Are Apps Going to Get Axed by Conversational AI? with Jacques Klick, Director of Market Strategy at Sinch

(Intro Narrator at 00:00:00) Today, we're talking to Jacques Klick, Director of Market Strategy at Sinch, about the rise of conversational AI and its long-term impact on applications. You're listening to Joel Beasley, Modern CTO.

(Joel Beasley at 00:00:19) When I was preparing for this, we have the same mindset, actually. So this concept of our apps gonna get killed by conversational AI. I think we're getting rid of apps, and it'll almost be like, if I want to say, "Hey, show me what's going on in life in a feed format from these three sources," my screen will just show me that.

(Jacques Klick at 00:00:41) Yeah. I think we're heading to sort of that singularity, if you will, or that single inbox. For me, I mean, the obvious one is we're already seeing this take shape. Right? You look at what Salesforce is doing with their headless UI. That's an example of a huge business application, a CRM, sort of making the decision of, "We're actually gonna be that invisible layer that connects into where our customers are." And I think the genesis of that is when we saw this explosion of user adoption for the LLM tools. Everyone started to go into OpenAI and Anthropic to do their workflows, to do their daily life activities. And then because of the rise of MCP connections and now native integrations into those applications, it's just the user interface that people like to take part in. And you hit on the conversational thing. That's exactly, I think, where I spend a lot of my time and my company's time figuring out because we're a communications company. Right? We're a global communications platform. We connect to email, voice, messaging platforms, and channels, and we've been sort of silently watching the rise of conversational AI capabilities and saying, "Wow, this is really gonna change the way that people interact because instead of having to click through 10 different companies' apps to do a certain task, I can just go into one interface, my text thread, or maybe it's an LLM tool, and then just prompt it with what I'm looking to do and get all that back." So you're exactly right. I mean, I think that's what's underlying all this. And now the agentic stuff is taking that even further. Right? Because now it's maybe that you had to do some connection into that third-party application, but now the agent is doing that for the LLM or for the messaging channel itself.

(Joel Beasley at 00:02:26) How long do you think, like, what do you think, how will it roll out? Will we be there in like five years, 10 years? Like, I'm pretty sure we'll be there in 10 years. But I think we're gonna get close.

(Jacques Klick at 00:02:38) Yeah. I think sooner. Yeah. I could see this in two to five years. I mean, like I said, when you see big companies like Salesforce, I mean, the company I used to work for, ZoomInfo, they announced the same thing. I think people recognize that in order to survive in this space, you need to become integrated where the user base is converging. And that's clearly these LLM tools. The massive user adoption they've gotten is crazy. And so I think it's happening very, very quickly. And even from an enterprise software perspective, look at the adoption of something like MCP, like a super technical—it's just a connector at the end of the day, but everyone ran when it first got rolled out by Anthropic. Everyone ran and sent their teams to go build MCP connections to make sure that they could be discoverable in those LLM tools by their customer base.

(Joel Beasley at 00:03:31) Absolutely. I mean, that's like, it's the new internet. Like, you have to be there. And ChatGPT just rolled out ads too, so I wanted to check that out. I had that on my list of things to do.

(Jacques Klick at 00:03:42) And just generally, like, I don't think you're finding people competing for App Store relevance. Right? Like, 10 years ago, that's what the big conversation was. How do I get my app ranked number one on the App Store? That conversation has completely gone away.

(Joel Beasley at 00:03:58) Yeah. You're exactly right. And, honestly, I was real concerned, obviously, about like the deepfakes, the ability to generate video, and all of that. And I feel better about it now, or at least I feel like I understand where the curve is headed. But one of the things that I didn't expect was the business that I've gotten from it. So because the podcast has been around for so long and because so many people have talked about it for the past 10 years, now when people will ask, "How do I reach these technology leaders," the marketing people, the Gemini will tell them. So I started noticing in my incoming leads and stuff that people are like, "Oh, yeah. Gemini or Grok told me about you." And I was like, "Oh, this is actually fantastic."

(Jacques Klick at 00:04:39) Yeah. What are they calling it, AEO now? Right? I mean, that's the whole thing. Is how do you become discoverable? And it's something that our company—so I work for a company called Sinch—we spend a lot of time thinking about because we're doing it for the humans to discover via the LLM tools right now, but I think just around the corner, it's the agents themselves. Right? So when someone just prompts, "I want to send a marketing campaign from my website. I have a house cleaning business or something like that," that the LLM tool itself goes and does that search. Well, who's the best communications platform to do that type of campaign that Sinch, my company, is the one that it's gonna pick? And so it's changing the game because the old ways of SEO was, "Alright, you know, I'm gonna compete on key terms and things like that," but now you're competing on agent capabilities. So the ability for the agent to easily connect into your system. How much time does it take for a developer or an agent to connect into your platform and already get to the build it needs to do? How much documentation do you have for that agent to truly understand how to correctly build that thing? That's a lot of the work that my company is spending time working on, and I think we all are because we recognize that's where this is kind of going.

(Joel Beasley at 00:05:54) So when you're building your docs and you're building your content, you're also thinking about how it will be ingested by the LLMs.

(Jacques Klick at 00:06:01) A hundred percent. Yeah. There's different frameworks you have to think about and different ways that you're setting it up. In some ways, though, it's not too different than setting your docs up for a developer. Right? Because a lot of these IDEs and coding assistant platforms, they do operate just like a superpower developer, and they're using the same frameworks, and they're using the same vocabulary and things like that. But I think it's more about just simplifying it and making it accessible. And in some cases, for us, where we're trying to compete right now is that native integration into those platforms. So IDEs like Cursor, Windsurf, and then the Claude tools and OpenAI and things like that.

(Joel Beasley at 00:06:39) Oh, explain that to me. So I'm a user of Cursor. How do you mean a native integration?

(Jacques Klick at 00:06:44) Well, something it could just be, like, skills, right, that we have packaged up and accessible within Cursor and Windsurf. So, you know, our company, Sinch, we're a communications platform, so we have expertise in different communications channels. So messaging, email, voice. And so when a developer who's using Cursor or Windsurf and is writing code and they need to do a certain communications task, all of a sudden, because they're saying they want to do a particular type of campaign—let's say it's a just a regular SMS campaign to notify customers of a change that's happening on their system—well, it's gonna download a Sinch skill because it has that native connection, and the skill is gonna tell them exactly, you know, here's how SMS messages get sent, here are the dimensions, here's the character count that you need to take into consideration, here's what your link needs to look like if you're gonna link something, and all of that just gets automatically ingested by the bot and it starts writing the code. And, again, so I think as companies think about how they want to work with these LLM platforms, you really have to think about, like, what is the domain expertise that you can bring to that platform to actually execute something really, really strongly.

(Joel Beasley at 00:07:54) Yeah. I'll move on from this, but I just personally am curious. Alright. So I haven't come across this. So if I'm using—I use Cursor. And so, like, I'm using it, and I'm building an application. Is it doing that in the background without me knowing, like, just trying to solve its own problem? Or do I actually say I need to enable this skill, or does it suggest—I haven't come across this, like, skill approval thing or whatever.

(Jacques Klick at 00:08:17) It can do both. I think it can certainly run in the background for certain skills. But usually, on the right-hand side of your screen, you'll actually start to see it say, like—and sometimes it's through the Claude code, like, kind of connector that you have to the IDE—but you'll actually start to see, like, "Hey, there's this Sinch skill. It can do x, y, z, it can send an RCS campaign, it can do an email," and then it will download that skill and just start executing. It's very similar, honestly, to if you ever use Claude, right, and now you're using it in the chat instance or Co-work, and you ask it to produce a document in a Word doc or a PowerPoint, it automatically goes and downloads the skill that it needs to do that. So, I mean, honestly, a lot of companies are competing on that skill and creating custom skills now.

(Joel Beasley at 00:09:01) Oh, wow. And then me as the developer, am I alerted to the fact that it's using that skill or something?

(Jacques Klick at 00:09:08) Yeah. Yeah.

(Joel Beasley at 00:09:09) And this is happening today? This is happening today?

(Jacques Klick at 00:09:11) This is happening. This is live in production today. You know, the technical capability there, right, is, like, you might want to be a Sinch customer. Right? So in order to have that skill actually execute, you'll need to actually call our APIs. And so at that point, you'll need to have some sort of API key that you put into the software itself. So that for us, that's the front door that we use so that people are not just—

(Joel Beasley at 00:09:36) Oh, okay. So, so if I'm a Sinch customer and I've got the RubyGem or whatever the package is installed, Cursor has that context that I'm using this API actively. And when it goes to do things that could use that API on my project, if there is a skill there, it'll ingest that and then know how to use it better. And if not, it's just gonna try the best it can on endpoints.

(Jacques Klick at 00:09:58) That's it.

(Joel Beasley at 00:10:00) Uh, thank you for helping me understand. Sorry for everyone who had to listen to me wrap my mind around this new concept, but I'm sure if it's new for me, it's probably new for a lot of people.

(Jacques Klick at 00:10:08) Yeah. It's really cool. And it's just something that, again, we're very AI-forward at our company and thinking about what's around the corner and how do we just make ourselves easier. And as a communications company too, specifically, but I think the rest of enterprises are gonna have to adapt to this model as well. We've always wanted to be the invisible infrastructure behind the scenes. Because it's not a great experience if one of our customers is trying to send a message. Let's say you're a large bank, Chase Bank, and you're sending a one-time passcode to one of your customers, and it says it's coming from Chase Bank, but delivered by Sinch, and it's coming from a number that's 665224. That's not a really trustworthy experience. Right? So we've always had to be that invisible layer. And now what we're finding out is that I think a lot of enterprises realize, "Oh, we're ending—whether we like it or not, we're ending up being that invisible layer through these things like skills and automatic connections into the LLMs." I mean, in some cases, they are telling you, like, "Hey, it's coming from this type of software or platform," but in many cases, we're all just becoming a little more invisible.

(Joel Beasley at 00:11:18) And you know what? You know who called this, by the way? Sir Tim Berners-Lee, the creator of the World Wide Web. So I did an interview with him, like, five years ago. I was so excited. And he just like, you know, when you're talking to someone who's brilliant or like, "Okay, I just, I feel like a monkey right now in this guy. Yeah. I'm just trying to, like, I'm just trying to keep up in the conversation." And he was explaining to me about the second version of the internet and how it'll work, and he said this was before the LLMs got popular because it was five years ago. He said, "What's going to happen is, like, right now, the interface and the data store are, like, married. So, like, you go to Facebook, and it's using Facebook's data store. You go to Chase Bank, it's using Chase Bank's data store." He goes, "What will happen is that interface and database will come to your local device, and then that will sync. And so, essentially, Facebook will ultimately—it'll still have its interface. But for 99.9% of users, it's going to just be an API. And everything's just gonna become this API that, like, you can perform actions locally in your own context, in your own applications, and then that data store will authentically sync with the other one." And I thought that's such a wild like, "Why would we have the need for this?" Okay. I can see it, like, app—and I was thinking of it in terms of apps because I didn't have the conversational AI. No one did, really, to the degree it is today in their head. And I was like, "I guess we could build apps like that. Like, that would kind of make sense so I could—" But now, dude, I built an app start to finish in 90 minutes, and it's on my iPhone running. I did it Sunday. I did it this past—I woke up. I had the idea. I was like, "I want this utility app." And then I built a basic version of it, deployed it to my phone. You know how many lines of code I wrote, Jacques? Correct? Zero lines of code. And, uh, and this is happening. I did an interview with this guy named Trevor from Soundstripe. He's in my town in Nashville, Tennessee. And he said, "For the past quarter, we have a new rule. No humans write code. Zero humans have written code. We review code, but, like, we don't write code." They use swarms of agents and the software called Conductor that I had never heard of. I don't know if you've heard of it, but it's—they're using these new softwares to, like, give outcome-driven explanations, and then multiple agents work together to achieve that outcome and come back to you when it's done.

(Jacques Klick at 00:13:43) And that's the most exciting part about AI. I think all of us relate to that moment where we built something like you did and we wrote no code. Like, for me, it was I used Lovable last year, which is a Swedish-based app builder company. And I was testing out running my own small business, and I created a website. And I had a pricing model that was in a rough Excel file. All I did was upload that pricing model and say, "Imagine I wanted a home screen on my web page where someone could pick, you know, certain classifications of what type of job they needed to book. And based on my Excel sheet, you could create, like, a custom price quote directly on the home page," and it did it, like, immediately. Right? But I think the second part of your point is that that's the reality. Like, yes, people are rushing to do this. And we just did a huge study at Sinch where we talked to 2,500 enterprise leaders, specifically in charge of their communication strategies, and talked to them about their AI usage. And it was surprising because to us, going into that research, we had had the expectation that, "Oh, probably a lot of them are still piloting things." Because at the end of the day, like, for our customer base, they're interacting with their end customers. So trusting an agent to handle that interaction, you know, a conversation, a notification about a flight check-in, a bank password, an update on your account, pharmacy pickup notification. That's a really tricky thing that you don't want to get wrong and you don't want to risk your brand reputation.

(Jacques Klick at 00:15:18) But it was surprising that 62% of them already had agents live in production. We're talking autonomous agents out there using conversational tools and interacting with their customers. And now what we're finding out is that 74% of those had to roll back the agent because of different failure points. And we can get into those different failure points, but I think that's the important piece here. You're right, everyone is rushing to these outcome-based, "Oh my god, I can repurpose a whole team and swarm them on this thing, and it can go do it autonomously." But then the governance piece really starts to show up, and that's what we're seeing when we talk to our users and our customers.

(Joel Beasley at 00:15:59) You're exactly right too, because one of the things that I see a lot of people make mistakes with is they'll hear an interview like this, they'll maybe hear something that I said about Trevor, and then they'll look to go try to apply it in their customer service center. But I think there are—I know that there are different niches of the market that are more mature than others on different types of tasks, right?

(Jacques Klick at 00:16:21) Yeah.

(Joel Beasley at 00:16:21) So there's some very basic tasks that are highly automatable, and you hear those people talk proudly about it and how they cut their cost and did this. And then you hear this other person who's in an area where it's not mature at all. It's barely making something happen, and they're like, "Yeah, we heard them do it. We're doing it, same technology." And then they're having to roll back. You think any of that's happening?

(Jacques Klick at 00:16:41) Yeah, I think a ton of that's happening. If I think back to when I speak to—or when I just scroll LinkedIn, right? How many times do you see an influencer who says, "Follow me and like this video, and I'll show you how to build this end-to-end thing." And then you actually try and implement what they did, when you try and use their packages and whatnot, and it all breaks down. Because to me, the thing about agents is it's all reliant on the connectivity across your different products and your different tools and things like that. These agents aren't yet able to fully move from a CRM system to messaging channels to all these different things. We're still—each company is figuring out those different connection points, and each hop that the agent needs to make to a different tool is a risk of another breakdown or a risk that it's going to interact with another agent. So I think it's pretty scary, and I think what you're hitting on is the approach that I think we give to a lot of our customers, which is start with the basic tasks. Start automating those things that are highly repeatable but very uncomplicated. Get that done. Create audit trails, because that's one of the biggest failure points we actually saw with the research that we did, is that a lot of people aren't even able to track where their agents are going and what they're doing. And so the question that we ask our customers all the time is, if you had an agent running and it made a mistake, would you even know about it? If it's off sending messages to your customer base, how are you even able to measure it or know that it's doing that so that you can roll it back if it's, for example, sending out personal information that it shouldn't be, or making mistakes and hallucinating like we know that these models will always do. That is how they're built.

(Joel Beasley at 00:18:37) Rightfully so. It's a feature of intelligence. Humans do it all the time. If you build intelligence—

(Jacques Klick at 00:18:45) Very often. We're doing it right now here today.

(Joel Beasley at 00:18:50) But you don't build audit trail tools, though. Or you're building them inside of your—let's just do some work here for a second and just tell people, what's the exact problem that they're experiencing today that your company solves?

(Jacques Klick at 00:19:03) Yeah. So, again, let's start from scratch. Sinch is a global communications platform. We handle about a trillion interactions every single year across messaging, email, and voice channels. And so messaging could be things like SMS. There's also some new advanced channels that I'd love to touch on. Voice is seeing this renaissance revolution where, I think, particularly with wearables and autonomous vehicles, the rise of voice capabilities and it handling voice calls and contact centers like you mentioned is really impressive. Email is still one of the top used and most influential channels, right? Email's not going to go away despite all the rise of this other stuff. There's still a use case for email interactions. And the thing that we help solve for customers, right, is unifying those channels, number one, because what we find is a lot of times companies are spending so much time just trying to operationalize a customer journey where a conversation is happening, support ticket gets logged or a web chat comes in through their website, they try and reach out to the customer via the phone call, they have that transcript, but it doesn't connect when the customer suddenly wants to talk over WhatsApp or suddenly they send an email follow-up. So the first thing that we really help solve is unify that customer journey across all those channels and provide the kind of interaction allotment of, "Hey, this is what this customer is asking for and what their issue was and how we help solve it." The second thing is just the pure infrastructure and scale that we're able to offer, right? So if you think about why would someone even need to use a company like Sinch in the first place? The reality is without companies like us, if you needed to send a marketing campaign to every mobile device in the US, you would have to negotiate that with T-Mobile, AT&T, Verizon. But instead, we handle all that back end so that you just say, "Hey, I need to reach every device on planet Earth," and with our platform you can. "And I need to send hundreds of thousands, millions of emails or text messages because Black Friday's coming up." Well, you need an infrastructure partner like us who has the experience and the scale and the connections to those networks to be able to make sure that those messages get delivered on time, reliably, and also, when you start to think of global scale, compliantly. So regulatory compliance in the EU is very different than the US. A lot of people are probably familiar with GDPR, but it limits the amount of marketing activity you can do and how you do it. And then I think the last piece is all about the different trust and safety protocols that we build within our platform. This is sort of what touches on the guardrails that I think you were hitting on. So PII protection—so personal information protection—capturing that at the network layer so that it's not even coming into our customer systems if, for instance, a bank is interacting with a customer and the customer, for some reason, decides to share their Social Security number when they shouldn't be doing that over the messaging channel or whatnot. Blocking that so that the bank doesn't even see it. It's not even getting ingested in their system. It reduces the risk. And then, yeah, all the cool stuff that we can actually do on the execution of those messages. So the interactive campaigns that we can help our customers build over messaging channels or over voice interactions and things like that. So it's all about infrastructure. It's about reach. It's scale, reliability. And I'm sure you've heard this from other experts in AI, but to us, the thing that resonates the most with customers today is trust and safety. They are investing more of their AI budget into that than in AI development itself, and that's what our research really came and showed.

(Joel Beasley at 00:22:41) I love it. Yeah, you guys, the website looks beautiful, by the way. It's sinch.com.

(Jacques Klick at 00:22:47) Yeah.

(Joel Beasley at 00:22:48) Is that what you do?

(Jacques Klick at 00:22:48) Our brand team.

(Joel Beasley at 00:22:50) Yeah, good job, brand team. Is that what you do? You push them to sinch.com. They click "Talk to an Expert" or "Try it." Can you self-service or no?

(Jacques Klick at 00:22:58) Yeah, you can self-service. I mean, we service—developers are one of our largest target markets. So we spend a lot of time making sure that people can just log in, start building with us right away, get a prototype into your system, and start sending messages within a day. That's how we like to operate.

(Joel Beasley at 00:23:17) That is awesome. You said new channels other than SMS. Are there just the social channels like Messenger and WhatsApp and stuff?

(Jacques Klick at 00:23:23) No, actually. I mean, WhatsApp is a good example. But even WhatsApp, I think you're going to start to see even more innovation hit that. But the big one that's coming to bear, it's sort of just like a Ferrari in a garage waiting to roar. It's called RCS. So that's Rich Communication Services. It's sort of the next evolution of SMS messaging. And if you think about it, I think this gets back to why apps are probably going to go away. It's a highly conversational method. So think about every time you've ever interacted in the past with a company. You get that green text from a message that has six digits or maybe it has the full nine digits, but you have no way of knowing if it's actually your bank or if it's actually a hospital or a pharmacy telling you for a pickup for your prescription. That's all changing, and RCS, it's just a new standard. So essentially, companies need to register in order to take part in it. And what that does on the back end and for the user experience is that suddenly, when I get a message from Verizon telling me that my bill is ready—this is already live in production today—I get a message that actually has a check mark, Verizon logo in the top screen, doesn't show a phone number, but it's telling me that this is Verizon.

(Joel Beasley at 00:24:35) That's what that is? Yeah, I thought it was this custom thing they did with Apple.

(Jacques Klick at 00:24:39) No. So, well, Apple and Android are the gatekeepers of the technology, RCS, and they're now—

(Joel Beasley at 00:24:46) Turning on—it's like an open technology and built—

(Jacques Klick at 00:24:50) And now they have to partner with the carriers, right, because there's that connection point. But, yeah, ultimately, from a device perspective, that's how it shows up. And the cool thing that you can really do is that now you have these interactive app-like experiences happening in your message thread. So that Verizon example, when it sends me my bill, within the text thread itself, I have clickable options—

(Joel Beasley at 00:25:11) Yeah.

(Jacques Klick at 00:25:11) —to automatically do that. And just think about how that evolves for, let's say, airline check-in. All the stuff I used to have to log in to my airline app to pick my seat and check in and all that stuff, it all happens in that messaging channel. So the whole app experience, it doesn't mean it necessarily goes away or it's not useful. It's that it gets repackaged in a text thread where we know a lot of us spend our time. That's where we're talking to our friends and our family members, and suddenly the brands are going to show up more and more there.

(Joel Beasley at 00:25:40) I would love—I fly once a week at least, and my goodness, how great would it be not to have to go to Delta to Southwest? I have to keep all these apps up to date. I've got to log in. I always forget my customer number. If they could just do what Verizon does where I saw it, it gave me something. It either was trying to sell me a new iPhone or had me pay a bill. It did something. Or maybe it was when I went to Canada and I got some message. But I noticed that—that's why I thought it was a custom Apple-type thing they did, because it had—it wasn't a normal bubble. It was—I could click on it and do a dropdown and some other stuff inside the message.

(Jacques Klick at 00:26:17) I mean, it's very—it depends on which market you're in. In France, it's been highly adopted, and so you're already seeing companies make these really amazing, interactive marketing campaigns, and it takes over your whole screen. But it's revving and ready to go in the US. It's just the adoption hasn't been there yet, and it takes a while to register your brand for it. But I think once companies start to see the value—and what we're seeing with our customers is they're seeing three X, four X engagement on those marketing emails that they send or the critical customer notifications they're sending. Think about, instead of getting a one-time passcode—we all hate that, right? Having just verified directly in the text thread where maybe it does Face ID directly in there and you're logging into your account from there. And then the next piece, we already touched on it, but the conversational aspect. So maybe it's I don't even want to click the "Pay My Bill" thing in Verizon. I just want to respond in a text to Verizon, like, "Hey, yep, approved. Pay my bill." And then the agent is able to converse and figure that out and do what I want it to do just by interacting with it via text. And then, of course, the next evolution, I think, will be voice.

(Joel Beasley at 00:27:27) This is wild. So anyone can go register for this RCS thing?

(Jacques Klick at 00:27:31) Any brand. I mean, we make it easy for companies if they want to do it with us. We can partner with you and figure out and make that registration process super easy. But I would say it's one of the most underrated things in marketing today that brands are not using enough because—

(Joel Beasley at 00:27:47) Is it tens of thousands of dollars, or is it pretty reasonable?

(Jacques Klick at 00:27:50) No, it's actually—the cost is very similar to your SMS, and there's almost no—it's a very low cost to register your brand in general.

(Joel Beasley at 00:27:59) I want to do this. So I get about 100 to 150 new text subscribers per month. I do comedy on nights and weekends. I go around and do comedy shows. And so people, they subscribe to get notified. They give me their phone number and their ZIP code, and then—because they see my clips online—and then I'll send a text out 48 hours before I'm going to be in their area. Like, "Hey, we have a show Friday in Chicago." Or I'll be like, "Hey, we have a show Friday in Chicago," and I'll send that out after this podcast.

(Jacques Klick at 00:28:29) That's awesome.

(Joel Beasley at 00:28:30) And but it's just a basic text. I use GoHighLevel, and I have something in there. How would I go from that scenario to using this Sinch RCS thing? How would I do that?

(Jacques Klick at 00:28:45) I mean, you—number one, just sign up for our API, and you can start building right away, and then someone from our team will be in touch and help you register. We're also right now—it's in beta, I believe, but we're about to release it. It's a business application, so it's more built for CMOs and non-technical people who want to be able to send these types of campaigns themselves. And so that's probably the best option for someone like yourself, I would say. Even though I think you do have the technical chops, you could build your own—

(Joel Beasley at 00:29:12) But I don't want—yeah.

(Jacques Klick at 00:29:14) It'd just be easier to do it through the app, yeah.

(Joel Beasley at 00:29:16) I'm a visual person, and I'm like, okay, if I'm going to design this RCS experience, you're going to want some interface that's going to help you design it to be cool and to test it. And I'm not going to build those interfaces. That's a lot of work.

(Jacques Klick at 00:29:29) Well, not to get back to what we talked about before, right? But RCS is a skill for us. Creating that campaign and—so, again, you're not even really—eventually, you're not even really going to have to do that. You're just going to type into your Claude code or IDE and say, "I want to send an RCS campaign, and I want it to be interactive in these ways, and I want it to look like this."

(Jacques Klick at 00:29:51) And we're packaging the skill to say, well, RCS, the dimensions of the carousel has to be x, y, z, and all these different categorizations that we're familiar with because we have that domain expertise. And then you're just prompting and telling it what you want it to look like, and it just does it for you. So you asked me, like, how far away is all this stuff? It's coming.

(Joel Beasley at 00:30:12) It's right now.

(Jacques Klick at 00:30:13) Seeing it live in production already. It's just the adoption, and there's so many different things that everyone's getting distracted with with AI things. It's just hard to keep track. I get it.

(Joel Beasley at 00:30:23) So I could feasibly get off this call, go design an RCS experience, sign up with you guys, and then have it deployed in, like, a couple weeks. And the longest part of that would be the approval process of me getting approval of the...

(Jacques Klick at 00:30:35) Rest of the day.

(Joel Beasley at 00:30:36) Yeah. I want to do it. I think that would... if I'm, like, a big tech guy, like, how could I not adopt the most interesting... I want my fans not to have a link to, like, oh, here goes the ticket page. I want it to be... I don't know what I want it to be, but I'm going to figure out what I want. I want it to be the coolest experience ever.

(Jacques Klick at 00:30:55) It's a huge arbitrage opportunity because it's available, people can use it, and very few brands actually are. So why would you not get ahead of the game? And even just think of it from the trust and security perspective. Think of how often you ignore those messages that hit your inbox where the number is...

(Joel Beasley at 00:31:14) It's like 844. And that's what I don't like.

(Jacques Klick at 00:31:18) And you just want to delete that. But if it came from joelcomedy.com...

(Joel Beasley at 00:31:23) That's what it is.

(Jacques Klick at 00:31:24) Okay. Yeah. I'm going to keep it. There we go. I'm plugging your show already.

(Joel Beasley at 00:31:28) Let's do it. Let's do it. Yeah. I try not to talk about it too much on the podcast. It's pretty rad.

(Jacques Klick at 00:31:33) I think it's awesome. I'm a huge comedy fan. Austin's a comedy town, so we love to have shows here.

(Joel Beasley at 00:31:37) I know. I'll be there in, I think, two months. We're doing a couple shows in the area. But back to this.

(Joel Beasley at 00:31:46) So, yeah, dude, I am just thoroughly excited. Is that business thing that you talked about, that you mentioned they could build, is that in, like, a month, a quarter, a year? Do you know when that is?

(Jacques Klick at 00:31:56) The software to help business users actually do RCS? Yeah. No. That's within this year easily. It's going to get rolled out very quickly.

(Joel Beasley at 00:32:04) Could I do it just with Cursor and, like, installing your skill and becoming a customer and just telling it what I want?

(Jacques Klick at 00:32:09) Drop the API key. Yeah.

(Joel Beasley at 00:32:10) Yep. Now I don't even tell it what I want as much anymore. I ask it for ideas. You ever do that? You're like, hey, I need some ideas here.

(Jacques Klick at 00:32:16) Yeah. I mean, how many times are we using Claude to be the thought partner and, like, how would you structure a prompt for this tool that I'm using? Every day, that's how I'm using it. It's scary.

(Joel Beasley at 00:32:28) OpenAI hardware. Is there... I didn't know that they were coming out with some hardware. Do you know about that?

(Jacques Klick at 00:32:34) Yeah. So I think this is, again, to my point of the way that we typically interact with AI today tends to be on our desktop and tends to be in chat, but that is changing very quickly. And I think this is an underrated story that not enough people are talking about. We've seen examples of companies doing AI tech in their homes, Alexa, you know, things like that, Meta's Ray-Ban glasses that can do things as a wearable device. But I don't think we've seen it come from a company like OpenAI where the model is so much more advanced.

(Jacques Klick at 00:33:09) And I don't know if any of your listeners have already started to use the chat interface with ChatGPT where you can just talk to it, and the quality is so good. Not only of the voice sounds human, but for us as a communication channel, we often think about the latency of that voicebot. Because if it's going under 0.8 milliseconds latency, now it's sounding like a human. Because when I end my sentence, it has already interpreted what I said, created its reaction, and then it's giving me a response automatically versus, you know, you interact with your Alexa, it's huge pause, choppy sounding. You know, if you interrupt it, it gets all flustered. This is, I think, the future. And so, yeah, OpenAI famously, a year ago, announced that Jony Ive, who was the head designer for Apple for a number of years, designed the Mac, iPhone, Watch, iPad—I mean, you name it—known for this very minimalist approach to user design for products, and they're working on some kind of device. There's a lot of speculation that it could be a wearable, it could be a pen, could be AirPods-type thing, but something that we will most likely interact with without a screen.

(Jacques Klick at 00:34:26) And it's going to be tied to that model, and it's going to help you do things and help you answer questions. And it's fascinating because I think none of us have had that experience with something out in the world tied to a model as powerful as OpenAI's model is, and tied to the connection points that a company like OpenAI has today. Right? We talked about it. Every application is now either through MCP or through a native integration into ChatGPT. So it goes beyond what the chat model could do. And now every single company is going to rush to figure out what's my use case for wearable tech, and how should my customer... how do I want my customers to interact with me knowing that ChatGPT is going to be kind of that middleware, for example?

(Joel Beasley at 00:35:14) Did they release, like, what the device actually is yet?

(Jacques Klick at 00:35:18) No. It's still very speculative. I mean, depending on... you might read a blog that says, oh, it's for sure going to be, you know, AirPods. For sure it's going to be a pen. I haven't seen anything that truly has made a commitment to anything, but the set releases, I think, later 2026, early 2027, and I think it will fundamentally change the game. I mean, as a communications company, we're thinking about this. Right? Because is the way that people suddenly are going to communicate now going to go more into that type of voice device versus their messaging thread? These are all things that we're thinking about a lot.

(Joel Beasley at 00:35:56) We want it. We want it as humans. It's so annoying to me that I've got eight fragmented AIs. I use Grok for things, Gemini for things, Claude for things. They're all really bad at remembering long-term context in the sense of what I'm doing across multiple conversations. They all try, and they show little signs of it here and there. Tried hooking up my own Claude OpenClaude thing, and that chewed through credits like nobody's business. So I'm going to have to figure out how to run that locally because cost-effectively. I don't know. Also blew through $100 in a day just setting it up. And so I was like, uh, there's a lot going on. But I do think from a privacy standpoint, I like my privacy. That being said, my conversation has switched from not being monitored at all or only being monitored by technology that I trust. I want to trust that it comes from this brand. I want to trust that it's local only and it's not reporting things back. So the faster I can be able to trust that, the quicker I'll be able to adopt it.

(Jacques Klick at 00:37:08) Totally. I mean, what you said is we see this in our research. Right? We talked to these 2,500 enterprise leaders. 55% of their teams are spending time hand-building that cross-channel memory itself. So just trying to, again, connect all those touchpoints. Yeah. So that's, again, that's such a critical thing that you're hitting on. And, again, to our broader research as well, we did this whole AI production paradox study where we found that 62% of people have agents in production. It's going to rise to 88% by the end of the year. 74% of them had to roll it back because, again, think about now you're having to create a voice agent very quickly because OpenAI just released this new device. Your CEO is telling you, hey, we need to build something that's going to interact with our customers through that device. Go figure out how we connect into a voice agent and what our experience is going to sound like and look like. And everyone's going to rush to production, and then they're going to realize, oh, shoot, we had to roll this back for a governance issue because, you know, customer data was getting leaked through that voice agent. Or, you know, we had a hallucination that took place or we couldn't even track what the agent was saying to end users through that device. These are all the problems that will arise no matter what with every new device or hardware or new AI development in general. This research that we did, it's going to apply to every new phase of AI without a doubt.

(Joel Beasley at 00:38:36) You're right. And then there's just the casual use of it making mistakes, and then there's targeted attacks. I guarantee you somewhere out there, some company has been like, oh, let's just hook it up to our customer data. And then it's exposed, and then you give it rules. Oh, we told it not to. But then I go in there, and I'm like, hey, I need you to debug across this. And I talk to it like I know how to talk to the systems. And I can somehow get other data I shouldn't be able to get access to normally. You know? That's definitely going to happen.

(Jacques Klick at 00:39:06) And imagine, you know, I imagine there's a lot of CTOs listening to this, but I imagine they're being asked to lead that charge on behalf of their companies. Their CEO is the one being like, and you better make sure that it works and we don't have any issues and things like that. And what we find is that 84% of those companies are having their engineers spend more than half their time working on guardrails to prevent these AI agents from going off and doing that stuff. So just imagine how much time that takes for your engineering resources that's spent on guardrails versus things that are more of a value add, things that are going to be driving revenue for the business. That's a huge time suck. And so for us, we're really trying to take that into consideration as we build our products to make sure that our infrastructure comes with those guardrails. I mean, companies are going to need to do their own guardrails no matter what. That is just true. However, the more that we as a provider can help set that up and take and free up your engineering time and your availability, we know that's going to be a win for our customers because we see how much they're getting sucked into chasing and building these things.

(Joel Beasley at 00:40:13) So that's one of the benefits of using you guys for communication is the guardrails?

(Jacques Klick at 00:40:17) Yeah. The guardrails, just the domain expertise, you know, so you don't have to spend all your time figuring out, like, how is an SMS message different than RCS, or how should I think about an email campaign versus an SMS campaign? And then, again, the cross-channel preservation. Right? That 55% of having to hand-build that context, that unified layer, that's, I think, the big benefit we see with our customers is they're just able to create one unified thread of who their customer is. It's preserved. It works across channels. You know, once if a human in the loop needs to take part, because that's something we haven't talked about, but we all recognize there still needs to be that human in the loop. And so arming that human with that understanding of all the interactions that this customer went through so that they can quickly help them resolve their case or whatever the use case is, that's super critical.

(Joel Beasley at 00:41:09) When did you join the company?

(Jacques Klick at 00:41:12) I've been here for a year and a half. I lead market strategy. So my whole job is to think about how is our market evolving. I spend a lot of time speaking to our customers, some of our most advanced customers. I speak a lot to developers, and I speak to industry analysts to really figure out, like, what's next? I mean, and that's why I'm thinking a lot about the hardware implications because, again, at my company, we've been thinking, you know, mobile apps and our ability to reach every device on planet Earth, which is awesome. But if all the user adoption is going to the LLM tools, a year ago, I kind of came into the business and sort of saw, like, hey, we need to evolve our AI strategy. Right? How are we integrating with these platforms and, you know, making sure that customers can use us on the back end, and we're the infrastructure that scales so that someone like yourself can go in tomorrow and start sending messages or emails and use our infrastructure. And if you ever get to the biggest levels of comedy and suddenly you need to send millions of messages, that we can scale with you and partner with you along that journey, like, that's the way that our business grows. We're that silent infrastructure partner that scales as AI. And I'm making a bit of a joke there, but even someone who's a small business, if you're sending 100 messages today, I guarantee that number's going to go up tomorrow because of the nature of conversational aspects of messaging. Because what used to be a one-way message that you notified a customer, now you're getting a response back from that customer, and they expect a response very quickly to what they just asked. So you just have to have a partner that's going to help you scale that communications layer.

(Joel Beasley at 00:42:52) Oh, do you guys handle that too, like, the LLMs that respond to text messages?

(Jacques Klick at 00:42:57) Yeah. I mean, conversational AI was a huge bet that we made because we saw that convergence of, oh my God, we have these channels. We're the last-mile delivery for our customers to reach their customers. But suddenly, we saw, like, oh, there's all these conversational capabilities, plus you have these Gen AI tools that help you create the conversations. So, yeah, we have our own proprietary tool. We call it our agent builder tool, and it's based on all the large language models. You bring whatever model you want. You select whichever model you want within our platform. Same goes with our voice agents. So if you ever want to deploy a voice agent for a telephone line or a contact center, you know, Eleven Labs, hugely amazing voice agent capability. You just select that you want to use their agent or you want to use Google's voice agent. But, again, it runs on our infrastructure. So our whole strategy has been we need to integrate with every AI tool and company possible, but ultimately what no one can recreate or replicate is the communications infrastructure that we power. And so that's what really excites us when we think about the future because applications will change, you know, the way that users interact with things will change, but the underlying truth of a message needing to deliver to a customer is never going to change, and we get to own that piece of the puzzle.

(Joel Beasley at 00:44:18) Oh, I love it. My favorite businesses are infrastructure businesses, and you guys sound like you've got some smart people over there. Did you get some good equity options in the company, hopefully, because it... I would want them.

(Jacques Klick at 00:44:30) Like all tech companies, yeah, it's definitely a part of it. But for me, I did join the company. I followed a woman who I'd worked with at a previous company, and she was just a phenomenal leader. And so for me, I saw the opportunity to work with her again, which was my number one thing. She's one of the best product marketers in the world, which we could do a whole podcast on what product marketing is and how it works.

(Jacques Klick at 00:44:54) But she was just a phenomenal leader. And then when I actually took the curtains back, I thought, oh, boring telecom technology infrastructure company. And then I peeled the onion a little bit and saw, like, whoa, this is a massive opportunity. I mean, we literally have the potential to reach every device on planet Earth, and we do that for 200,000 customers. And just the rise of AI across each of these distinct channels, it was such a huge opportunity that I was like, I have to capitalize on this. And I've been super happy with the company, and we're Swedish headquartered, but we have a huge US presence and a huge APAC presence as well. We run all around the globe, and we just have a great culture of people building and getting excited about building. Yeah, it's just a really exciting place to be.

(Joel Beasley at 00:45:42) Gotten to visit HQ?

(Jacques Klick at 00:45:44) Yeah. Yeah. I was in, we actually did a huge executive off-site in Malmö, Sweden. Yeah, actually a couple weeks ago. Yeah. Big, you know Malmö?

(Joel Beasley at 00:45:52) That's a cute, oh, I've been in that little town. It's so sweet.

(Jacques Klick at 00:45:55) It's like a little stupid city.

(Joel Beasley at 00:45:57) Yeah. Yeah. It's really, I have a friend that lives out there, actually from the podcast.

(Jacques Klick at 00:46:03) That's awesome. Yeah. So we, I mean, but that was, like, it wasn't an executive off-site like in a traditional sense. Like, yeah, we had some fun, and the Swedes are just, I love their personality. Like, US company off-sites, like, it's very rigid and, you know, in some places, maybe things get wild at like the happy hour or whatever, but this was just like fun. You know, we were doing karaoke. People were getting on stage dancing. Like, just that, they just have a better sense of, I think, just kind of letting themselves go and not taking themselves too seriously, and they have a dry sense of humor. I just love it. So it was a fantastic time.

(Joel Beasley at 00:46:37) Small world. Wow. Good stuff. And then is Lovable out there in Malmö too?

(Jacques Klick at 00:46:43) No. They're, I think they're in Stockholm, but they were our big first actual partner when it came to, you know, integrating natively into app builders or LLM tools because we had that Swedish connection. But it was just such a crazy experience because I was a very early user of Lovable. I mean, I first started building like a year and a half, almost two years ago, like when they first came out. And I was just immediately like, this is changing the game. And then a couple months ago, for us to announce that we were like the first communications platform partner they were gonna use is just amazing to see that genesis come together. And it just came because we had an employee who used to work at Sinch, who now works at Lovable, made the connection. They saw the opportunity, right, because from their perspective, they know people are gonna build apps, they're gonna build websites, and those things are gonna need to communicate with end users. And so we tested it with our email product, worked really well. So we're opening the door to do the messaging side and the voice side, and this is our land grab that we see. We see that we need to get deeply integrated with those models to get access to the customer base, to have the user experience be more seamless, and then we're just the platform that ends up sending the messages in a compliant, reliable way.

(Joel Beasley at 00:48:01) Dude, I love it. I love, and I'm a big fan of Lovable. I'm a customer. My comedy sites, I built with Lovable, and it was brilliant. So.

(Jacques Klick at 00:48:09) It's amazing.

(Joel Beasley at 00:48:10) Yeah. It's so cool to be able to just like send a message, like, add these three tour dates, and then it just does. And then.

(Jacques Klick at 00:48:16) Yeah. And when it finds something that it breaks, it like says, hey, I found this break. Do you want me to just fix it? Sure. Fix it. Yeah. Yeah. Yeah.

(Joel Beasley at 00:48:23) All right. Last thing we're gonna wrap up on, what is something that you are learning right now as a leader?

(Jacques Klick at 00:48:29) Honestly, it's something that I learned a long time ago, but I think it's never been more true today, and that is to go towards the hard problems to solve. You know, I've always, in my career, been challenged. I was a former management consultant on the strategy side. I've been a product marketer. And there was always times where you could have just stuck with like, hey, let's just do the basic, easy, stay home safe, kind of fix an issue, hope that it works, whatever. But I always had mentors that kind of pushed me to go find the most complicated problem, the one that everyone else is running away from, and just dive into it. Because number one, what's the worst that could happen? No one else wants to touch that problem, so you're gonna be seen as someone who wants to go and challenge yourself. You're gonna learn from that experience. And if you do end up solving it, it's gonna help you in that next iteration. And why I think it's never been more true today is like, because all those easy problems are now getting solved by the AI and the agents. If you were someone who just moved text from one box to another, that is going to be automated. And so the people that are gonna succeed, I believe, and what I'm trying to teach my team and push my team to really start to learn and develop as, you know, facilitators and orchestrators of these agents is how to solve complex problems while using these new technologies. We talked about the limitations that each of these models or these agentic platforms have. You have to know those as the human in the loop, and you have to be aware of when to use Copilot versus Codeium versus Claude for what practical use case. So I think just for everyone out there, like, go run at the hard problem. You won't regret it in the end. What you will regret is just sitting back and being like, oh, I'm just gonna keep my little easy task log going. And, yeah, the tools are helping me today, but quickly, they're gonna replace you.

(Joel Beasley at 00:50:26) Well said. We're gonna put a link to, we mentioned this research thing several times. Yeah. This episode. Yeah. We will have a link in the show notes below for people to get access to that full research. And then for listeners that wanna connect with you, LinkedIn is a good spot? Or?

(Jacques Klick at 00:50:42) LinkedIn's the best spot. Yeah. I mean, I did all the research that we did. So we talked to 2,500 enterprise customers across, you know, all different types of countries and industries, all people that have experience actually launching agents into the wild and putting them into production. So if you do wanna connect and you wanna learn more about that research, go to sinch.com/ai-production-paradox, or just connect with me on LinkedIn, and I'd be happy to run you through the research. Yeah. I'm super available to do all of that.

(Joel Beasley at 00:51:14) Amazing. Well, Jacques, we made a podcast. How do you feel?

(Jacques Klick at 00:51:17) Yeah. This was great.

(Joel Beasley at 00:51:19) Thank you so much for listening. And if you found this episode useful, please share it with a friend or colleague who you think would get value from it. And if you have topics that you'd like to hear discussed on the podcast, either add me on LinkedIn or send me an email [email protected]. Every time I get an email or LinkedIn message, it absolutely makes my day and inspires me to keep going.