This is why AI pilots die on the CFO's desk

Mikolas, thanks for having me. Excited to excited to have the discussion.

That's a great, great question. So I I think, you know, all the experience I've had up until this point led me to starting my own company. And I I, you know, over the years, I've always had a very entrepreneurial bend. Even working in, I started my career as a network engineer at a very large media company, Hearst. And even while I was there, I was always trying to get into the latest and and and greatest things. So, like, how do you start A podcast, like back in the early, early days. So I was always like trying to get on like at the forefront of whatever it was, interactive media, et cetera. And I was lucky enough, you know, through the years to then leave a company that I was at for quite a long time, 17 years, and join AWS in the very, very early days of cloud adoption. And during that time, not only did I meet incredible people, but I kept gaining experience in not only technology, but business. And you know, I think it just kept compounding and compounding. And then went out and I spent 10 years in private equity backed startups around the AWS ecosystem, Databricks, Google, Snowflake, Anthropoc, OpenAI. And what that led to was realizing about six months ago, it's a it's an old saying, or maybe not an old saying, but it's a saying that Andy Jassy had at AWS where there's no compression algorithm for experience. And it's the first time in my career where I felt like there, I was, I was meeting the intersection of my experience with this business opportunity, which is AI adoption and what's going to come through this revolution and the technical capability that is now democratized through AI. So for me, it was a perfect time to to start a business. And I'm really excited about where we're headed.

I I think the the number one thing that we're looking at when we we look at any of these mid market enterprises, you know, smaller enterprise companies, even small businesses, is what are you looking to do with the business? Are you looking to save money? Are you looking to make money? I'm sure you're probably trying to do both, but technology is, you know, I mentioned that it's democratized, but technology's actually, I don't want to say it's gotten easier. But it's definitely been something that is more attainable and accessible to an everyday person now. Now there's right ways to do technology and wrong ways to do technology, but understanding business process, measuring what that return on an investment could be through AI, and understanding what that ownership looks like, what it's doing, how it's going to perform, so that you hit major KPIs within your business. I think that's the most important thing. And that's typically where we start is understanding that because Too many times, Miklas, we're seeing companies are just trying to, and and and they don't mean to do this, but they're being forced, whether it's the board, whether it's their customers, they're being asked, hey, we need to adopt AI. So they'll they'll just launch an AI pilot and really not sometimes look at what the business impact could be or should be. And that's where we typically start.

Yes. so there's a there's a few patterns that are emerging that we've seen over the last couple of years. But if you look at if you look at companies right now, and I'm you know, I have a a revenue background as well. So if you look at companies, it all starts with sales. If if you don't have sales in a company, you don't have a company. You could have the greatest product out there, but if and no one knows about it, you you really you're not selling anything so we typically are starting with sales there's different ways to create sales like how do you create more productivity out of your sales team so there's there's patterns we're seeing around that with ai adoption some sales teams are you know have that lap you know like really locked down but how do you take as much toil out of their work so that they can focus on only what they can do which is create relationships you know do customer service All the things that, you know, a relationship based sale is based on, which is trust building, et cetera. Those are things that AI can't do today. But that's where we like that typically where we start. And there's a lot of things that that go on in the back end operations, et cetera. But I think sales has a really unique opportunity in this AI adoption curve right now.

I think there's two main reasons. I think you do not have someone typically this is what we see. You don't have someone being accountable on the business side. And you don't see someone or the company measuring exactly what we were talking about, like return on investment, total cost of ownership. All of those things add up to a lack of adoption. And what happens, you start seeing, or if they are adopting, you start seeing what I would call a lagging indicator, which is At the end of a quarter, a month, whatever it might be, a CFO sees a massive, massive bill potentially, if things are implemented maybe not not quite the right way. They see a massive bill for whatever the frontier model, whatever that might be. And they're like, Well, wait a second. You know, we spent X amount of dollars. Mr. Business Leader, Mrs. Business Leader, what did you save? How what what kind of efficiency did you get? And and nine times out of ten, what we've seen is that it's We we we weren't measuring that or that that's going to come later. And and that's where people run into some issues because sometimes the plug gets pulled, all sorts of things like that, that can happen. So I really think starting with the business, understanding the business outcome, understanding what you're measuring and having accountability is is going to make a pilot become production successful.

That's a that's a really great question. I've seen some amazing ones. One one that I that's a little bit near and dear to my heart has been a resource allocation machine learning slash AI application. What I mean by that is that there are companies out there that have professional services teams, crews that are out there, whether they're fixing or contractors out in big you know, working on big construction sites. Working on oil rigs. one of the b one of the hardest challenges for anybody that is managing those types of crews is understanding where to put those crews next. So, like, what's the you know, who are the people without projects? What are the projects without people? When is this project ending? How are we gonna solve where we're going to put people? And I've seen some really interesting use cases. We built an interesting use case around. A resource management, AI-enabled resource management that allowed a company to really allocate their crews in a much more efficient way. And as we're measuring this, we're seeing a significant return on investment by putting people in the right place based on geography, skill set, et cetera. Where before that was either tribal knowledge, it was spreadsheet, it was a lot of work to get that. And you're starting to see a a lot of that type of adoption take place. And that's That's probably a boring use case. There's so many more, but that that one is one that's a little near and dear to my heart.

It is that's an amazing use case. so we're, you know, we're one of the things that we like to say we're very agnostic on the LLM side. So Anthropic, OpenAI, Grok, amazing, Meta, Gemini, obviously from Google, and some of the open weights models as well. there w one thing that I think we're we're priding ourselves on right now is the ability to utilize whatever LLM. makes the most sense for the business that is the most efficient, most cost effective, and gets the best results. so building harnesses that are able to do that. I mean, I have a board of directors. they're not physical. They're a board of directors of LLMs that whenever I'm doing work or I'm I'm checking like, hey, how are we pacing on bookings? How are we pacing on revenue? How are we pacing our delivery? I have a board of directors that that actually looks and you you declare a chairman of the board. My chairman of the board this week is clawed. It might be something different next week. But that that allows us to really try out all of the different models and understand where where we're at. So if you look at our stack right now, using all the LLMs, Notaker, and I don't get paid for this, but the best note taker that I have seen by far, I recommend it to anybody, is Granola. Granola AI is amazing. it's great for in-person meetings, it's great for any Zoom meeting. You know, Google Meet. If you're if you're like me differing with different different clients, you're you're on Google Meet one time, Teams. I I had a WebEx meeting last week. No one's had a WebEx meeting in a very long time. So it's interesting. So I think granola is a huge part of it. And then we run all of our applications headless. So our CRM, we use Attio, that is completely headless. So that's all running through our harness, like whether it's cowork or codex. We're working through our CRM completely headless. all of the applications that were that even QuickBooks, you know, anything that has to do with invoicing, revenue, that is all done as headless as possible. And that allows us to integrate in within the actual harness itself, whether that's cowork, whether that's a custom harness, whatever that might be. And next week, Micholas, if you ask me, we might we might be using other tools, who knows? But You know, they're all they're they're all you know, changing and accelerating and innovating at an incredible pace right now. it's hard to keep up.

Yeah, I've I've tried over, you know, I would say, I guess it's two and a half. I have I've I've I've used I I think over the last couple of years there's been lots of them. I'm not gonna pick on any one of them. There's the built-in tools that are that are part of the meetings. what I've noticed about granola is that the actual ability to recognize who's speaking.

So it it will pick up that context. and then it's a b you know, so you you get really good transcripts and then the ability for any of the LLMs to take that context, take what's what's happened in there. The real power is what comes out of those notes. Those notes then power my to-do list. So that gets published in something called things, or it gets published in notes in in Attio, or it gets published in notes in a Google Doc, whatever it might be. those things are all really, really That the tighter that integration is, the better. The one use case that I love is the in-person meetings. Now, you have to tell people that you're recording them. So I'll be in an in-person, you know, I'll be in an in-person meeting. I'll say, Hey, just so you know, I'm going to I'm just recording this with my note taker. And most people are totally fine with that. That that's not an issue. And then that allows me to when I get back to my desk and I'm I'm looking at stuff like I missed something in the conversation, or if I was sitting there taking scribbling notes on my iPad.

I would have missed some context and all the note takers do a really good job of capturing a lot of the actions that take out of it. But I think the power is having that brought into whatever harness you're using. I'll I'll use cowork as an example. We're using cowork at the end of the week. Every day I get a brief again, how many tasks I had, what what meetings we had yesterday. And then at the end of the week, I get a weekly status and what we call a loose ends report. What meanings took place? And that there were actions a sign that we did not action. And then you go through it, you triage it, and and you get back. So it's it's been a wonderful way to create efficiency and allow you to do more with with a lot less.

I I couldn't ag I couldn't agree more. Why why would you this seems like a very simple concept, but it's hard to become an expert in every application. And when you can create headless applications using natural language in a in a Cloud Co work, it it's it's outrageous. You start you have these aha moments and You know, I don't want to pick on any SaaS vendors, you know, my friends, some of my great colleagues, they call it the SaaS pocalypse that SaaS is going to be, you know, potentially come down. I don't believe that. I I believe that there's going to be a combination of things, but I'm asked all the time by clients, hey, I have this massive, massive bill with insert SaaS vendor here. Salesforce is is one that comes up a lot. It's a massive, massive bill. Can we code? our own CRM that meets the exact needs of our business rather than trying to get this to match it using a lot of consulting hours, a lot of licensing. So it's interesting. We're in we're in the early, early days of of this, but headless to me is is an amazing opportunity.

Amazing. Amazing. I I so the the the the the subject that comes up a lot, Nicholas, we always talk like if you're talking to your colleagues, talking to customers, talking to people that are really immersed in the technology infrastructure of AI, there's a lot of debate on whether there's you know the so called business moat around LLMs. And what we're seeing right now. Is that it's an arms race, right? Like you're seeing every week, last week Gemini, you know, launched Argon. It's incredible. Meta, you know, launches their latest version of the Frontier LLM. Like there is. And then, you know, next week Anthropic's gonna come out. So it's an arms race. And you know, some people are picking one and sticking with it. Some are saying, Well, I wanna be agnostic. I want to be able to use whatever, whatever we want. So it's a really interesting time at the acceleration of not only the frontier models, the open weights models. I I saw something, I think it was like yesterday or the day before, that Kimmy might or Quen might only be six weeks behind Opus, which is incredible to me because it's, you know, Opus is one of the best LLM, if not the best LLM I've used thus far. It's just Weeks is weeks is not a big deal to me if you're if you're looking at open weights. So it's just an amazing time right now. And that that brings up like the business conversation about AI adoption. Is it going to be more about the LLM or is it going to be the services and implementation around these LLMs? And that and that's really the bet that we're taking with strong tide.

It's i yeah, I have a use case that I use grockbot for. So would I you if you adopted just a few months ago and you already have like a s a setup, if it's very locked into your workflow, it's hard to you know change some things. But in parallel, I started playing with MetaMuse and Grokbot. And what I really liked about Grokbot was and if if you watch the folks from SpaceX who are Incredibly brilliant. The way they tell you to set it up is create a chief of staff agent. Have that chief of staff agent then create worker agents depending on the work you need done. So you create this chief of staff agent. I I've created one. And you tell that chief of staff, hey, I'm looking to, you know, whatever it is, I want to know all AI news that happened in the last day. I want a daily briefing on it. It creates. That that agent for you. Then you create, you know, and these are, you know, they're they're running 24-7, they're doing everything you need. I've I've seen and I used this the other day, and I never thought I would would do this, but I, you know, the other day there was something that I mean it was something silly, like protein bars. They were they were expensive somewhere. I'm like, I wonder if I could create a shopper agent. So I go to Groundbot, I'm like, hey, you know, this specific protein bar I noticed when I'm on Amazon, it's you know, I don't know, it was like ten dollars. And it created an agent looking for different price variations, all you know, 24 hours a day looking for this. And it found it significantly cheaper somewhere. And that's just that's just like a small use case that you could see, like when you, you know, you could see the capability and the opportunity. it's just the limit is creativity right now. That's that's really the limit of any of this.

Don't worry, that Optimus robot is coming. And that and that's going to be, you know, I talk to people about this too. You know, that that's an interesting if that does happen, like what what happens when, you know, I'm not saying these are not expensive things, but you know, there'll be there'll be financing like there are for cars to buy one of these robots. And exactly what you're saying, Nicholas, your dishwasher. You want the oil changed on your car, it's going to be able to do that. It's going to be able to paint, you know, maybe a room for you. It's going to be able to do your lawn. I it there is just so much that I I don't know if I ever want to outsource any of those things, but that's that's where we're headed.

Ha ha. I agree.

It's it's crazy. Nicholas, I spent some time a couple of years back at a company called Boston Dynamics visiting. There were a client of mine at one of my previous companies, and I've never had up until that point, I had never had a physical, like visceral reaction to something. And I saw they had their their robot atlas. Now they have the dog spot that can walk around, and then they have Atlas, which can mimic human movement, can do you know, somersaults, it can fall. And I I had a reaction that was physical seeing this. I was like, my goodness, this is, this is, this is wild. Like we are in the future now. And that would, that was an incredible experience. So we're we're closer there, but I'm with you. I am probably not going to be trusting any firmware updates in my house with a robot.

best ways to put yeah, a couple of places. Strongtide.ai, LinkedIn, always get some notes on there. it's probably two of the best places. Also have an Instagram presence for both personal and and strong tide as as we start building more and more of a brand that's personal as we get into these AI journeys right now.

Thank you, Micklas. Great conversation.