Will AI be the End of SaaS? with Rob Woollen & Marwan Mattar of Sigma

(Intro Narrator at 00:00:00) Today, we're talking to Rob and Marwan from Sigma about whether or not all the advancements in AI will lead to the death of SaaS. You're listening to Joel Beasley, Modern CTO.

(Joel Beasley at 00:00:15) So I was actually pretty excited about this because this is a topic I've been thinking a lot about. Will AI be the end of SaaS? Rob, what do you think?

(Rob at 00:00:28) You know, it's been on my mind a lot as well. I think one of the things I'm just inherently fascinated by is, if you look at, say, the last fifteen years, every company would just buy up these SaaS apps. Right? It was so simple to buy something. You could just sign up, and next thing you know, your company had like a hundred of them. And there was a lot of great things about them. Right? The fact that you could just start it up so easily, you didn't have to provision hardware anymore. All of that really was a game changer from the generation before.

(Rob at 00:01:00) But if you look at where we're at now, I'm just fascinated by this idea that you can essentially dream up an application you want, ask AI to build it, and you start seeing things just appear almost instantly. And to me, that is going to be a complete game changer. Now how we actually get there securely, how we actually get there and get people unlocked with this idea, I think is going to be a huge transition. But yeah, to me, it's one of the most fascinating things going on in the AI world.

(Joel Beasley at 00:01:32) Yeah, because we'd buy the SaaS applications that had some beautifully shiny functionality that we needed. But then we had to figure out how to integrate it into our own system and bring the data and bring it back and connect it back. Now it's almost the effort of going through that entire process is slowly becoming almost equal to just having AI build it for ourselves on top of our existing dataset.

(Rob at 00:01:56) Yeah. I mean, every SaaS app you bought, the next step would be, now we've got to customize it to our company. Right? No one just runs Salesforce. They run Salesforce with a thousand things they've configured on it. And so if you think about sort of this promise with AI, instead of having to take this generic thing and customize it, it's like, just build me what I actually want. Now that is going to be a huge transition for people. Right? People don't necessarily think that way today. Right? Their first thought is, I need to go find the product that does X, and AI is going to unlock, I could just ask for what I want.

(Joel Beasley at 00:02:29) Now that's exciting. And you were at Salesforce for a while. Is that correct?

(Rob at 00:02:33) Yep, six or seven years.

(Joel Beasley at 00:02:35) Alright. How long ago was that?

(Rob at 00:02:38) See, it was 2007 to '14.

(Joel Beasley at 00:02:42) So you know Parker then?

(Rob at 00:02:44) I do.

(Joel Beasley at 00:02:45) Yeah. No, he's great. We've had Parker on the show. We also have had, I think Taher is the correct pronunciation of his name. I'm not sure, but he's commonly referred to as the father of SSL. I think he's in a security role currently at Salesforce.

(Rob at 00:02:58) I know Taher as well.

(Joel Beasley at 00:03:00) Okay, cool. Marwan, are you in the Salesforce club? Are you?

(Rob at 00:03:04) No, no, no.

(Joel Beasley at 00:03:07) What's your background, Marwan?

(Marwan at 00:03:09) Yeah. So I spent the beginning of my career in academia doing the bachelor's, master's, and PhD in computer science, and then I spent a long time in the gaming world. So I spent a number of years at EA and a number of years at Unity, building AI products for the gaming world before I slowly transitioned over to enterprise, first at Scale AI, and now it's Sigma.

(Joel Beasley at 00:03:32) But gaming's more fun, though. Right?

(Marwan at 00:03:34) Gaming is fun for sure. The culture in gaming is really interesting. It was also really nice to work in gaming but not have to build the games. Right? Just kind of be gaming adjacent. Right? So you get to help the folks that are really creative, but you're not kind of in there dealing with the challenges of building a complex game.

(Joel Beasley at 00:03:54) Do you think that AI is going to kill the SaaS model?

(Marwan at 00:03:58) I think in the long term, certainly that's a possibility. And I think, like Rob alluded to this, the idea that you can kind of dream up something and build it and not have to customize it. And not just that there's a one-to-one mapping between the SaaS apps that exist today and what you'd want to create in the future and just customize, but the idea that you can build an entire workflow that is the synthesis of what would be multiple different apps in today's world. So the idea that you're not just kind of dreaming up a SaaS app, you're also just dreaming up a solution to the business problem that you have, that's kind of what I get really excited about as well.

(Joel Beasley at 00:04:35) So you're both at Sigma. Rob, you're the co-founder, and Marwan, what's your role there?

(Marwan at 00:04:40) VP of AI.

(Joel Beasley at 00:04:42) VP of AI. Okay, awesome. And what's the main problem you two are solving?

(Rob at 00:04:47) Sure. So Sigma builds a cloud platform for AI application development and analytics. So essentially, we take the largest data sources in the world, things like Snowflake and BigQuery and Databricks, and we build an AI application and analysis platform on top of that.

(Joel Beasley at 00:05:06) Oh, so you're seeing this unfold right now?

(Rob at 00:05:09) Oh yes. We're very much seeing it and frankly trying to push it forward ourselves.

(Joel Beasley at 00:05:14) Do you think it'll be one of the things—I'm just trying to have fun here and trying to understand how this is going to roll out. Like, a lot of people, I think, would be scared that, like, oh, all the SaaS is going to go out of business. But usually the way I see these things kind of roll out is people will start using it today. Like, they've got all their SaaS tools. They've got all their data in their current spots. And then there will be new tools like yours that can connect in their data, and then they could build AI stuff on top of it. And so for new stuff going forward, they're very likely to use that because they're not having to rip and replace legacy systems or existing stuff. Are you seeing people—and so this is my question, I guess—are you seeing people replace their existing SaaS or are you seeing people just essentially solving their new problems this way?

(Rob at 00:06:00) Both. So, you know, as you'd probably imagine, and I think as Marwan alluded to, it's not that people are trying to rebuild exactly the SaaS app they already have. Right? They're thinking about sort of what would I actually want to, how do I actually want to solve this problem in an AI world? I want to build something probably that's slightly different and probably that is itself more intelligent. So you see people building a different kind of application, but a lot of it is also people that, frankly, they weren't empowered beforehand. Like, their problem they had was too specific for any general SaaS vendor to solve. But once they could actually ask AI to build it, they were now empowered to go do it themselves.

(Joel Beasley at 00:06:41) Oh, that's pretty cool. I'm actually really excited about what you guys are doing. Marwan, tell me about this demo-to-production gap in AI.

(Marwan at 00:06:51) Yeah. The way I like to think about this is if you went back to twenty years ago and you wanted to build an AI demo, you had to think very clearly about the task you want to solve. You had to go out and collect the data that would allow you to train a model for that task. You had to acquire compute. You had to set up training jobs. You had to train a model, and then you got to test that model on that one very simple task you had in mind. So the idea of building a demo back then was incredibly involved and took a lot of effort and a lot of know-how, frankly. And today, you can basically prompt a demo, right, which means the time to build a demo has gone from what used to be days and weeks to minutes and hours. The problem then with that is a demo is cool. It shows the art of the possible, but it doesn't capture the realities of the world. Right? It doesn't capture the fact that there's a lot of edge cases you have to consider. It doesn't factor in the cost and the latency and the user workflows. So it ends up being that it's very quick today to build a cool demo, but it's increasingly hard to get that demo rolled out or that product behind that demo to be rolled out and actually reliable to a large number of customers.

(Joel Beasley at 00:08:05) Is this an area that your product helps solve?

(Marwan at 00:08:09) I think we think very, very deeply about the guardrails that we want to give the systems that we built so that our customers can be successful more easily. And so the systems that we build, we try to be very mindful that our customers are going to use them in all these really complex ways. And that for them, you know, 90% accuracy is not good enough. Right? It has to be really accurate. And that goes beyond some of the work that we do with the show-your-work and, for example, products like AskSigma, where we really try to give users the tools to understand how well this is working.

(Joel Beasley at 00:08:42) I know everyone's always like, don't get too carried away with demos, and then you hand it to a founder, and they just make the sale. You know? Sometimes you've got to do that. Rob, you guys—one of the things I liked when researching your company and your background—was that you don't give up. You're a persistent person. It took a little bit to find the product-market fit for this. Can you tell me about that journey?

(Rob at 00:09:04) Yeah, absolutely. You know, I think in a lot of ways, we're a pretty unusual company in that almost from the start, we've actually been pursuing the same sort of problem. We've been trying to make it so that, essentially, this large-scale data was something that almost anyone could actually leverage. Right? Previously, it's only really been accessible to people that could write SQL or Python. And so we're unusual, I think, in that we have always been working on the same problem, but we didn't really, frankly, know exactly how to solve it when we started. And we kept trying and trying and trying different iterations. At least three times, we threw everything away and started over. Even as late as 2020 to 2021, after we already actually had a little bit of revenue, we actually rebuilt the product yet again. And, you know, those transitions in hindsight are easy, right, especially if they work out. But at the time, as you can imagine, it's super stressful trying to figure out, like, are we making the right decision? Should we actually start over? And yeah, it's been a journey and, obviously, very happy to see where we are now, but it takes a lot of persistence.

(Joel Beasley at 00:10:19) What would be one piece of advice you'd give to a founder who's thinking about giving up right now?

(Rob at 00:10:25) I think to start a company, you have to be entirely irrational. And what I mean by that is, like, any rational person would say, like, your odds of starting a company that's successful are, frankly, very, very low. You know, like, why would you ever do this? And so you almost have to have this irrational inner belief that whatever you're doing, it's going to work out. And I think throughout the journey, sort of reminding yourself of why you started the company, why you feel so passionate about it. And certainly in the low points, having other people to talk to and sort of having a co-founder, I think, is super important. There were definitely points where either my co-founder would be down and I would be up, or he would be down and I would be down and he would be up. And we definitely sort of pulled each other through some of the hard times.

(Joel Beasley at 00:11:10) Were so—co-founders, you guys are both there from the beginning? What was the moment in time where you two got together and you're like, we want to work together? We want to do this as a business?

(Rob at 00:11:22) So we were both what's called entrepreneurs in residence at Sutter Hill Ventures. So we were both hanging out at a VC firm sort of working on ideas. I didn't know Jason at all at the time and had, you know, had no plans to start a company with a random person. But at some point, the partner, Mike Speiser, that I was working with mentioned, like, hey, you know, the ideas you're working on actually seem to be very similar to what Jason's working on. And then from there, it just became, we started trying to work together. And now, what, eleven years later, we're still going strong.

(Joel Beasley at 00:11:56) Nice. And then, Marwan, what drew you away from gaming and into this Sigma world?

(Marwan at 00:12:05) Yeah. At a very personal level, I've always been really excited about building systems that are close to users. Right? So for me, the gaming world was all about how do you build the tools that let the game developers be successful in making games. And so Sigma kind of really checked that box for me, which is, you know, you've got a very large number of customers, a very large number of users, and so being able to have impact on users is something that has always been a consistent theme in everything I've done. But specifically to Sigma, I felt like the product was really strong. Right? The premise that you can go all the way from just connecting to a data warehouse and going all the way from visualization to building reports to building enterprise data apps, for me, kind of felt like it set the foundation for the building blocks you need to actually build intelligent systems that can help users go from data exploration to decision making. So it kind of felt like the right fit for me.

(Joel Beasley at 00:13:05) When building these types of systems, let's say you're mentoring or talking to an incoming class of engineers, like the next generation, and you can convey one or two core principles of systems engineering, what would they be?

(Marwan at 00:13:17) Probably the first one would be the importance to not get carried away in starting small. I think in building all these AI systems, it's really easy to want to start big and to kind of include a large number of use cases in your very first experiment. And so the idea that you can start small and start very focused, I think, is a little underrated. And then, of course, once you start there, incrementally adding use cases and adding and evolving how you solve the problem kind of goes hand in hand. The second is to be incredibly metrics-driven. One of the lessons of AI machine learning is the no-free-lunch theorem. Right? Nothing comes for free. With every improvement in accuracy, there's some trade-off somewhere. And so really understanding kind of what you're getting and giving away as you evolve the system is really important.

(Joel Beasley at 00:14:07) You're a smart guy. I like you. Rob, good choice in teammate. How did you two meet, by the way?

(Rob at 00:14:16) We had hired a recruiter actually to sort of—we had the idea last spring that we really wanted to get more kind of foundational AI expertise in the company. Right? I feel like, you know, I've been building enterprise software for, gosh, thirty years now. So I feel like there's a lot of things in enterprise software just because I've been doing it so long, I sort of almost inherently kind of have an idea of what we should do. Whereas I felt like a lot of the AI stuff was newer to me and newer to the organization. And I wanted someone with deeper experience to sort of have that breadth to advise us on kind of the right direction. And yeah, we found Marwan.

(Joel Beasley at 00:14:56) You said, go find me the best we could possibly find, and they brought you back this champion. I love it. I am curious. So, like, last year in my mind from the conversations that I've had, agentic was a big deal. Like, that was a huge buzzword and theme in the AI space last year. This year, it seems like these fleets of agents or whatever words you want to use to describe a plurality of agents working together. Are you guys seeing that happen at all?

(Rob at 00:15:26) You know, I've heard and seen the idea that you've gone from directing a prompt or a chat conversation to now I've got — I almost think of them like interns. I've got this fleet of interns that I'm gonna go direct. I think I can't decide honestly how I feel about whether it's the right direction or not. Part of me likes the idea of being able to massively — you know, I've always wanted to have a lot more time to do stuff, right?

(Rob at 00:15:58) And so you feel like, oh, if this could really unlock me, and I can see a bunch of use cases even in software development where it helps a lot. On the other hand, it still feels very human directed, right? And to me, I'm very much interested in more autonomy, right?

(Rob at 00:16:17) When is the AI gonna reach the point where instead of me having to tell the intern every little thing to go do, it really has enough intelligence that it's automating and really taking things off my plate?

(Joel Beasley at 00:16:28) Yeah. And some people are trying to get to that end by just throwing a thousand interns together in the same room and saying — dude, by the way, Rob, that's the first time I've heard the intern analogy, and it is so great because they can think, they can show up, but they're just not quite there yet.

(Rob at 00:16:45) I think I credit my co-founder, really.

(Joel Beasley at 00:16:48) Yeah. We were all early in our career. We can laugh at that, you know. Just like when you're in your twenties, you think you're so smart and you're an adult, and then you get into your thirties and then your forties, and you're like, oh, wow, I was an idiot.

(Joel Beasley at 00:16:58) I was alright. So 2026, do you think we're gonna see more of this orchestration happening, Rob?

(Rob at 00:17:06) I do. I think that you're gonna see more of people trying to manage this type of orchestration across the agents. I think you're gonna see more agent to agent workflows. So, you know, much as we've talked about all these different applications coming together, I think one of the most powerful things is if you can actually have workflows that go from agent to agent across systems. So I'm looking forward to seeing a lot more of that.

(Joel Beasley at 00:17:33) Okay. And this is a good time to bring up — you actually are having your first ever conference designed around workflows. Is that correct?

(Rob at 00:17:40) Yes. We've been waiting to have our conference and, yes, we're in March, I believe. We've been really passionate about — I think, frankly, too few people understand what AI can already do for them today. And so what I hope comes out of this conference is people really unlock their imagination of these are all the types of applications that I could build on top of the Sigma platform today and really change almost people's mindset of what's possible.

(Joel Beasley at 00:18:11) So the workflows are gonna become a bigger part. They're part of the conference on March 5, so that's the theme around it. And then what's the one takeaway someone's gonna get from this conference?

(Rob at 00:18:23) To me, it's really the art of the possible. I think when you go through these big transitions, too few people — it takes a while for people, the human side of it, to actually understand what can you do in the world today. So the technology is often ahead of where people's mindset is at. And so I hope people take away — I don't need to buy the same old SaaS apps I've been buying for ten years. I could actually change the way my company works.

(Joel Beasley at 00:18:51) And what's the pain point someone — I always like to bring this back because we have all these different listeners. They're at different points in their career. Some of them are leaders, first time managers. Some of them are VPs of engineering.

(Joel Beasley at 00:19:02) Some of them are CTOs and CIOs. Knowing that that's the three areas of people who listen to the show — what is the major pain point they're experiencing that they're like, oh, they can diagnose that they might have a need for Sigma's product to look into it today? What type of pain points are they experiencing?

(Rob at 00:19:22) So it usually comes from a few different directions. One is simply scale of data. So, you know, traditionally, people love Excel spreadsheets. Excel is like the original application development platform. People email spreadsheets around to each other.

(Rob at 00:19:37) No one calls it an application. No one calls working in Excel programming, but that's frankly what it is. But, you know, you cannot do that with any sort of large scale data. You can't do that with any data you actually have to secure and can't have just running around on people's PCs. And so by bringing this data to a centralized platform, by building the application workflow, the analysis, integrating that all into the application, that's essentially what Sigma solves for them.

(Rob at 00:20:10) And one of the things I like to point out to people is if you rebuild an existing SaaS app on the centralized data, it actually gets much better than the siloed app you had previously. Because the problem in the past would be, you know, you would have a support, for instance, app. It didn't know anything about the sales side of this. So it just gets a ticket from a customer, and it's like, I don't know whether this is the biggest customer in the world or the smallest one. Being able to actually integrate all of your data together means that every one of your applications you build on that centralized data platform actually is much better than the previous siloed version.

(Joel Beasley at 00:20:47) Nice. Marwan, you wanna add anything to that?

(Marwan at 00:20:50) Yeah. I think I mean, this touches really nicely on the earlier point around what customers really want is solving business problems, right? Not stitching together these individual SaaS apps that were all designed with 50 different use cases in mind. But the fact that you have this one central data platform, and then with Sigma being able to build your workflows on top of that, they can stitch all the way from what's happening in support, what's happening in your product, and actually get to the point where you're taking actions, which then write back to your data systems, is really where I think the power will be in the coming years.

(Joel Beasley at 00:21:24) Tell me about what AskSigma is.

(Marwan at 00:21:27) Yeah. So AskSigma is our product for answering questions of your data. So it's basically a way to simply ask natural language questions, and we return back answers based on the data that you have in your warehouse. And one of the things that it offers you also is rather than just giving you an answer, it shows you then all the steps that it took to arrive at the answer. So then you can introspect at every step, make sure that the system delivered an answer that was reasonable.

(Marwan at 00:21:58) And then from there, you could take all the elements that were created for you as part of this analysis and build them into a Sigma workbook and continue off and build workflows. So it helps — for me at least, it helps solve the problem of data discovery, data understanding, and then query generation.

(Rob at 00:22:17) Nice.

(Joel Beasley at 00:22:18) So you got AskSigma, and then what other main parts of the application do you have? You have an AI builder, got some MCP stuff.

(Marwan at 00:22:25) Yeah. So AI builder is what Rob alluded to in the beginning. It's a way to build entire apps in Sigma all through natural language. And there's a couple of things that are really cool here, right?

(Marwan at 00:22:36) So the first is you can get started very quickly by just prompting the system to build what you want for it. And what you get back is a set of Sigma assets that have been built within the Sigma platform. So not a bunch of code that you then have to understand, and it's all built within the governance that already exists within your Sigma platform and your underlying data warehouse. And then you can — it's kind of like your Copilot. You can prompt it, but then you can also go and manipulate it the way you normally do it manually by moving things around.

(Joel Beasley at 00:23:07) Interesting. So let's say I've got a thousand episodes of this podcast. There's always these moments where, like, oh, who told me that? I know they had twins or three kids or whatever. We have them all in one giant Dropbox.

(Joel Beasley at 00:23:19) And the Dropbox chat is not that smart. It's mediocre at — I don't know if you've used it, but it's okay. Is there a way I could take all that data and put it into Sigma, and then I could talk to AskSigma about all of my past episodes?

(Rob at 00:23:37) You can. We have the facility to integrate with things that read unstructured data like that. I would say where we really specialize, though, is more on the structured data side. Imagine you were bringing in a bunch of tables from a SaaS system that tracked your audience and the engagement of how many people watch each episode. Being able to actually do data analysis on those types of things is probably where we would say we were more specialized.

(Joel Beasley at 00:24:07) Okay. So you guys specialize more in the structured datasets and then insights and transformation of those datasets?

(Rob at 00:24:15) Right. All the way from analysis to if you then wanted to, let's say, build — let's say you had sponsors and you wanted to build a partner application with them where you want to expose perhaps the data to them or make it so that you could build workflow on top of that. Every time maybe a new subscriber shows up to your podcast, you wanna do something — that type of thing, you would build that all in Sigma.

(Joel Beasley at 00:24:37) Oh, okay. That's pretty cool. Are there any new features that are coming that you're excited about that you're allowed to talk about?

(Rob at 00:24:46) I'm probably most excited about the one that Marwan just mentioned, which is this idea of going from natural language to be able to construct applications on our platform. I think, you know, as we alluded to earlier, the idea that the cost of building an application being driven to zero means that you don't need to ask permission, you don't need to worry too much if you try to build something and it doesn't work. You only wasted a few minutes of your time. I think one of the biggest problems we have today is that most ideas just die because there's too much friction, right? I don't want to go open a ticket with IT.

(Rob at 00:25:24) I don't want to go talk to some vendor. If you can get rid of all that friction, then a lot of the creativity that people have can very quickly be realized, and that's what I'm really excited to see.

(Joel Beasley at 00:25:36) Yes. Just to have the idea and be able to go do it. I don't have to go hunt the vendor, negotiate the contract, go through months and months of all this stuff, negotiate the support situation, get approval from all the security people on both sides. I can just — I need the answer to this question. I can invoke this AI to help me achieve it.

(Rob at 00:25:58) Yeah. And part of the fun of building this sort of general purpose data platform like we do is that, you know, people do all sorts of things with it that we frankly never imagined. And so seeing — I'm especially interested now — unlock, make it even cheaper for them to go do things. I'm really curious to see what they're gonna go do now.

(Joel Beasley at 00:26:16) Yeah. It's very exciting to watch this. Now do you have any — well, here's a broader question. Have you guys gotten to the point where you've developed any sort of insights or reports on what you're seeing happen across your customer base? Novel use cases and things like that?

(Rob at 00:26:35) I mean, there's definitely — I would say there's combination of the down the middle use cases, right? What are the most common things people do with our platform? You know, given our data background, it's probably not surprising that we have a lot of penetration into financial institutions, people doing analysis on everything from their internal finances to sharing with their customers. Also, any sort of supply chain, any sort of operations like sales operations, marketing operations, anything where there's large scale data where you have to make important decisions, we tend to get those as the, I would say, the common use cases.

(Rob at 00:27:16) I will say sometimes it's amusing to have the ones that are the less common use cases, like these novel things people do. There's one customer that is, I believe, into poultry. They manage chicken — breeding chickens, and they built essentially an ERP system for poultry breeding on our platform, which was not on my — yeah, it was not on my top list of things people are gonna go do with Sigma, but it's awesome to see them build that out.

(Joel Beasley at 00:27:47) Kick that over to marketing. That's what people wanna see.

(Rob at 00:27:51) I knew that would be a popular one here.

(Joel Beasley at 00:27:53) Right, right. That's because it's so interesting. The hard thing in marketing in the B2B world, you know, is how do you stand out in a way that's edgy enough but still technically appropriate, you know?

(Rob at 00:28:04) Well, and it speaks to this point earlier about there's probably not a lot of people out there saying, hey, I'm gonna go build a SaaS app that specializes in poultry breeding. And so instead of having to hope you find some vendor that does a poor job of it, you can build the thing yourself.

(Joel Beasley at 00:28:22) Oh, gosh. The graphics and the potential namings are running through my head. It is a marketing dream.

(Rob at 00:28:31) Somehow I knew it was gonna spark a smile.

(Joel Beasley at 00:28:34) It is. It does. It's hilarious. It's amazing. Let's do some leadership conversations.

(Joel Beasley at 00:28:40) Lot of reason why people listen, become better leaders. Rob, the first time that you were a manager of people other than yourself, going from individual contributor to first time leader, tell me about that.

(Rob at 00:28:54) Oh, I had a pretty shocking transition. So I was a CTO of a group at Salesforce, CTO of the platform. And on our product management team, we didn't have a leader of product management. And it was becoming, frankly, an increasing problem in that group. And so after a while, the executive in charge asked me, like, do you wanna take on this role?

(Rob at 00:29:22) And so it was the first time I was managing people. It's the first time I'd ever had a product management title. And so I jumped from being essentially a CTO with a pure engineering background to running product management and managing then, you know, a bunch of product management leaders. So it was a little bit of a trial by fire. I think in retrospect, it was hugely valuable for me to get that experience once I got to Sigma, because it really let me see not just the technology side, but also interface with customers and see a little more of the business side.

(Rob at 00:29:58) But it was a big transition.

(Joel Beasley at 00:30:01) You just jumped right in. Did you like it? You could do it, right? You were able to do it, but did you like it?

(Rob at 00:30:08) Yes and no. After about a year and a half, two years, I missed writing code, and I missed being on the technology side. So I actually went back and was the CTO again.

(Joel Beasley at 00:30:21) I know what you're talking about. Once this show got popular, I stopped — I did hands on software engineering for seventeen years. And then the first three years of the show, I was still doing it. And then the last five years, I've just kind of — I still manage two production apps, but it's not an everyday thing.

(Joel Beasley at 00:30:37) I only need to poke my head in there once every couple weeks, take a look at it. But I do kind of miss it. And sometimes I feel like a fraud too when I'm doing the show. My god, I'm not doing it all day every — but there's a lot more than just hands on keyboard stuff. Marwan, what about you?

(Joel Beasley at 00:30:53) Tell me about the first time you became a leader.

(Marwan at 00:30:55) Yeah, it was definitely much softer transition than Rob's. This was, you know, at both EA and Unity, my role was basically starting up new AI teams. And so in both situations, over time, I was kind of the natural person to start managing that team. So it was always, like, you know, the transition was more of managing folks that were previously my peers.

(Joel Beasley at 00:31:22) Nice. Nice. And then, Rob, if you go back and talk to yourself, like, right when you make that transition, what's one thing that you had a time machine, you go back, what's one thing you would tell yourself?

(Rob at 00:31:33) That's a great question. I think one of the big lessons I've learned over the years is just the importance of having the right team. And so, you know, frankly, a lot of things at work go very well when you have the right people, no matter what you do as a manager. And when you have the wrong team, like, there's very little you could do if you're the best manager in the world to make it great. And so I think, you know, in particular at that time, I had a couple of great people and a couple of people that probably were not the right fit.

(Rob at 00:32:00) And, you know, making those decisions sooner probably is what I would have advised myself in the past.

(Joel Beasley at 00:32:07) Oh, yeah. That's, like, the key first-time leader mistake until the pain becomes, you know, you watch that pain play out, and the pain of not letting them go becomes so great that you just have to. It's like, yeah, that's a tough one to learn. You know? Marwan, what about you?

(Marwan at 00:32:27) My instinct is those are kind of similar as who you bring on the team is the most important decision you make. Like, I always feel like the P0 of your job is just who you're hiring onto the team—both cultural fit, technical background, appetite for the type of work that you do. Right? There's nothing that beats having the right people that are incredibly motivated at rest state, right, and not feel like you have to constantly kind of get them to be motivated. I think prioritizing that has probably been the biggest value add.

(Joel Beasley at 00:33:01) I agree. One of the early leadership things that I was told was, you can take people who are running and point them in the right direction, but it's really hard to get people to run. And so, yeah, finding those teammates that are, like, already running and they're trying, and then you can just help coach and nudge them in the right area versus trying to wake up every day and get someone to start moving—that's a sad existence. All right. Hiring advice.

(Joel Beasley at 00:33:30) Now, Rob, I like behaviors, human behaviors. You have direct reports on your team. You have a handful of them. When you're looking at your direct reports, you obviously can't invest equally into all of them. That's just not possible.

(Joel Beasley at 00:33:45) What are some of the behaviors that you see a direct report doing that you think to yourself, Rob, you're like, you know what? I'm gonna spend some extra time with that person. I'm gonna pull them aside because I think they have real potential here.

(Marwan at 00:33:57) I'm smirking. Can't talk about me, though.

(Rob at 00:34:00) I was smirking because Marwan is one of my direct reports.

(Joel Beasley at 00:34:04) Oh, is he?

(Rob at 00:34:07) Oh. I'm probably an interesting case, you know, in that I admit even to people that I manage that, like, my passion is not people management. I only directly manage three people. And as part of that, I frankly focus on people that are, you know, senior enough that I try to be a resource to them, you know, when they're in need. But, you know, if someone actually needed to be, like, actively managed, they're too junior for, honestly, me to manage them.

(Rob at 00:34:40) Like, I'm not particularly skilled at it nor, like, is it obviously my passion. I do love people development. Like, I love giving people advice about, you know, what I think they should do or what I think they could do with their career. But the day-to-day, I feel like my passions are much more on, like, building the product. And so I try to sort of balance that and be very clear and make hiring decisions. You know, Marwan has relatively low expectations of how much management he's gonna get from me.

(Rob at 00:35:12) On the other hand, he doesn't need it.

(Joel Beasley at 00:35:15) So the behavior that would stand out is independent ability to execute and achieve outcomes. That's who you want.

(Rob at 00:35:21) That's a much more positive spin on it.

(Joel Beasley at 00:35:23) Yeah. Well, that's what we do here.

(Marwan at 00:35:25) Yeah. Rob's slowly underselling himself as a manager here in terms of interest for sure, but abilities are there for sure too.

(Joel Beasley at 00:35:34) So, Marwan, your team—you manage people directly as well?

(Marwan at 00:35:38) Yeah.

(Joel Beasley at 00:35:39) So what behaviors stand out to you when you see them happening where you're like, I'm gonna spend some extra time with this person and help mature them up? Rob's like, I'm only dealing with mature people. You've gotta help them make that next step, Marwan.

(Marwan at 00:35:51) The one that stands out for me that I feel like I kind of maybe take, I view as important, even though I feel like typically it doesn't get the importance it deserves, is folks that make it challenging to have a collaborative environment. Right? The folks that, like, you know, feel the need to be the voice in the room or don't invite everyone in the room to, like, have an opinion. And I say that because one of my kind of core tenets is that to solve these really hard problems we're trying to solve, you need everyone to kind of contribute to this. And everyone's opinion has to be important, and everyone has to have a perspective on this.

(Marwan at 00:36:34) And that's the only way we're really gonna be able to solve these and build these really complex systems we're trying to build. So I really am kind of very sensitive to folks in the room that kind of mute some of that. So that's one of the behaviors I find myself reacting to more strongly than maybe I would otherwise.

(Joel Beasley at 00:36:51) Yeah. So that's a good negative behavior that would be like a huge red flag. What's a—because what I have found in my history is that you need the team to be more focused on the outcome than the individual. If we're all in agreement, it's like the Navy SEALs. Right?

(Joel Beasley at 00:37:07) The idea is to get the whole team to complete the mission successfully at sacrifice of yourself, not to make it all about yourself. So that's a red flag. I love that. It's a good red flag. What's a huge positive? What's a green flag?

(Marwan at 00:37:22) Yeah. It's usually the folks that are—I find a big delta between, like, how good they are in, like, one-on-ones versus how much they contribute in kind of bigger team meetings. Right? So those are folks that I feel like are incredibly smart, have a lot of really good ideas, and I really want them to grow and feel confident to share that more openly.

(Joel Beasley at 00:37:46) Okay. So if I'm your direct report and I've got brilliant ideas when we're one-on-one and we're together, I speak clearly, I have value to the customer. Like, I'm saying all the right things, but then if I'm quieter in a bigger team meeting, you're gonna work and help improve that person to express that in the team meetings. That's a good thing. I like it.

(Joel Beasley at 00:38:05) You guys have a good team at Sigma. Come on now. As we start to wrap up, I just want you guys to share, like, one thing that you want this audience to know. So, again, the audience is CTOs, CIOs, VPs of engineering at tech companies of all shapes and sizes. We're gonna plug Sigma Computing's website. We're gonna put everything in the show notes. But what's the one parting thought that you want them to have? Marwan, I'll do you first and then Rob second.

(Marwan at 00:38:33) The biggest one for me is when you're thinking about AI systems, it's really important to be very clear on the use cases and to push your product team to be very clear on what the evaluations of success are.

(Joel Beasley at 00:38:48) Excellent. And, Rob?

(Rob at 00:38:50) I think as a CTO, the job is gonna change a bunch over the next few years in that, you know, we're clearly seeing software development change. Right? The, you know, the amount that an LLM can do for me, the amount that an agent can do for me—like, no matter, you know, they're maybe not perfect yet, but they're certainly gonna change software development pretty dramatically. And so I think for any CTO, it's really about a combination of being pragmatic.

(Rob at 00:39:21) Nothing changes overnight, but things definitely change dramatically in three to four years. And I think, you know, making sure your company is ready for that, making sure you're looking forward, you know, have a plan for that—that's frankly all of our jobs.

(Joel Beasley at 00:39:37) I love it. Sigmacomputing.com. Is that correct?

(Rob at 00:39:40) That is correct.

(Joel Beasley at 00:39:41) Is that the correct website? All right. We did it, guys. Rob, Marwan, we made a podcast. How do you feel?

(Rob at 00:39:46) I feel great. Thanks for having us.

(Joel Beasley at 00:39:48) 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].

(Joel Beasley at 00:40:06) Every time I get an email or LinkedIn message, it absolutely makes my day and inspires me to keep going.