L L M technology to parse the question and then also to construct the answer. And then does the whole, you know, vector database lookup and all this stuff behind the scenes to get the right answer. And then we built all this stuff around it to do curation. Cause like the biggest problem with enterprises have is like, they don't want to just unleash an L L M directly to their customers. They want, they want to make sure it says exactly what they wanted to say. And so we had to build a whole curation system around it that does things like, here's all the questions we've seen. Here's all the answers that we would provide, our technology would provide. Do you want them? Yes, no. Oh, I want to change the wording on this one, right? And so we have all of that and it can even take actions. And if they don't have the
12:16 Speaker 00 answer through the, we call it AI agent, or that's our product name. If they can't get it from the AI agent, they can easily from the same chat window escalate to a real person. And it has their whole customer and like customer journey already summarized for them. So it says like, oh, here's what they were asking the bot about, right? Here's what, what was, what they got, what they didn't get from it. So now when the agent, the real human takes over, they got it right there. It's already, and that's kind of our other strategy is we want to own the whole customer journey. Because you've all probably had bad experiences too, where like you start on like an online chat that is clearly a bot. And it doesn't help you and you escalate and then you kind of have to repeat yourself.
12:56 Craig S. Smith? Right.
12:56 Speaker 00 To the person you're now talking to or provide the same information again. A lot of times it's because it's a completely different system.
13:02 Craig S. Smith? Yeah.
13:02 Speaker 00 And that's what we're also trying to solve. The way you can get the whole picture of a customer is if you have all those endpoints. Like we even do video meetings. So if you're a salesperson, you can go from messaging someone to a call with someone. So now let's schedule a meeting. It's all in the same timeline. You can see everything you talked about over a meeting, over a call, over a message. We can summarize all of that. And that that's like our strategy, right? Like that's we think it should have been the way things work, right? That's kind of like the real unified communications.
13:30 Craig S. Smith? Yeah. The customer service end of it is is getting kind of crowded. It seems to me just based on the.
13:39 Speaker 00 Yeah, at this conference, as I can tell, yeah.
13:41 Craig S. Smith? Number of pitches I have from companies. You know, there's, you know, Sierra.
13:47 Speaker 00 Uh huh.
13:48 Craig S. Smith? These guys. And there's a new company, Crescendo, that was, you know, sort of has humans and and AI all under one roof as opposed to like having a BPO. And then your chatbot and trying to make them talk to each other. How do you stay ahead of the the pack?
14:11 Speaker 00 Yeah, there's so much competition. I think when when ChatGPT came out, I think every single smart person in the world said, wow, this would be amazing for customer support.
14:21 Craig S. Smith? Right.
14:21 Speaker 00 And so now, since that those large language models came out, I think it feels like half the startups are trying to build some kind of customer support by. The way we stay ahead is one, we have our own AI team. So we've been started, we started on this whenever before people did start on this. And we already had the technology of an AI team and inference and all these things you need to run AI at scale. So when this whole world of like large language models came out, we already had all the team to do it. We had the scale to do it. We had the data to do it. So we've already been cranking as fast as we can. So we one think we're ahead of those. But even then, the thing that differentiates a lot of us or us versus maybe some of these startups is anyone.
15:02 Speaker 00 Can go build a chatbot. You know, you can usually figure it out. There's enough stuff out there and but it's really hard to build like a global real time communication platform. And I mean, not just chatbot, like to have a full contact center that what businesses expect is they need all of it in one. They need the ability to like, yeah, go online and talk to a bot on a website. But then, like we just talked about escalate to a real person. Or transfer to a real person or add someone to a call or maybe switch to screen share. We have a built in where you can call into a call center and they get a link and you can see the screen share of the agent. So the agent could show you what they mean when they're, you know, like, that's the type of stuff that you kind of have it to have an all in one.
15:44 Speaker 00 And it's really hard to build. Like that. That was our background with Google Voice is that you have to build a worldwide platform. We're in over fifty countries. Pretty much day one, because we are also telephony. And that is expected. You have to work with the phone number. And if you can't, like, you're missing a huge piece of that puzzle. And it's really hard to build that piece.
16:06 Craig S. Smith? Yeah.
16:06 Speaker 00 So we, we, we've been spending our lives building it. And it's not something you could just like learn in your college class at Stanford and they don't teach it.
16:16 Craig S. Smith? Yeah.
16:16 Speaker 00 They teach mobile apps. I can build an iPhone app, but they don't teach how to build like a worldwide, real time, reliable communication network. We have to work with carriers around the world and have partnerships. We, we have to do all of this in house. And that is so hard to build. That that gives us all those advantages of customer journey. It gives us the data advantage. These AI startups don't have the data. They're buying conversational data online. That's not really relevant. So it's kind of that whole pack together package, right?
16:46 Craig S. Smith? Yeah, the, the voice or, or, uh, voice data, I mean, the captured conversations and that analysis. You said that voice is is the last online.
17:01 Speaker 00 Offline.
17:01 Craig S. Smith? Offline, I'm sorry, data.
17:03 Speaker 00 Data set, yeah.
17:04 Craig S. Smith? So, you know, it seems, do you do other things beyond customer calls? I mean, you mentioned, you know, teleconference, like group meetings and things like that, even live meetings. And there are a lot of tools now that transcribe those meetings and they're hooked up to a chat bot. Then you can query the the meeting. But it seems for an enterprise, that would be a pretty powerful tool. If you could capture all the conversations that go on in meetings with, you know,
17:43 Craig S. Smith? with the knowledge of the people attending and mine that for insights.
17:50 Speaker 00 That's how we use, we use, obviously, we use our own product internally. So all of our meetings internal are with dialpad meetings. The, I would say, to be honest, so we have lots of companies who use dialpad meetings instead of any of the, you know, other competitors for internal and external meetings. Most of our customers are using our meetings product as a, as a seller or a support agent with a customer.
18:17 Craig S. Smith? Yeah.
18:17 Speaker 00 And because the thing about our platform is like you, again, I just told you a little story about like, hey, I first start texting you, I call you on an outbound call, then I set up a meeting with you. Having all that data and conversation in one thing, so they use us for the meeting with the customer, even though they might use something else internally for their own internal meetings. But that gives them all the other features we've built, like live coaching. Like one of the features we have is called playbooks, where to like tell you, here's all the things you need to figure out on this call, right? You need to, you need to find out their budget. You need, you know, this like kind of typical sales playbook stuff, right? But you can customize it with us, but you can say, I want you to first initial call, find out their budget, find out their use cases, right?
18:59 Speaker 00 And it knows when you ask the question, it tells you to ask it, it knows when you asked it, and then it does a little summary of their answer, right? And so that now is all combined with even your call you had with them a week earlier. It all goes to your CRM, it goes into our system too, but it can go to your CRM. And now you have the full picture.
19:18 Craig S. Smith? Yeah.
19:18 Speaker 00 And then we can do even cool things like, you know, some of the stuff we do is like a churn prediction. You know, for support side, same thing like they might, one, we can analyze that, but getting back to your original thing. Support people, so you call into a, you know, and have an issue. They can hit a button that gives you a unique link to join our meeting product. Where now you can be face to face, which increases satisfaction. They can share their screen, you can share your screen. We even have like remote desktop control, so they could take over your screen if they want to like do the thing for you. And again, it keeps that conversation together. So you just switched the video from a regular phone call, but it's keeping all that data in one place. So like, that's where, that's kind of like where, why we have so heavily invested in video.
20:02 Speaker 00 Is because future, it's customer relations is going to be a lot of that too. You kind of just need it all, right? So we don't want to miss anything.
20:08 Craig S. Smith? Yeah. Yeah. The, is there a consumer side to this?
20:13 Speaker 00 We have a lot of consumers who buy dial pad. That's not what we're going after. There's a lot of, you know, one person, two person, you know, businesses using us. My parents are using it for their home phone system. I'm using it with my wife, but it's, we don't stop anyone from signing up, but it's, it's pretty much mostly, yeah, for B to B.
20:33 Craig S. Smith? Yeah. And is it, you don't stop people from signing up? Is it prohibitive for, for an individual?
20:40 Speaker 00 No. So one of the first things we did, and this was our Google Voice background. I think we were the first, the first product where you could sign up for a phone number online and get it instantly. Was Google Voice back in like 2007 or eight.
20:54 Craig S. Smith? Yeah.
20:55 Speaker 00 And so we took that and we said, most of our competition was contact sales.
20:58 Craig S. Smith? Right.
20:59 Speaker 00 So one of the first things we did was like, no, you're going to be able to do everything and sign up online. You're going to be able to go and pick a number, search numbers, do everything, pay and get a phone calls or get messaging instantly. So we always prioritize that. So that's why we do have a lot of just people who come to our website. And sign up, even if they're not a business, we don't, we don't enforce that.
21:18 Craig S. Smith? And then how do you charge? Is it pay as you go?
21:21 Speaker 00 It's per seat. So most of the stuff is per seat. It's if you're just using our regular dial pad. Connect it's, you know, that one's the cheaper one and that's more for just regular. I need a separate number and more controls for my business line. And then it goes into like the contact center is more expensive. Because it has all these other things and call routing and skills based routing and all the extra AI that goes on it. But even our base one has AI and transcription and summaries. So yeah, it's and then our contact center has a very small per minute charge depending on international calls, toll free calls, things like that.
21:58 Craig S. Smith? Yeah. Why is it taking from a consumer's point of view such a long time to for this to penetrate the market? Cause I still, you know, get that chat bot that doesn't do any good. And, you know, I gotta wait and then it's somebody at a call center in India. And they don't really ask me the same questions I've already answered. Why is it taking such a long time for enterprises to adopt these technologies?
22:32 Speaker 00 Well, they haven't, they haven't found out about dial pad yet. So that's, that's part of it. You know, we, we announced we're over 300 million in revenue and just last year. So we're, we're getting there, but we have a long way to go. It's a massive market. The, the main reason is one, you to do what I'm telling you, you have to have the whole thing.
22:49 Craig S. Smith? Right.
22:50 Speaker 00 You can't just be an AI agent bot. You have to have the whole thing to have that amazing customer experience and not really anyone's built it yet. Because they either were traditional legacy call center where they don't have AI. Or they don't, they're, they're catching up last minute to AI, right? It's not their like specialty or you're an AI startup who's building an amazing bot, but then you're asking to integrate. You're like trying to reach, but integration is even hard and it's expensive because they have to make their margins. And a lot of times they're using a third party for their AI. So it becomes really expensive and it's, and it's also just so new. Like if you think about like with these, the chat GPT, that was not that long ago. That really was, those large language models were, were a real breakthrough.
23:34 Speaker 00 And it wasn't really possible until then, but now. People were playing with them, but to do all this stuff to make it enterprise ready is a lot of work. Like you can't just release it to the world. And I think there was a company that said they replaced like ninety percent of their support agents with with their own homegrown chat G P T. You know, bot, and they just pulled that back and said it wasn't working. You know, like it's because you can't just unleash this amazing technology and give it directly to the consumer. It'll say crazy things.
24:04 Craig S. Smith? Right.
24:05 Speaker 00 And you need all this extra stuff around it and you need compliance.
24:09 Craig S. Smith? Yeah.
24:09 Speaker 00 It's it's just it's not easy to then get it out the door. And it's still only been a couple of years since people have been able to really, really have it available. Because even before then, it was not available, like G P U's weren't available. If we wanted to use Google or OpenAI, they didn't have enough bandwidth. So like even if we wanted to roll out this amazing chatbot, so did everyone else in the world. You couldn't, you couldn't scale it. And so we then immediately said, we can't wait on them. So that's why we started even having our own, our own A I's running on our own machines. And in some cases, even down to the C P U because we needed availability.
24:43 Craig S. Smith? Yeah.
24:43 Speaker 00 You know, we, we just, one thing like we had, um, I put this on LinkedIn. I was saying, I think we have the world record for generative A I summaries of calls and meetings. And it's only because we're able to scale it. And because we built it ourselves, it's cheap.
24:58 Craig S. Smith? Yeah.
24:58 Speaker 00 Whereas if you were using someone else, it'd be so much money.
25:01 Craig S. Smith? Yeah.
25:01 Speaker 00 And so it's, it's just kind of the luck of it's so new, that's why.
25:04 Craig S. Smith? Yeah.
25:05 Speaker 00 But it's coming.
25:06 Craig S. Smith? Yeah, everyone's talking now, I mean, enterprise first has to adopt this, but everyone's talking about these multi-agent systems. Are you guys exploring that? Are you integrating that into your stack? And what would the agents do if you are working on that?
25:29 Speaker 00 Yeah. What do you mean by exactly like, what do you mean by multi-agent?
25:32 Craig S. Smith? Well, agents that can take action, so I'm guessing.
25:36 Speaker 00 Oh, yeah, yeah.
25:36 Craig S. Smith? Yeah, you know.
25:39 Speaker 00 It's like the agentic.
25:40 Craig S. Smith? Right.
25:41 Speaker 00 AI, yeah, like the things that actually can do things.
25:43 Craig S. Smith? Yeah, I'm sorry, because call center agents.
25:46 Speaker 00 Yeah, yeah, but that also doesn't help us that like now.
25:49 Craig S. Smith? Yeah.
25:49 Speaker 00 Everything's called an agent, even if it's not a call, a call center agent. Yeah, so that's, that's stuff that's already, we already do some of that. And we're doing even more advanced stuff with that, but that's part of what our bot can do, our AI agent. Is you can say, hey, I want to schedule a meeting. Right? And it'd be like, oh, let me go reach into your scheduling system and say, what, what, what times are available? And then it'll bring back, okay, we have these times and they can click it and then it writes it back to that same system. So we already have, that's what a lot of these like workflows. So before agents, it was like kind of called workflows, which is like that if this, then that flow. And you need a lot of that because you need to control the inputs and outputs again.
26:32 Speaker 00 You don't want it to just, again, try to do the whole thing. You want to say only if they ask this, can they then do this? So a lot of permissioning around it as well. Yeah, we already have that, but we're starting to do even more. So like things like we have the concept of holiday hours, like setting your, your business hours. And that would have been really hard for you to go in and, you know, find it in the interface of a thousand different things and set it. Well, from our own assistant inside of our web settings, you can say, hey, I want to change my holiday hours. It says, okay, what did you like to change it to? Okay, I want, I'm open on weekdays this time and weekends we're closed. And it's like, okay, we're going to respond with, is this correct? You want these changes? Yes. And it changes it. So that is, I think that is the biggest trend you're going to see this year is all this stuff that it's taking now.
27:17 Speaker 00 This amazing understanding of the language and taking action with it. That's why you're hearing about it all over the place.
27:23 Craig S. Smith? Yeah.
27:23 Speaker 00 The agentic explosion is, is happening.
27:26 Craig S. Smith? Yeah, yeah.
27:27 Speaker 00 Yeah, we're right there.
27:29 Craig S. Smith? And, and for you guys, uh, I mean, it's, you have a, a strong business, but there's kind of an endless market. I mean, a lot of people chasing it, but you're, you're in the game. So what's, what's the future? Are you, are you?
27:48 Craig S. Smith? Are you just working on building market share or developing new products and or new directions? And if so, what are those directions? I guess agents or AI agents is one of them.
28:04 Speaker 00 I mean, our focus is on customer communication and whatever is going to make that better or more efficient. And it's it's it's a combination of all of it. I'm not I'm not a big believer that it's going to take over the world and you're not going to need, you know, a support agent anytime soon. And even if it does, it'll be able to do some really like common tasks for you, but you'll always need an escalation path. You always need to be able to then get to like have a complicated discussion with a real person. So, one, yes, all that automation. We just launched last year, our next generation AI agent, which is using all the latest technology in our own large language model. Then the next phase is voice. So taking the same platform, which is all the same stuff, right? If they ask
28:47 Speaker 00 this or want this, do the same thing, except in text form, you're going to do it in voice form, right? And then if they have a problem, you can switch to a real person and they can talk to you. Then there's all this stuff we can do with the data, though. So like, we do a lot of live coaching. And I told you, like, the playbooks and like, we even do like, automatically score a customer conversation that from the agent's perspective, like, is the how good is the agent at doing this? Did they welcome them? Did they, you know? And so we have like QA scorecards, we call it with AI, where it automatically grades the call. So again, you can train your agents. So there's a lot of just enhancing humans. It's like, I still believe is a lot of what we're doing. I don't just like even and I'm I'm a programmer, like, I don't think it's going to replace me, but it's making
29:30 Speaker 00 me faster, like so much faster. And it just means we're going to do more. And then, yeah, the next part is we still want to automate. I think there is a lot of automation. And but the data part, sorry, the data part is with all this amazing conversation.
29:45 Speaker 00 You should be able to, there's so much gold in there.
29:47 Craig S. Smith? Yeah.
29:47 Speaker 00 Like, which customers are happy? Which ones are unhappy? We'll be able to tell you, which we're working on right now, we'll be able to tell you which customers are likely to churn, based on the conversations they've had. We've found already that you can tie those conversations to, when you connect it to like, you know, their status, or if they churn, we can learn off of that. And we'd say, hey, this is what we should be doing. And now here's what you should be saying to avoid or cross-sell, or how do you counter if they're upset so that they become a loyal customer? And so it turns out, we're not trying to save like, you're, you're making so much money off of how much happier customers are, or your sales is doing so much better because of all this information you're now able to provide. Because you don't have to repeat yourself. Because you do know what your sales team talked to them about two months ago.
30:32 Speaker 00 So, and analyzing all that data, that's, that's like the future is like, there's this all this data now. We gotta take it, we gotta give you insights.
30:40 Craig S. Smith? Yeah, yeah. Is there going to be a point at which AI, uh, voice and, and
30:51 Craig S. Smith? large language models and agents are, are seamless or strong enough that, uh, they'll do a lot of the customer interaction?
31:01 Speaker 00 Yeah, I, I, I think there's gonna be a lot of simple stuff and I, there's some people even talked about it here.
31:07 Speaker 00 If you get an answer fast for your problem, or your problem's fixed, you don't care if it's an AI or not.
31:12 Craig S. Smith? Right.
31:13 Speaker 00 It's whatever gets it done. You don't want to be on hold for 20 minutes and then another 20 minute conversation just to do some simple thing. If, if, if a AI agent can do it for you, that's gonna make the customer happy. And I, and if you look at a lot of the interactions, especially in customer support, 80% probably are like, you know, are simple things that a bot could do.
31:34 Craig S. Smith? Yeah.
31:35 Speaker 00 And then it's a win-win for everyone. Your best people are solving real hard problems, instead of doing like tedious same thing over and over again. And the customer's happy because it's, they're getting their answer faster.
31:46 Craig S. Smith? Yeah. Uh, are, are you, I imagine that you track research and what's happening in the broader world of AI.
31:55 Speaker 00 Yeah, I'm obsessed with it, yeah.
31:56 Craig S. Smith? Yeah.
31:58 Craig S. Smith? What do you, uh, do you have any, where do you think the most exciting areas are?
32:06 Craig S. Smith? I mean, I'm sort of obsessed right now with Manus, this Chinese. Which I don't have access to yet, but I've watched plenty of videos of people that do have access and it's it's amazing. I mean, it's not something that I think
32:24 Craig S. Smith? companies like OpenAI or Anthropic couldn't do themselves.
32:28 Speaker 00 Yeah.
32:28 Craig S. Smith? But, but I don't know, but what's your, your, is, and is that the most interesting? Part of what's going on or, you know, reinforcement learning, Rich, something?
32:40 Speaker 00 I think that made the, that made the headlines. It seems like it's a lot of, people don't really know enough about it because it's not, they don't plot it, they don't have access, we haven't done run tests on it. So that one's more, in theory, it sounds amazing, but when you look at it, at least just from demos, it looks like it's just a lot of heavy compute and using the same technology, but like reinforcement stuff, like continuing to re-prompt and re-prompt and re-prompt and doing it faster. So it's hard to know like if it's like a breakthrough technology or if they just wrote some amazing things around the traditional models. And I, we don't know yet, like until we get more information. I, I think
33:17 Speaker 00 we're still at the beginning of this breakthrough. And that's why, like, the future is taking, taking all this stuff and now automating as much as we can. But I don't think it's going to be auto magical. I don't think it's like, just take this thing and connect with it and it's just going to do it. I think there's going to be a lot of vertical where you have to go in and build, you know, specific bots for legal, specific things for automating your, whatever workflow you have in your company, but it's going to be very specific to you. I don't think you're just going to be able to take this, this new amazing thing and say, go do this. I don't think it's going to work. It's going to need, and even then there's privacy stuff and like, they have to have access. That's the problem. Like for all these agents, you can tell it all you want, but it can't log in for you. Like,
34:02 Speaker 00 I want it to solve like, hey, I want to book a vacation, go find the best place, do reviews, look at the reviews, book me the cheapest flight. But like, it doesn't have access to do that. Like, the whole internet's blocked bots. And so like the next phase you Probably heard about too is that MCP protocol. Where it's ba- it's basically a way to connect your, these, all these things that can do actions to the L O Ms. And so you're gonna see a lot of like, that's enabling the agents as everyone builds this interface layer. So that the L O Ms can talk to their product to do things. That's like the new way of A P Is.
34:36 Craig S. Smith? Oh, that's interesting.
34:37 Speaker 00 Yeah, and that's like the next big thing, like Anthropic built this and it's, it's, yeah, kind of the next big thing.
34:41 Craig S. Smith? So, so a website, like an airline website would, you, you have to log into your account. But if you're an agent that's been vetted or something, it gets access.
34:56 Speaker 00 Yeah.
34:56 Craig S. Smith? Is that what you mean?
34:57 Speaker 00 Yeah, so you would, you would go in theory, and this is all relatively new.
35:00 Craig S. Smith? Yeah.
35:01 Speaker 00 You would go to that, you would go to the, so Claude Anthropic's Claude has support for this and OpenAI, I think, will have it if not. But it's similar. You'd basically go and say, I want to add this airline to my chat G P T. As an example, and then you'd authenticate. So you'd then quickly authenticate that, oh, it's me. And so now, when you say book a flight, it has access as you to now go do things with that airline. That's the next wave. But the problem is those businesses still have to enable that. They can't just crawl the web. There are some stuff to crawl the web, but it's still not perfect. That's like really, it's all over, is this MCP, uh, Model Context Protocol, is I believe what it's called.
35:46 Speaker 00 But you'll hear a lot about this, I think, in the next, yeah.
35:49 Craig S. Smith? Yeah, okay, M, M P C.
35:53 Speaker 00 M C P.
35:54 Craig S. Smith? M C P.
35:55 Speaker 00 Yeah.
35:56 Craig S. Smith? A little dyslexic.
35:57 Speaker 00 Yeah, if you just Google it, you'll see a million things now about it. But it's, it is, this, this year is about agents, like doing things for you.
36:04 Craig S. Smith? Right.
36:04 Speaker 00 And there, a lot of the enabling of that is with that, that technology.
36:08 Craig S. Smith? Yeah.
36:08 Speaker 00 But that's what we're doing too, like, because you, again, makes sense for customer support, for sales. You want to solve the simple things, like I just need to change my billing address.
36:16 Craig S. Smith? Right.
36:16 Speaker 00 Like little things that it should be able to handle. But yeah, privacy, security, like authentication is a big part of like the roadblocks to that.
36:23 Craig S. Smith? Yeah. Okay, this has been fascinating for me. I hope it's fascinating for.
36:28 Speaker 00 I love talking about this stuff.
36:29 Craig S. Smith? For users or for listeners. Uh, yeah, and I'll, I'll log on to dial that, maybe I'll become a user and individually.
36:38 Speaker 00 Awesome. Yeah, yeah. Awesome. Well, thanks for having me.
36:42 Craig S. Smith? Yeah, great. Thanks.
Previous #257 Ankur Banerjee: How cheqd.io Is Building Trust for AI Agents with Decentralized Identity Next #259 Anjney Midha: a16z's Strategy to Turn AI Startups into Unicorns Page updated November 3, 2026 · transcript and summary are machine-generated
