300 |💥 AI done right - behind the scenes of the most advanced AI company I work with, with Ari Supran, CEO of Sonance

22:26So this is what our Sonance brand guidelines would do. It injects the logo, the word mark, the fonts, the color scheme so that everything has a consistent look. And this is still, again, a work in progress. We're still evolving how this works, but it's really been, you know, it's almost magic, right? The employee doesn't even know that it's happening in the background. They just start building, could be as simple as an artifact in Claw of like a simple lightweight app or something more sophisticated like this and it just comes out looking exactly the way we want it consistently every time without them having to know any of that or really do anything.

22:57So that's exciting. Let me skip over to another. So our iPort brand selling into enterprise, we have different verticals that we sell into restaurants, gyms, spas, retail, et cetera. And I thought this was a really good thing. Our AI leader, Derek, has his own agent who does coding for him that he can delegate a project to. And And this is a good example of where our, Derek just told me this last week, our marketing department at iPort wanted to make a microsite for iPort specifically for the restaurant industry and asked Derek, should I use WordPress? How much do you think this will cost? And he said, well, send me your design brief.

23:34And the next morning he came back with this microsite for the restaurant industry and looks exactly like our iPort traditional website, but it was designed specifically for the restaurant industry as a little microsite. And what I love is, number one, how beautiful it looks, the animations, everything's so fluid. What's mind-blowing about it is that his, not him, his agent built this in about four hours. But what really blew me away is that he asked his agent to put a CMS into the website as well. So what he can do now is just come in here, just like a PowerPoint slide or a keynote slide, and make any change you want to any aspect of this site.

24:13because his agent built a CMS into the website that I can then change a image just as easy as that. I can change copy as easy as this and treat this like a PowerPoint slide. And the code is written behind him as he makes changes with this WYSIWYG interface here. And that just starts to really get my ears shifting of like, wow, we could make microsites per client with our iPort business so that we have something very specific to them. We already do custom renderings of their space with our product, but give them a website or a microsite that they could go to. And then when I think about our customers, because we do a lot of AI evangelism in the industry and trying to help inspire our customers.

24:56And quite frankly, I can think of 10 to 15 of them that are just doing mind blowing things and transforming their businesses and creating internal apps that their business will run. But one of the challenges in our industry is giving the client, this typically wealthy end user, a proposal. And today it's a series of PDFs and builds and materials that could be as easy as a microsite that can be auto-generated with brand focus from the integrator to the client, showing all the different subsystems of home automation and all the things that they're doing in a rich way that is so easy to build and then make changes to before sending it out to the client.

25:30So I thought those were some good examples to share with you of a gamut of different things. And there's so many other things. I think about our factory up in Nevada and just have an individual up there that has been building internal apps and really building one super app that the factory is beginning to use to create more transparency of the process and where orders stand within the process between custom engineering, finishing, assembly, et cetera. and just creating more transparency and clarity for our internal team as dashboards and other things to know where we stand in the process. And he's just doing some amazing stuff.

26:05And I think that's so exciting to me because I think that in the old world, there was sort of two ways to really add value to a company. You could either be this like amazing individual contributor, maybe a top salesperson, or ultimately you could be a people leader. And I think in this new world, there's this third avenue that I'm so excited about, which I like to refer to as orchestrator, that, like I said, I think the key here is they don't just build for building's sake. They actually understand the business well enough, get up out of their chair, get away from the keyboard, go talk to people, find out what the business pain points are and understand what's possible.

26:39And then they're building solutions that solve problems and they're not stopping there. This is the key. I think they understand that the tool they build is only as good as their ability to get people to use it, to know how to use it, and to ultimately create a success factor of are we getting the right business outcome from this and not stop until we are and then own it all the way through and i think that is really exciting because i think that in the next two to two and a half years our intention is to completely transform just about every aspect of our business we may still be selling the same types of products that we were today but every other part of our business we plan to transform using ai as well as getting i think the two key points here these models and you and I've talked about this before.

27:21I personally think the models, frontier models today, like Opus 4.8, GPT 5.5, already more capable than to do just about everything that we'll ever need them to do. So that is not the bottleneck, in my opinion. I think the bottleneck, if anything, is getting data clean, organized, and mapped in an intuitive way. Our plans there are a data lake house with Azure Databricks, and we're building that out now. And then secondly, is getting context right. I think that is the biggest unlock that anybody can focus on, which is, especially if you want to stay model agnostic, which is to all of these systems that Chachapiti and Claude have memories.

27:57But we want to go beyond that and start capturing all of the context in the business into something referred to as a knowledge graph that we believe will not only capture all that information, but I think the key is that it can inject the right context into everything everybody's doing all the time with AI so that everything's more relevant to our business. Everything hits closer to the target from the get-go. So it might be something like a good example. We've got an individual building an app for customer onboarding and we have a very diverse business. We have between the way we do business in Europe differently than in the US with residential dealers, commercial dealers, enterprise customers.

28:35There's so many different nuances to this. And so he built an app or is building an app that will assist in onboarding a customer, which today is a very broken process here of a system of emails going into spam because they're docusigns. And we're building an app now that has sort of, we like to refer to as the pizza tracker of where does this stand? Whose approval is it waiting on? All in one place. And that app that he built is so powerful and amazing. Yet when he showed it to our sales leaders, they were like, this is incredible. However, you miss that there's this nuance here. We do things differently here, there, et cetera.

29:08And so he's having to go back and make all of these changes. Our belief is in the future when we can connect his cloud code to that knowledge graph, the very first iteration of the product is just going to be so much closer to accurate and what we need because today cloud code doesn't know those nuances, nor does the gentleman who's building the app, who's a technical resource and doesn't understand all those nuances of the sales and go-to-market strategy. but our knowledge graph will. And so we're going through a process right now of capturing this knowledge, codifying it in a way into this knowledge graph that I really think between getting data right and getting context right, that's where we really get the unlocks.

29:44And that's the direction we're heading right now.

29:46Isar Meitis:I love everything you said. We can probably spend the next 10 hours deep diving into each and every one of those, but I'll pick a few. I'll start with explaining the knowledge graph and why it's so critical. And I'm in the same path with my company as well. The idea of a knowledge graph is connecting the dots, right? Those of you who don't know what that means. So if you take a simple example, if you allow the AI to write proposals for you right now by giving it access to, let's say, a transcript of a call. So that's awesome. It has access to the transcript of a call. But maybe there were five calls.

30:16Isar Meitis:Maybe there were 65 emails before that. Maybe there's information in other departments in the company that already worked with this client and they know stuff from the past. Maybe the technology that they're using is a little different than the average and it needs special attention. that information, and I'll enhance what Ari said, that this person doesn't have that information. Most likely, nobody in the company holds all the information that is required to do this. That information is broken between people's heads and multiple systems. If you build a knowledge graph and give the AI access to it, and it's not magic, like there's a process, there's cleaning the data, there's documenting that, like there's a process.

30:53Isar Meitis:But if you follow the process, at the end, the AI can figure out where the right data is, pull the right components in without breaking the context window and still benefiting tremendously with whatever it is that it is trying to do. So that's one aspect, just to clarify on the last one. But I actually want to go to the point you started with, which is the people, right? And I think both you and I really believe that that's the key to everything in a company. What is the current structure from a people perspective? And also, what were the steps to get you there? Because when you and I started, obviously, was you and me.

31:24Isar Meitis:There were no other people that were involved in the process. Sure, sure. Well, let's see. From a people standpoint, the process goes back to about two years ago, making those investments and bringing in technical resources. We brought Elliot Amador in, has a lot of background in machine learning and AI, and has really been the backbone of a technical team. We brought in Thomas, a recent UCLA grad that has a background in some programming, as well as acoustics, which was a really nice find for us, who's been instrumental, moving Derek into that role. So we've got that core team. We found a gentleman named Joshua on our team that was an electronics test engineer, or is an electronics testing engineer, but had a proclivity for AI and started building tools.

32:05And so he's kind of joined that core team. And then we started, you know, we started off with a committee.

32:10Isar Meitis:And I think we realized that committee is not really the right thing for us. We didn't really need a committee to come together to make decisions. What we needed was a community of people to share knowledge and to inspire each other and teach each other what's possible. And so that led ultimately to this orchestration team. I will say that we, going back two years ago, I remember starting to, maybe it was two and a half years ago, starting to come to our head of IT at the time and say, have you seen this? Have you tried that? And realizing quickly, and he and I have a really good relationship, that he was so head down trying to keep the lights on of our current ERP and other systems that it just wasn't It wasn't fair really to expect him to also be trying to stay up on all the developments in AI.

32:51So we built this technology and innovation team sort of in parallel to our traditional IT team. I would say that that worked well in terms of getting us the traction that we needed and getting going. And at the same time, I think we'd let that go as two teams for too long. And we're actively bringing them together as to one team with one set of priorities because there's so much interdependence between the AI efforts and our traditional IT efforts. from security and networking, all sorts of things like that. And we are clearly, I know I keep saying this, but we are on a journey. We are not at the destination.

33:21I don't know if there is one, but we're making great strides. And we also are fully recognizing that we have, we're out there, maybe close to the frontier of what's happening. And we're just going off of instinct in some ways. There's really not a playbook here, but we are trying to stay connected, talk to a lot of different peers and companies and customers of what they're doing and try to stay abreast and be nimble. I think that's the key for us is to just make sure that we don't lock ourselves into any one provider model agent framework. We haven't really touched on agents much in this conversation yet, but that's certainly something we see in the future of introducing agents as coworkers to our employees in Microsoft teams.

33:58And by the way, I should have mentioned this earlier, all the apps that we'll build internal apps, I should say, we, all of those will appear to our employees in Microsoft Teams. So what we learned is it's really hard to get your app into Microsoft Teams for the world to see. It's really easy to get your app into only your company, a single tenant app, to put it into Teams is easy. And that's where agents will show up as well. Like I said, Wes, we haven't introduced agents. We're using them on an experimental basis. I have one, Derek has one, many of my other senior leaders have them. But eventually we see an opportunity to, we're building an agent that essentially spawns new agents that we have the ability to control what it has connections to who it answers to who can talk to it and what its role is and lock that in so that an employee can't override that but they can use that agent to help them get work done that's exciting and also very much something that we're keenly aware has a change management human side of it that we need to be careful about one thing i just gotta always go back to is before all this AI stuff started, we had a really healthy culture.

35:02We're a privately held company. I can't speak for other companies in the world that are publicly traded or private equity and what their motivations are. And I don't have this massive call center of a thousand people doing the same thing. I've got really hardworking, passionate employees that at times feel stuck because the processes and the systems that we've had are inflexible and they feel like, well, it is what it is. And I think that culture is actively changing right now to say, if it's not the way we want it to be from the customer experience and the employee experience, we can change that.

35:34And we are seeing evidence of that throughout the company right now. And that's what gets me so excited to come into work every day. And I'm seeing so much from our customers too, that are doing some things that inspire us. And there's, like I said, a handful, maybe 10 or 15 of them that are doing things that, you know, they're really building internal apps to run their entire business on, from scheduling to proposals to you name it, all into one app. And that's exciting as well, because we played a role in sharing knowledge with them, learning from each other. And I think that just speaks to our pillar of authentic partnership and our ability to help our customers be successful, not just with our brand, but just in general.

36:14And so AI education, AI evangelism has been a big part of that. And we're seeing tremendous exchange of value with our customers as a B2B company that believes in these types of permissions.

36:23Isar Meitis:Yeah. When I, when I work with businesses and we start by writing, you know, the why, like why do we want AI? And, and, you know, that's the core for everything. And I always go, at the end of the day, we do what we do to provide value, but we don't do what we do to provide value to the owners of the company. We do what we do to provide value to the people in the company, to their families. Maybe they can get two more hours a week at home and play soccer with their kids or whatever it is that they like to do. We just provide value to our suppliers and distributors and consultants and like everybody in our ecosystem.

36:54Isar Meitis:And AI really has the ability to do that. And that is very, very rare. You know, there was the old saying that has been true for years, you know, faster, better, cheaper, pick two. And it's the first one, the first time ever you don't have to pick. You can do faster, better, cheaper, more valuable with less time for the sake of everyone. And as much as there are definitely the negative sides of this, like what is this going to do to the world and relationships and the economy and the workplace in general. There's a lot of open questions that nobody has access to. From being able to build value to everybody involved at scale, this is the most exciting time in history.

37:37Isar Meitis:I want to ask you one more thing that I think would be very helpful for people because you live through this literally every single day, every single hour, and you think about this all the time and you have an incredible team that is involved in that creation, which has created this, you know, snowball effect of just growing and attracting more people, a flywheel effect if you want. But there's a lot of people out there who are just getting started. Like they barely know the difference between Claude and Chachapiti if they even heard of Claude. If you are in a leadership position in a company right now and you're just getting started, what are the three to five things that you would say, okay, sit down, write this stuff right now and do these things to accelerate your success or your chances of success.

38:22Isar Meitis:Again, you and I don't have a crystal ball. We don't know where this is. Chances of success in implementing successfully. Yeah. I would say first and foremost is if you're just getting started with the adoption, somebody in leadership, don't delegate it. Somebody has to have their wow moment. And I don't want to sound like a broken record, but I think that what I keep hearing people say is things like AI is this, AI is that. It's not this. It can't do that. As if it's this monolithic one thing. AI is not one thing. And the model matters, but the product that the model that you use the model matters as well.

38:54And what we've learned is that these coding harnesses like CloudCode and Codex are extremely capable. And while they are coding harnesses for coding and not everybody, so I'm supposed to speak here to the people that are new, just understand that you don't have to be a programmer to leverage coding because what is behind every spreadsheet and every PowerPoint deck is code. And so what we've figured out is that these systems like Cloud Cowork and now the Codex app, not just ChatGPT, but OpenAI's Codex app are really this gateway into doing knowledge work with AI that is so much different than chat.

39:29And when I say chat, I don't mean ChatGPT, Cloud Chat, same thing. This chat idea that I ask a question and get an answer, we are past that. And I think that is the number one thing I would encourage anybody out there to go experience for themselves is open up Claude's desktop app, go to Claude Co-Work, it's on Mac or PC, and point it to something. You know, you don't upload files into it like a chat interface. You point it to a folder, a traditional folder. I'm on a Mac, so that's Finder, Windows Explorer, that has a, could have 57 documents in it, four spreadsheets, PDFs, Word docs, it doesn't matter.

40:02And ask it to just familiarize yourself with the contents of the folder ask me questions to make sure that we are on the same page and then this is what i want you to do i want you to in those and it's an action word here create something build something uh design something analyze something actions because this is not chat we're not here to get an answer we are here to get stuff done and we are here to delegate to a model in a product like cloud co-work and i think once it starts to analyze the files and start to ask questions of you, it is that moment that I think most people that I've turned on to this start to realize that this is different.

40:39Because what they see is that these questions are no longer generic questions to clarify what you want, but they're incredibly insightful questions that are based upon its analysis of the files in your folder.

40:49Isar Meitis:And so much so that even if you don't know what you want to do, there's obviously a great technique called reverse prompting, where you simply say, given what you see inside of this folder, what do you suggest we do? And I take that, I learned recently, just take that even a step further and say, what do you suggest? Let's say I'm a busy executive and I need to help visualizing this data. What is the suggested prompt that you think I should give you in order to get there? And it writes out the prompt for me. I say, go do it. And right in front of my eyes, hands off the keyboard, I see graphs built and tables built and all sorts of things that, you know, I can fumble my way through still from my Excel skills, but it just happens magically in front of me.

41:27And I think that's the stuff that is most important is finding this real practical value. And Cloud Cowork's been that gateway for our company to see the real value. And I've got so many examples throughout our organization of people that are just crushing it by doing different analysis that they never could have done before, creating dashboards, creating visualizations to these aren't necessary apps is just simple stuff that I refer to as, or is referred to as knowledge work. I think that's where we're really seeing the traction. And we haven't even really kicked off our formal cloud cowork workshops and education that start this Friday.

42:01I'm excited to really upskill the employees because back to your question, the why behind this is just very clear for us. We truly believe that AI is something that can unlock and upskill our employees to be able to free up their time from the menial tasks that are involved in their job to free up their time from more value-added activities that help our customers, to help our employees, to help our culture. And we're seeing that throughout our organization. And again, I can't stress this enough, that doesn't happen with chatbots. Whether it be Cloud or OpenAI, it doesn't matter. It happens with agendic delegation tools like Cloud Cowork and Codex that I think are really the, it's really the foundation of what we're doing right now, which then leads to these dashboards can become apps.

42:47So it's sort of the gateway, the on-ramp to seeing the potential and power of AI. For us, it's been Cloud Cowork and our employees are by and large seeing it, they get it, and they're coming back to us saying, thank you. Not what are you doing? You're taking away parts of my job. And to be clear, with 360 employees, I'm sure that we have some employees that are concerned. And I'm concerned over about AI in the world, you know, but I know our company, I know our culture, I know our motivations and I know why we're doing what we're doing. And I know our founders and our leadership team and who we answer to.

43:19And we answer to our employees and our customers with the right intentions of adopting AI to really leverage this, what I think is a total arbitrage moment of intelligence right now, where the people that are using these tools have an ability to differentiate themselves and different companies have the ability to transform and differentiate themselves in new ways that never existed before. And that's what we're going to continue to do. So my long winded way of saying my advice is don't delegate this, have your own moments. I suggest Cloud Coworker Codex to do knowledge work and then let that spread organically and get more and more people creating tremendous practical value using these tools.

43:56And then from there, you've got the traction, you've got the confidence and you can start to invest. I'll say this, we wouldn't be doing a lot. We wouldn't let domain experts do some of the things that we're doing with apps if we didn't have the technical team behind them. but we wouldn't have had the courage to invest in the technical team had it not been for leaders of the company having their own moments and realizing that there's really something here that we need to be ahead of and start adopting. And I think that's my advice.

44:23Isar Meitis:Awesome. I think that's an awesome way to end this. Really, really incredible conversation. I truly appreciate you willing to come and share what you've learned with your own blood, sweat and tears in the last three years in the AI journey. I really appreciate our partnership and our friendship through this journey together, you and your team really inspire me and push me to work harder and do some of the things that I do. And I think that's maybe the biggest thing is surround yourself with people who are experimenting with this. They don't have to be coworkers. They don't have to be somebody you hire.

44:54Isar Meitis:That's not a bad thing necessarily, but it's not a necessity. Surround yourself with people who are tinker and experimenters in this space. Because like Ari said, there is no playbook. And if there is, it's changing on a weekly basis. So even if there was one, it doesn't matter. The amount of times Ari and I said, oh, no, no, no, we got to dive into this or that. And it's just changing all the time. But the concepts, like the concept of agents. So it was OpenClaw and then it was ChatGPT and then it was, you know, whatever open source tool. And then it's now it's Cloud Cowork and Hermes. And like, but the concepts and the experimentation are what drives this forward together with something that I love that you pointed.

45:34Isar Meitis:because I think is the, if you want the secret sauce, is how do you create real excitement in the organization? And it is starting small and just showing people what's possible. And then the right people get excited and then they will join and they will excite more people and that just organically will grow. And it takes time. Again, it didn't happen overnight. It took three years. But if you don't start, then you're not going to get there. If people want to follow you, learn more about what you're doing with AI or learn about the company or whatever it is, what are the best ways to do that? Sure.

46:04You're welcome to reach out to me. My email is ari, so A-R-I at Sonance, spelled right behind me here,.com. And then you can also reach our whole AI team if you have questions or just want to bounce ideas off of us, which is very similar, but ai at Sonance.com. So reach out to either one of those email addresses and we will get back to you and love to connect and learn what anyone else is doing out there. We were definitely a sponge trying to continue to make sure that we recognize that, like you said, there's no playbook. We've got plenty to learn, plenty yet to prove to ourselves. And yeah, it's been exciting.

46:39By the way, one thing I forgot to mention, if you're new and you haven't figured out skills and what skills are yet within AI ProNZ, that that word is just foreign to you, get into Cloud Coworker Codex and start building some skills. I think that is a major unlock of capability that really highlights the difference between chatting and really delegating things using skills. So those are sort of the entry points that I think are best, but reach out. We're happy to help. We're happy to talk. And thanks again for having me. Really appreciate it.

47:07Isar Meitis:Thank you. This was absolutely awesome. I appreciate it. Thanks, Cesar. Thanks for pulling our journey. I appreciate it.