Why the Best CMOs Are Becoming Builders with Mike Maziarz of MindBridge

Mike Maziarz 0:00
I never thought I was going to have to completely reboot my skill set. But, you know, Claude Code came out, Opus 4.6 kind of turned on. I took one look at that and I said, oh my God. Like I've got to go back to school, like mentally, and I've got to rebuild my skill set from the ground up. And I've got to invest hours and hours a day into learning this new tech stack and what's possible from it. Hello, everyone.

Kevin Kerner 0:29
This is Kevin Kerner with Tech Marketing Rewired. I'm your host. And today I'm talking with Mike Masaris, who is the chief product and marketing officer at Mindbridge. Mindbridge is a really cool company. They use AI to find errors and fraud in financial transactions. And Mike runs both the product and the marketing team behind it. It's kind of a unique combination, a pretty rare seat to have both in one person. What I find amazing about Mike, and in our conversation, we talk a lot about this, is a CMO who actually builds, he's taught himself agent architecture. He was an engineer by trade. And he built a system called Tropo, which is an operating system for AI agents. Right now it's about 5,000 files now, and his agents wrote every piece of the application. So it's really cool and pretty incredible to see a CMO that's so active in building an AI. So I'm really excited for this one. I think you'll enjoy it. Let's get to it. This is Tech Marketing Rewired. Mike, welcome to the podcast. So glad to have you on. Kevin, thanks for having me. I'm glad to be here. Yeah. Really, really appreciate it. I know you're quite busy in the product and marketing role there that you do at Mindbridge. I want to talk a little bit about that. Maybe to get started, you could just give a quick bio and then a little bit about what MindBridge does because it's really fascinating. Do I have to tell you my age? Uh no. I'm pretty sure I'm older than you.

Mike Maziarz 1:58
So okay, okay, okay. I guess we don't need to be able to do that.

Kevin Kerner 2:01
We'll talk afterwards.

Mike Maziarz 2:02
Sure. I started my career in technology. Um, I started as an engineer working for consulting and systems integration companies. It's kind of where I got my passion for technology. Uh the thing is, I wasn't a great engineer. I was an okay engineer. I was really drawn more towards business and process. And so I entered uh kind of the software project management track. And I took my career there. Did that uh before the dot com bubble even popped. Um eventually, when in the heart of the dot-com era, I became a product manager for the first time. Uh that was at the end of the 90s, and I loved product management. I loved getting my hands on with the product and the customers. After that, I took my career in the direction more of product marketing. I really did kind of like the front office function, getting involved in the sales process and the pursuits and kind of positioning of the products that I had. From there, I went and started my first business, which I ran for seven years. It was an awesome journey. Um, took that up to 35 people. Uh, it was a security education company called Safe Light Security. Uh, ran that for seven years and sold it to a competitor uh that loved our business. And uh it was time to get off the treadmill of of running a small, a small business after seven years. And after that, after I sold that, I really kind of firmly put my um career into product management and marketing. So I uh became a CMO and VP of product at the same time. So I was like kind of the first time I took on that um dual role as head of product and marketing. I then, you know, we then sold that company. I went on to just focus on being a CMO uh for a couple of years, started another company in data analytics, AI-based uh data analytics and data integration, and then ended up here at Mindbridge. So it's it's just like a lot of tech tech journeys. It just happens to have a lot of years in it, but we we kind of um roll with the flow over time and it just you know the knowledge accumulates and continue to have fun.

Kevin Kerner 4:08
And MindBridge is an AI native uh product that focuses on the financial industry, right? It's kind of above all of the um activity that's happening at a company and it senses uh things that might be at risk, put you at risk, correct?

Mike Maziarz 4:22
That's absolutely right. Yeah, we provide like an independent layer of oversight. So we look at all financial transactions on a continuous basis. And so we're looking at full population monitoring, we're looking for errors, uh, fraud, or even just insights that we can get in the data. The key piece is we operate outside the system of execution. So we're not within the ERP, we're outside and independent of those. So what we do is we provide uh a layer of governance for CFOs, and that was important. It's always been important to kind of have that layer of insight and oversight, but it's even more so now as agents and uh agentic workflows start to come online and new agentic features start showing up uh in the ERP systems. Yeah, so CFOs, I mean, they have a real challenge. They have to sign off on their numbers every single quarter, right? And uh for public companies, it's a fiduciary responsibility. Um, they could easily uh lose their job, or even worse, they could go to prison. Yeah, they could go to prison uh if it's materially uh bad. So we do provide that kind of independent layer to make sure that doesn't happen. On the auditor side, that's the other market we serve. So we serve the uh audit firms as well. And so they embed Mindbridge into their audit methodology. So we become an audit analytics layer that kind of gives their auditors um kind of superpowers, if you will. They can look at all transactions in a in a data set versus just doing that kind of traditional, traditional manual sampling.

Kevin Kerner 5:49
Yeah, so cool. I mean, it's uh the fun I could talk all day long about the fintech side of AI. There's so many platforms, especially as a small business that that we could use that that need that do things for us financially, agents that work on our financial systems, but that security above it, it's always tricky because you just don't know what you can trust right now with AI stuff. Um, I was gonna ask you, or it's kind of interesting that you're both a CMO and a CPO product. You have both these product roles. I've talked to a lot of CMOs that are in this kind of healthy, I don't know, competition or d discussion or I don't know what you call it with the chief product officer. Um and you know, have that relationship needs to work really well. And you're that, you're both, like you're both product and marketing. What's different because you own both things?

Mike Maziarz 6:36
Is it is it's yeah, I mean, honestly, I've got that same competition on this shoulder and that shoulder.

Kevin Kerner 6:40
So it's just yeah, right.

Mike Maziarz 6:42
I still have those arguments in my head, Kevin. Um you know, and and in there, you know, that's really a shame though. It really shouldn't ever be competitive, right? Because we're all should be aligned in an organization. I know it's easier said than done. There's always competing factors at um in firms and different ideas and opinions, but it's those differences and opinions that end up, you know, it drives um, it just drives confusion, it adds extra work, um, you sell things that you shouldn't be selling, and you have poor implementation of products. So the better you can be aligned, um, it's just the smoother everything is for everyone, customers and uh and employees. Um so back to your original question, you know, like what makes it, I think it was what what makes it work?

Kevin Kerner 7:26
Yeah. Yeah. Or what I I mean, you see you see it from both sides. Um it's certainly different now that you control both things, but so you can you could tell a CMO, like, if you're a CMO with a CPO, like these are the things you need to do to align because you're doing it in this the same role, right?

Mike Maziarz 7:43
Yeah, I think the biggest thing that's changed uh in having both of these roles is is really some perspective for sure. But I like to take it back outside a product and I really do think of it, we bring this, we ground our company into what we call P0. So P zero is a rallying cry here at Mindbridge, and that stands for priority zero, and that's our customer and customer success, right? The value that they receive from the product. So everything we do emanates with that, absolutely everything. And so if you kind of zoom back out to like the customer situation, the people that are actually using your product and trying to get value at it, and then look at the rest of the business like as narrowly as possible, like just like a laser beam through a series of hoops. Like that's how we want to take it. We want to take that customer's view and perspective and line it right up into your go-to-market, right? So it has to go through what is required in order to make a customer success and implementation successful, what's required for the product and what we need to do for our roadmap, how are we going to sell and uh position it in the sales pipeline? And how are we gonna mark it in in the market? And if you get that line alignment up, like very, very crisp, crisply, like a laser beam, it just makes everything so much smoother.

Kevin Kerner 9:00
Yeah. So for if you're gonna give advice to a CMO that's working with a product, like how do you do it? What are the critical discussions you need to have? Where do they start? Like, how do you make that happen?

Mike Maziarz 9:10
Yeah, I would start with customer. I would start with P0. So what we've done is we're implementing a framework, um jobs to be done. It's it's an it's an old framework, but it's really effective. So we're taking everything we do, everything um that a cut every bit of value that we want a customer to receive from the product, uh, we instantiate it in a jobs to be done, and then everything aligns behind that. And so that should allow you to like really have a conversation with any department head uh in the company and align it with your CEO. And it really, really is fantastic. So if, for instance, you can get together and have a kind of a governance committee around what are the jobs to be done that we want to solve for in a company? You start with that. Like that's what we want to do, that's who we want to be, that's what we want to solve for. And if you get agreement on that, it makes it really easy when things come in from outside of the product organization to say, hey, we want these features, or can we do this, or can we take the product in this direction? Well, let's see. Does that help to solve one of our customers' JTBDs? If no, you kind of have to open up a whole new thread. And so I I've always found also that salespeople, sales organizations, and marketing teams want to do the right thing. They want to sell a product that's going to be successful, but they don't always have the crisp certainty of knowing exactly what they should be selling and and um how they should be positioning it. And I find it's particularly challenging in some of these denser, complex B2B uh enterprise B2B uh applications and platforms like MindBridges, because the the people doing the work, the engineering team, the product managers, haven't necessarily lived in the shoes of those customers. Uh versus some product teams and software, you know, you're using it every single day. Um, if you were a, you know, if you're developing software that's consumer software or something for individual users, you can kind of experience it every single day.

Kevin Kerner 11:09
Yeah, I bet the you said the word feature. So I would guess that that's the if you're doing it wrong, you're probably having a lot of discussions about feature. You know, what is the next feature? When are we going to launch the feature? How do you get to that jobs to be done? Like, have you done things to get the actual jobs to be done for the customer? Is it voice of the customer or research? Like, how do you really nail that down?

Mike Maziarz 11:30
Well, we started what we did, honestly, to start is we started by mapping out what we actually did very well already. So we basically took an inventory. We've got um we've got a matrix of job jobs to be done. It's hundreds of JTBDs.

Kevin Kerner 11:45
Wow.

Mike Maziarz 11:46
That that would be used within a finance for you know, finance organization in an audit firm. There's actually hundreds for both finance and then another 150 for uh the audit firm itself. And then what we did is we kind of narrowed it down. Say, okay, we don't do that particular job. We don't really solve for this. So we narrowed that focus down and then we mapped it out against our roadmap. And so we have a much more narrow, crisp list. And I I find it interesting because what we've been what we've been doing is re we've been rethinking how we operate. We've been, you know, our our CEO Les wants us to be AI native, right? Like everything, like let's rethink how we shape our business. Even as a mature organization, we have to think uh about adaptability and being agile and how we're gonna operate into the future. What we're doing with our product designers that used to be very, very much focused on user experience and you know, you can imagine layouts of screens, colors, um, the experience that the user receives in in navigating through the app. And we've put them responsible for jobs to be done and making sure that they they know absolutely everything about what's going to drive value for that user. And it's just a it's just an eye-opening experience for them because now they can be experts on that. Uh and that's that's how you ground, that's how you get ground the good product design.

Kevin Kerner 13:06
Yeah, so good. So yeah, I I focus on the customer. It's just the age-old thing that you're supposed to do. And that's it's so hard to do. But it's so simple, yeah, right. Uh, I wanted to ask you too. I went I was on your site and I watched the little, there's an intro video of what you guys do, and it talks about the pressure. If you are a um uh CFO today of of from the board to try to do all the AIs, like and you I think you even mentioned it there earlier. And it seems like if you when you talk to CMOs, the board and exec team also has a lot of pressure on the CMO to do the same thing. So there's some, there's some there's a lot of uh pressure, expectation, I think, from executive teams to move AI into the organization. Um when you're as a CMO and maybe CPO too, how do you think about uh the build versus buy? What's the strategy to make AI work inside your organization? Like how how do you get there without putting the team at risk? So like just like CFO would be thinking.

Mike Maziarz 14:08
Yeah, yeah, it is it is a there is a lot of pressure and there is a big challenge to do this, right? Um first of all, you got to have perspective, right? This is not a let's buy some technology and rule it out, and then it'll magically we'll magically be AI, right? AI native. It's it that's not how it works at all. So the way we've been thinking about it is very much a journey, right? And incremental and looking for the wins that we can build upon. And uh it's particularly challenging, I think, on the marketing side because we're dealing with um customer data, we're dealing with prospect data, and you know, we kind of get uh tied up in in privacy laws and regulation and access to connecting systems like our CRM into the BDRs and automating those types of things. So it it what we're trying to do is not um let those roadblocks get in the way. We're looking at the things that you know we can do that are safe and add a lot of value right away. Um the other thing I'd say that um is very interesting is many companies right now are thinking about this kind of from a process perspective. It's being driven from the top down. Like, what's that first workflow you're gonna automate? Like, how are you gonna do this? Um, what's our token budget? And they're they're kind of thinking at the solution level. But I've come to find out we have to really work hard on the education uh pieces as well. Every single person out there, every marketer out there, every product expert out there, everyone in a software or technology company these days needs to take self-education seriously. Like, I'm of the I never thought, I never thought I was going to have to completely reboot my skill set. But, you know, Claude Code uh came out, Opus 4.6 kind of turned on. I think in February, I took one look at that and I said, oh my God, like I've got to go back to school, like mentally, and I've got to rebuild my skill set from the ground up. I've got to invest hours and hours a day into learning this new tech stack and what's possible from it. And I think if you shortchange that process and or think that some sort of technology or process or some tool your company's gonna put in, you're really missing out on a fundamental um shift in learning this kind of new kind of framework, if you will. Everything is going to be built right now on top of these kind of the lowest level concepts of, you know, um of like markdown files and file systems and data organization and rethinking like you're working with um new coworkers and they happen to be agents. So like you have to think completely differently about how you're interacting with them and integrating them into your day-to-day workflow.

Kevin Kerner 16:57
Yeah, it's interesting you mentioned the coworker thing, because I do think, and this is just starting to harden for me as we're building agents and stuff on this side, is that you is initially you start thinking about uh AI like a software program, like you would normally, like a workflow. But when you make the switch to think about it as an actual worker, an augment augmentation to an existing worker. And this worker might be a um checks and balances worker, and this one might be a uh, you know, some sort of uh creative worker, et cetera. They they're augmented digital workers that can work with your your team. It's different than building software. It's like building a job description. And um helps uh that just that framing of like, okay, we're not buying software here, we're actually building uh these digital workers. I don't know if you see it the same way.

Mike Maziarz 17:47
I 100% see it the same way. And I think the difference is you really do have to start with yourself. Uh, there is a lot of focus at my company and other companies. How do we get these workflows integrated, these person-to-person workflows? For me, I've been focusing more on how do I get multiple uh agentic workers working together on different pieces of work. So I don't have to spend as much time uh transferring that context between them. How do I I've basically changed like as much as I can possibly consume and write and store, I do in markdown files right now because it's easily consumed by the agents. Um, they're all tagged, uh, they all have unique identifiers, they all have uh metadata in all of these files, so the agents can find them very easily. And I think if it's organized in that way, you're gonna see the value and the productivity, you know, really go up. And it isn't like it is a genuine, it is a genuine learning process and an evolution. My thought process on it, my techniques have changed over the last seven months dramatically. And um just in terms of how you think about them, what they can do, what they can't do, you know, where you can, you know, let them loose, and and when you really have to, um you have to be very active and be like kind of not human in the loop, but you know, human in the lead, so to speak. Which is a good thing.

Kevin Kerner 19:11
I'm definitely gonna ask you a few questions about like what works and what doesn't for sure. But I know the reason why I was asking you about build versus buy is you're a very unique senior individual, uh CMO, in that you've really gone down the rabbit hole on on uh what you're building. You built some cool stuff. Um uh uh I have been looking into ways to um build knowledge you know sets and looking into obsidian and all those type of things. You build a tool called uh uh Tropo, which I wonder if you could tell us about like how did you get to the idea? How did you learn to do that? I mean, you're an engineer too, but how did you learn to do this? Um then what does it maybe talk about the core thing that you built? Because I think it's really interesting for CMOs to hear how far you've gone into into actually building stuff.

Mike Maziarz 19:58
All right. You want a story then? Yeah, yeah. Let's let's go. We have to back up, we have to back up uh to the beginning of the year. So every year, you know, January 1st really starts in December. I really like to take a very fresh um look at our positioning and messaging, you know, what's working well, what's not working well, what's changed in our product roadmap, and what's changed in the market. And when I looked at those four dimensions of our positioning and messaging this year, a lot changed. Our roadmap was solid, you know, a number of things we could have improved in terms of our positioning. But what radically changed was this concept of AI. Um, it went rapidly, rapidly in January to, oh my gosh, we can really imagine that AI and agentic concepts and technologies and workflows will be entering the office of finance faster than almost anyone thought. Like it was like a five or six year like like ramp that like Gardner was predicting, like uh in mid-last year. And by January of this year, that accelerated dramatically. People started, they started getting on board. They're like, at a minimum, we have to start investigating, we have to start looking at us what happens when agents or start touching our transactions and start touching our ledgers. So it changed very radically. And that allowed us to rethink our positioning. So we went from being kind of an auditing solution where you could look at a bunch of transactions after the fact and you use our AI to audit or to look for errors, anomalies, and fraud to a requirement for this kind of new and necessary layer of oversight that you had described, it sits above and across all these financial uh systems of execution and kind of watches the stream of all this data flowing through and it's looking for you know, errors over there. That might be off over there, that might be possible fraud over there, and doing that on a continuous basis. That was never really necessary to be fluid and continuous. Before. But with the speed of agenc workflows and AIs coming into play, CFOs are now seeing the need for governance that they didn't have to have before. And that governance actually isn't a constraint. Gardner says that governance allows the best firms to accelerate. And that's what makes for breakaway companies, is if they get the governance right. And so we're uniquely positioned now to be able to provide that independent layer of oversight. So that's that's just the setup for the stope to your question. But what I had to do is really rethink our messaging and positioning. So first of all, of course, I used AI to do this. Um as a longtime marketer, I found it genuinely rewarding to use um ChatGPT as my uh my kind of my marketing assistant to go out, research frameworks for messaging and positioning, refine it like over and over and over again, and work with me like a partner. And I didn't have to hand it over to my whole team. I actually did most of it. Um I mean, of course the team helped, but I did so much of it with this AI partner. Okay. So, but the result was the result was okay, here's my 30-page beautiful messaging framework that is tight. It's got elevator pitch, hundred-word pitch, it's got positioning in it, it's got different um different messaging for different markets, and it's got a like, you know, messaging ladder that's going to evolve over time. So it's beautiful. And so then I was like, okay, well, I've got this messaging. Like uh a time old um challenge for marketers is how do you roll that out to the rest of your company?

Kevin Kerner 23:40
Comedy do it. And here you go.

Mike Maziarz 23:41
How do you how do you do enablement? How do you get the salespeople saying this? How do you get the product people understanding this messaging? And it's really, really hard. Usually you got to go and you got to do presentations and you got to schedule presentations, and um you roll it out and people understand 3% of what you're trying to say. And then they're constantly asking for a bespoke piece of content for their particular customer or their particular industry, and it's it's a lot of work, it's a it's a challenge. So I had a little, I just I honestly had an epiphany. I don't usually say that a lot, but I had an epiphany. I said, well, what if I just loaded my messaging framework, you know, into my prompt? I just start asking questions like do a battle card. Like, okay, I want a battle card against this competitor, or how do we relate to this technology? And I'm like, oh my God, this is really good. Like these, like using an agent, especially with in research mode, where it can go in and pull out additional web resources along with my messaging document was powerful. So I said, Oh wow, what if I pulled in my positioning framework as well, which had a lot of you know good, meaty things around our positioning and why we did it and competitive. And so I pulled that in and it was even better. So that is called, you know, context, right? We all know that. That's the context that you feed into an agent, and and this is not remarkably new. But this was back in February, you know, like January, February. I'm like, this is working really well. The challenge happened when you start pulling too much of this context into one window. It kind of fills it up, an agent gets confused, or it certainly gets loaded up. It can, it can only take on so much messaging. A 30-page document's a lot, add the positioning in it, add some quotes or some market research, and it becomes a lot. So I started thinking about okay, how can I compress this? How can I make this agent consumable, not human consumable? And uh, long story short, I just started using the tools to develop a little web app, you know. So it's a a server-based web app with a little bit of a front-end, front end. It ended up building what's called a rag-based workflow, which is retrieval augmented generation. And what that meant is it would take my prompt, what I'm trying to do, a battle card, and it would say, okay, it was like it pre-processes that prop processes that prompt. It says, Oh, okay, what context do I need in order to fulfill Mike's uh request? Oh, I need positioning. He's looking for a battle card. So it'll read in just the positioning pieces that it needs from this context library, which is all indexed, right? So it pulls in just the pieces of context it needs, then it goes out to the web and it does its competitive searches and it brings it back to the smarter LLM and it pulls it together and it puts it into a competitive battle card. And that worked fantastic. It worked out so well. We called it the Mind Bridge operating system, and we started rolling it out um to people in the company. And I will tell you, it is an absolute, it's been an absolute, absolute game changer.

Kevin Kerner 26:44
Yeah.

Mike Maziarz 26:45
Um, it was a challenge even getting people to use it, but honestly, you don't actually have to create one-off battle cards. Like everything's ephemeral. Like you just ask it what you need. The person in the field can ask it the question it needs. I actually asked it. I asked Mo right before this meeting. I'm like, I'm going into a podcast. I'm sure the host, Kevin, is going to ask me what MindBridge does. Give give give me a natural human language uh pitch. And it, I'm I'm just like, I did it. Like it was such it was so natural. I could have read it out loud. And uh it just sounds so smooth. So that that's kind of how I got into the tools, um, Kevin. But from that experience and rolling it out, that's when I decided I was gonna go all in, get in deep with Claude Code, really try to understand it, build out an entire agentic, um, like an agentic operating system, if you will. This is where your agents can live and they have a a place to store their work. Um, they have um templates that they can use, they have an event uh event messaging system so they can all communicate with each other on the file system um without using even Slack. It's like basically like Slack for agents, uh JIRA for agents, and Notion for agents, all in one. And I did that one for my own productivity, uh, but two to learn. And and I think it really did help me a lot to just go in and dive in deep.

Kevin Kerner 28:12
It's amazing. It's amazing that you did that. And I see how you kind of got to it. I'm gonna show a picture of the uh uh graph here, Tropo. This is Tropo, your application. Can you see this on your sign? I can see it. This is uh Mike's site. It'll get it will give you some information about what he's doing. There's a lot of really good information here about just generally how uh agentic is working for him. But this this what I liked about this piece was here, this um graph that you put together of the actual like knowledge itself, what Tropo is. Uh and you know, you've got all your different nodes here that it's reading from. I was wondering, I had a question about the agents and the um skills that drive this thing. Did you were you creating your own custom agents and skills? And or did you use any public skills that are out there that are like, okay, these are great too. I might as well load those in.

Mike Maziarz 29:13
Yeah, yes, it's all it's all um it's all very much custom. Um so every agent has been groomed. Um it's been groomed over eight, eight years, like eight months. So every agent has a life cycle, if you will. They're born, um, they do work, and they all retire. And they all have a unique identifier within their generational history. Uh, my one with the most sessions in it is 194 sessions. It's kind of my main uh architect uh agent, and it knows its lineage. Like it'll say 47 generations ago, Argus said this, and it it has the ability to kind of pull up that context, um, which is really wonderful because they don't they with the right context around them, they don't forget. Um they're ephemeral, they like they they they are born and they die, and that's that's all an agent really is. But it's the we call it a studio, we call tropo a studio because it's like it's like walking into an artist's studio, right? You have a place for everything, there's instructions, or it's kind of like a workshop, if you will, like a really well like a craftsman's workshop where there's all these tools around and there's knowledge about how to get stuff done. Um, so yeah, all the agents are bespoke, they all have a history and a lineage, they can communicate with each other and they know each other's lanes. So they'll assign work to each other in their lanes. Um and once you get it dialed in, it's it's super helpful in terms of skills. Like this is a question I get all the time like, oh, do you create Claude skills? Well, just just kind of the approach I've taken with this. I can nerd out for hours on this, Kevin. So just stop me if I get off. But um skills are skills are just a definition file. Like it's just really like this is the agent, um, this is what you do, these are your boundaries, this is what you're trying to achieve to get something done. And maybe there's some additional like Python scripts that help it to do it, or API calls or MCP calls to other resources. So that just defines an agent and gives it a little bit of a mini studio of what to do and how to operate. What I've tried to do is have kind of a multi-agent studio. So it's not just that one uh one agent, it's multiple. And skills in our in our system are just called tools. So they're called tropo tools. And those are Python scripts. All the tools are Python scripts that give an agent, it turns an agent from just doing determinist uh like reasoning behavior to deterministic behavior. So it gives it like a little, it gives it a little app it can run to do certain work, like go create a file. It just admit it like it actually pulls a template, it just executes a Python script and it creates that. So what we've really done over the over time is created a lot more of these tools for the agents to use. We've given them the ability to search for them and uh a little help system so that knows how to do it. And if you get that right, you're really not uh dependent on whether you're using um whether you're using Codex or ChatGPT, you're using Claude, you're using um cursor. I use all of these. I even use one um from an open uh an open way company, uh ZAI. It's a Chinese-based uh opening model, and it works fantastic. So all my agents can live in all of those harnesses, and none of their memory is in any of any of these harnesses. So you don't get locked in.

Kevin Kerner 32:36
Yeah, that's fantastic. Um, and um I'm assuming you when you download this, all of this is stored locally, right? You're storing it literally on your local desktop, right? Yeah. So it's pretty much it.

Mike Maziarz 32:48
There's a download there. The download is just downloading a zip file, and all it is is a folder structure. And with instructions for when you start up an agent, it will, it'll literally, there's a all um agents will read something called agents.ai, um agents.md. Um if it's claude, it'll read claude.md. That's kind of like agent instructions where it, when it's in a place, it doesn't know. So it'll read that. And so it starts with those files and it it it introduces that agent that you launch to the tropo environment, and then it loads some other boot files that let, you know, kind of informs on the environment, and then it just walks you through it.

Kevin Kerner 33:25
Amazing. I was so excited to do this when we had our pre-call, but that's been a while, but I'm definitely gonna get into this. I'll have to go down the rabbit hole. Two questions. What does your team think about this? Because they're looking at Mike and they're going, Mike, what is happening? And then what does the executive team think? You are like a power user of this stuff. What's the support you get from the Uz X and how does the team feel about it?

Mike Maziarz 33:45
Yeah, I mean, that's a really good question. Um so the team has been very enthusiastic about embracing AI on the, especially on the so I I've got two teams. I've got a the marketing team and the product team. The marketing team embraced it right away. They all use Mo all the time. It has been a productivity boost. Um so they've loved it. They are using AI also, they are like gaining their own skills. Um, they're not all coding, they're not all using Tropo the way it's been designed to use, but they're using some of the concepts. Um I had to realize early on that going in deep means, you know, you really can't push it on people. Like people have to come at it at their own kind of their own pace and their own perspective. I'm one of those um I'm of the age that they say where you know you become a systems thinker in technology over 30 years. So like you think in terms of systems and how things are done. So it's just a like a wonderful thing that this AI thing happened because we can now we can now interact with a team that can make your vision reality versus that website. I didn't write a single HTML. I I don't go to the place where it's hosted, which is uh uh Vercell. Like it like my agent it it assembles it all and it pushes it through a publishing pipeline and it just goes out. It's like I wouldn't even try to change one word. I would have an agent change a single word or a single error. So anyhow, not everyone's ready to do that. In terms of the um, in terms of the uh the executive team. Yeah, I think so like our CEO has absolutely embraced AI. Our C COO has absolutely embraced AI. Our backers and our board is very much like, let's let's do this, let's get on board. Um that is going to be the future. Um but we have to go at a certain pace, right? Because we have to move the organization along. Uh, we have to deal with like we're because we deal with finance data and we're in uh the office of the CFO, we have lots of uh certifications. SOC 2 Type 2, we're going for ISO 40 2001. And so we're very cautious about adopting new technology and integrating it into our process and workflows from a security and a privacy perspective. But everyone's embracing it, and they all know that we're in this for a journey. It's not about a quick, a quick hit. Um but but yeah, no, it's I think they've been very, very much welcoming of it.

Kevin Kerner 36:23
We'll just say I always talk with CMOs I talk to, uh I always say, you've got to get into this stuff. And I point them at people like you that say, these you can do this. It's not magic. You don't need to be a software engineer, you can actually do it. Don't you think it makes you a better leader of your people to understand these these things than not understanding them? Like if you didn't if you didn't understand how that all this stuff works, would you be, would you be at a disadvantage?

Mike Maziarz 36:47
Absolutely. I 100%. Um this is a fundamental shift. This is I I just did a part of a keynote for um our user conference this week. And one of the things I talked about was the parallels to what's happening right now to the industrial revolution. And it really is that big and and and monumental. It didn't just change like a little bit about work, it changed everything about work. You know, where it was done, what skills were needed, what capital was needed. The only difference, the main difference now is it's happening like 20 times faster than the industrial revolution, maybe more than 20 times faster than the industrial revolution. So, with that fundamental change going on, if you're a leader and you don't really understand the dynamics of what's happening, you don't know where it's going. You don't have that vision to see where it's going to be in two or three years. And that puts your whole entire organization at risk. It's almost not responsible. Um and I can even give like I'll give you a different parallel where like I'm because I'm a I'm a I'm a I'm a software guy, I'm like a technology guy, I've gone through a lot of different industries. Whenever I enter a new domain where I don't have deep expertise, like accounting and auditing, I didn't have that when I started here. I was at a disadvantage not knowing that. How can I make the best decisions about our positioning and messaging and product until you really go in deep enough to be intelligent about that market? Um and so that's just the education I had to do when I got started, or I get started at any particular company. And I think this is just so fundamentally different because it's going to change the nature of work forever.

Kevin Kerner 38:26
Yeah. If you're a CMO who hasn't gone as deep as you as you have, what's the first step you should take to getting there?

Mike Maziarz 38:34
First of all, I'm not necessarily advocating that everyone kind of goes. I mean, this is it's it's uh it could be a you could get some uh odd looks. I happen to my wife and I happen to be empty nesters. Um so you know, it's kind of like a hobby to me. To me, I was like, That's right. I told my wife, because this there's a significant token budget on a monthly basis. Um I'm not gonna tell you what that is, but she's been amazingly accommodating with me. But I said, I said, Tracy, um think of this like I'm going back and getting a second MBA. In school, yeah, it's another degree, right? Yeah. That was a hundred thousand dollars, right? Or something like that. A lot cheaper than that. In 2000, right? So, like, you know, like I so anyhow, you don't have to do that. Um, but what you should do is you should read. There's so many resources like this, this um, like this podcast, there's so many out there. I would read and educate yourself and get a little dirty. Spend a hundred bucks a month on get at least a basic coding package and try to build a little website or just try to build a productivity app or just try to build a, yeah, like a productivity app, like organizing your notes. Like I'm a horrible note taker, and I'm even worse at kind of organizing those notes. And it does it extremely well. So I would just take on a project that you can get behind and that helps you. And then I think you'll you'll find it'll be a very natural progression of wanting to learn a little bit more and wanting to learn a little bit more after that.

Kevin Kerner 39:59
Yeah. The other thing I the other I think it's great advice. The other thing I always tell them is like, don't we all as humans, we have this desire to know the answer. And so when something comes up that you don't know how to do, your initial reaction as a human is like, well, how what am I gonna do? How am I gonna figure this out? And in this case, it's better to, when you're experimenting and you're free thinking, it's better to act to give in to the fact that you just don't know. But you have this thing that knows all things. So act as dumb as you can. It's better to just tell the thing, I have no idea what I'm doing. I don't know how to load claud code, I have no idea what I want to build a website, but I have no clue how that connection happens. Talk in plain language and it will just tell you the thing in a very nice way. Kevin, it's amazing how fast you get there.

Mike Maziarz 40:43
You are so right. So, for instance, so if you do get claud code, get codecs, and think you're gonna go build something, right? You start right away and you say, I'm not an engineer, right? This is what I do, but this is what I'm looking to do. How do I get started? How do I think about this? You just ask those questions and you will get the answers. Not only will you get the answers like we used to get, like doing a Google search, it's gonna go set up a GitHub account. It'll set up your Git account. What's GitHub? Oh, GitHub will do this. Oh, you need GitHub? I'll do that for you. Yeah, no problem. What is that? And it'll like give me access to your file system and I will go and create, you know. And like it's just really remarkable.

Kevin Kerner 41:26
It's incredible. So this is really, really good. Okay, I have three questions I want to ask you that I ask every guest and then AI roulette here, real quick. Um, these could be short. Um I put a little benchmark study together to see how these are answered. So for you, the first question is what's one AI workflow you put into production this year that had some positive result? I think I know what that is, and one you killed because it didn't work. Like the and we're not gonna do this, it's not gonna work.

Mike Maziarz 41:52
Sure. I'm gonna surprise you on the first one because we already talked about Mo. Uh, that was on the marketing side. On the product side, AI is just an amazing thing that's happening to software companies right now. You know, everybody heard about the SaaS Pocalypse uh back in May or so. And if you went in and bought software companies back then, you made a lot of money this year. Um it's not going away. In fact, AI is making companies like ours more productive, more competitive, being able to innovate faster. So that's that's one thing. Secondly, AI, it's not about repla it's not about replacing what MindBridge does with a prompt. It's never going to happen. That just is not how LLMs work. They don't do that. But what's amazing is integrating AI capabilities and agentic capabilities into the seams of software, in the seams of deterministic platforms like we have. I'll give you a few examples of workflows that are working exceptionally well for us. One is we deal with data. And if you can't get the data into our system, we can't add any value. It's a big data pipeline. You push the data in, we do a lot of work, and we produce risk scores. So we're we've got it to the development of an agentic um integration agent. So it will integrate, it'll ingest, and it'll map or transform the data to our data model. And that's something that's always been very challenging and very services intensive. Historically, yeah. Exactly. Exactly. Yeah. And so, but it's something that actually agents are really good at. It can look at this data model. It knows our and so that's that's one example. Another example is because this is data analytics, um, you know, there we've got more PhDs in our company than I've had in certain companies as employees. Um it's a little it's data science heavy, it's it's machine learning heavy. And so when you configure a data pipeline, it's quite sophisticated. You have to know all the how all the knobs work together to get the result that you're looking to achieve. And so we are developing ADA, which is it's basically an agentic design tool, right? So it's an agent that does our configuration of the design. So we will ask the user questions about what it's trying to achieve, and then it'll configure the data pipeline in order to achieve those results. It'll map it into the jobs to be done, basically. So that has been a very, very challenging thing for us to do in the field, but now it's becoming a lot easier. And the third agentic workflow that's in there that's just absolutely amazing, is we put out a lot of data when we analyze data. So statistics, risk scores, curves, plots. Having an agentic interface into that data is just tremendous. An agent makes sense of it. Like it's like, oh, this is what you're trying to understand from this big data set that was just analyzed. So it turns things from being you've all seen, you've seen data analytics dashboards. They're dense and there's curves and they're like, what am I looking at? Now you can ask a business question of all of this, and the our agents now can sort through that and really get the user what they needed. Because they're often just business users that need to know kind of where their risk lies or where they should be looking. So those are three workflows that we've implemented that are just it just they're just adding incredible value for our customers.

Kevin Kerner 45:22
What about on the marketing side? Anything you did? Anything one thing is.

Mike Maziarz 45:29
I'll give you a challenge on the marketing side. It has been a challenge, honestly, to get the um get the agentic workflows online and approved through security and privacy. Because we want to tie it from like from like our BDRs who are out there doing calling. They need data outside of our from our CRM. That has to kind of flow in. But they also want to match that up to okay, well, what is this person that I'm calling on doing in Google? Like, do a Google search, look them on LinkedIn. And so assembling all that data into an object that will like be a nice little cockpit for a BDR or a salesperson, it like it's very achievable. We've we've prototyped it, but honestly, we haven't rolled it out yet because we're trying to make sure that we do all the right checks and balances for security. We're also a Canadian company, which we have um so we're um we're uh privacy security. Yeah, like some some different laws. Um but yeah, so those that's one area where we've definitely had uh some challenges. I will say we haven't pulled the plug on anything. It's just it's really a matter of refining and you know, like persistence and just getting it better.

Kevin Kerner 46:39
Well, that just kind of answers the second question too. I was gonna ask you what's still stuck in pilot, something you expected to be running by now that isn't. And you would expect to being able to deliver that type of intelligence to you. Well, it sounds like you built it, you just can't get it through because of privacy and security and those type of things, which will probably get fixed at some point.

Mike Maziarz 46:56
Oh, it it will. It will. I think we just have to make sure we get it right. I mean, we just have to make sure that we get it right and the governance is there, and um, we would rather take our time with security and privacy than um than rush with it.

Kevin Kerner 47:10
One more question. What's the one thing your team won't hand to AI and why? You're saying, nope, we won't give this to AI.

Mike Maziarz 47:17
Yeah, yeah. There has been, there's been some pushback in pockets of engineering, um, especially around data science-y things that are complex. Um so that that's a little bit of a challenge uh right there. But I think that as uh AI is being used to um they're using it as a design partner right now, and I think that's successful, but it's not really generating the code there. And on the marketing side, it's wonderful for so many things, like we talked about with Mo, but the team is not bought into some of the, especially like the social uh posting uh uh content for the website that comes out a little too structured or not the right sense of humor. Or so I think there's still refinement to do for that, but it's not a complete throwaway uh because they'll use it as a starter, like kind of a starting process, and then they have to refine it themselves. Yeah. But we haven't fully automated those pieces, yeah.

Kevin Kerner 48:13
With all these uh advanced models now, Astra, and now you have 5.5, it's just they just can't quite get there. But I do think they will get there at some point. I I really believe it will. The nuance will get there.

Mike Maziarz 48:27
Yeah, I think the mindset is one of the mindsets I have is it's just I don't know, I read it somewhere, someone talked about it. It might have been a Google, Google, Google DeepMind uh person years ago. It's just like, well, what is possible if you look out whatever period of time and say the level of compute is gonna be 10x what it is right now, and the cost is gonna be one tenth of what it is right now. So, whatever time frame that may be, like if you have that attitude that it's gonna be 10x better and one-tenth cheaper, like or 90% cheaper, like it just puts you in a different mindset. So you just gotta keep trying on it.

Kevin Kerner 49:06
Yeah, right. I don't think it, by the way, will replace humans because we're we're just so creative and there's so much we're unpredictable. And I think AI has a hard time becoming as unpredictable as uh the human mind, and because we can go all over the place. So getting to that's gonna be really tough in our lifetime, I think, but who knows? Okay, I wanted to ask you one one last question. This is AI Roulette. I have your profile and some things, and actually a little bit of this conversation going on in here, and the AI is gonna ask you a question. So this is coming from directly from robot. What model? Okay, here we go. Mike, you said if the laser beam gets out of whack, there's an easy person to fire. I think that's what you said in our pre-call. Um, that person runs product, and that person runs product and marketing now, meaning you. So when the website promises something that product can't do, which half of you gets fired first? Marketing.

Mike Maziarz 50:06
I'm sorry, marketers. We're the easiest. I'm sorry, it's so savage.

Kevin Kerner 50:15
Just savage. Savage response.

Mike Maziarz 50:18
It's it's so obvious, too. I mean, what's the average tenure of a CMO these days?

Kevin Kerner 50:22
It's yeah, yeah. Well done. Well, Mike, this has been awesome, incredible. I could geek out with you for a long time. If people want to get a hold of you, open to people reaching out to you in different ways. How should they um emails on the website?

Mike Maziarz 50:36
Emails on the website, mike at tropo ai.com.

Kevin Kerner 50:40
Oh, Mike, this has been a really fun one. This is really good stuff, and I really, really appreciate you being on. Thank you so much.