← Startup Stories - Mixergy
Loading summary
Transcript110 lines
- [00:00]
A
Every time I see my buddy, he tells me how much his business is growing. He created a company where he uses AI to create AI software for clients and it's really taken off. It's an interesting model. So I asked him to do an interview to walk me step by step through how he built the business. Here goes. Tarun Thumala is the founder of Press W, the AI engineering firm the Next New Thing presented by Zapier, the AI automation company. How much revenue are you bringing in right now with your agency?
- [00:33]
B
We're going to be doing well over $2 million this year.
- [00:36]
A
How much did you do last year?
- [00:38]
B
Yeah, we did over $2 million last year.
- [00:41]
A
Impressive. Give me an example of a typical thing that you do for your clients.
- [00:46]
B
Yeah, so, so we're largely custom AI application developers and so, you know, AI applications come in a lot of flavors. Either we are helping them advance their own products, especially if they're like something like a SaaS, or a lot of our work primarily is on building like internal operation applications that help them do their workflows better, smoother, faster with their teams.
- [01:09]
A
Is there an example project that you did that gives us an understanding of what you do?
- [01:14]
B
Sure. There's a lot of companies that for example, process a lot of documents. We have a case study on our website. You can go look at it's background checking company. They, they look at a ton of different court systems and every single court system in the US they have, there's over 5,000 of them that pull different records from and you need to examine that record and each one look looks and feels a little bit different. So this is, you know, unstructured data that we're looking at and we need to figure out whether or not a person has any sort of like infraction in that region. They just do so many documents. Right. Because every person that comes in, they have to go and do this work for, um, we were able to like build a little system that you know, can automate up to like 90, 95% plus of those types of documents.
- [02:02]
A
What do you charge for something like that?
- [02:05]
B
We have a pretty typical kind of custom software model, I would say. And I actually think it's a little bit broken the way that consultants and the other, you know, historically these types of service businesses have built because they're usually like a price per hour type business, right. And you have like resources that are spending on it. So we still use that as estimators for effort. But it's a, that's actually probably one of the trickiest parts about what we do because we are constantly getting faster and better at what we do. So how do you charge, you know, more value based rather than just like raw time spent? That's tough. Usually what we will try and fix is either or we'll try and do is either sink into a fixed fee model where we just know exactly what we need to go and do. We can estimate it ourselves internally and just give a, give a fixed cost. Especially if we understand how big of a problem this is for that customer. Or more often than not we're like, look, we think that this problem is going to take us a month to solve. Here's one engineer. And so we have a forward deployed engineer model where we're like, here's one engineer. It's going to take them a month. But this is a capacity plan and you're just paying for a whole dedicated engineer for as long as it takes them to go solve that problem.
- [03:16]
A
Okay, man, I get it. Because the problem now is that the better you are, the faster you are, the less you get paid and you don't have any incentive to speed up if that's the way the structure is. Right?
- [03:27]
B
Yeah, exactly right.
- [03:29]
A
You know, the interesting thing about you is that you started out with an idea that many people did when they went into consulting, which is, I'm going to create software. I don't know what yet, I don't know how we're going to pay for it. But if I could do consulting where I build software for other people, I'll be able to find the best developers, bring them in house. I'll be able to understand how to build and how to sell and how to create by watching what my clients do. That was the original idea. Before we go into how you changed it, what made sense about it?
- [03:56]
B
I think that there's, there were two macro factors in particular, actually, let's call it three. One was a little bit more personal, like when, when we sold our last, you know, AI product company.
- [04:08]
A
Mm.
- [04:09]
B
I was probably 24 at that time. Um, and my entire life had been either spent in academia or on that last company. You know, I had a couple internships and things like that, but I really didn't understand a lot about how the world worked. And AI is a very context dependent sort of skill in industry. Right. Like you need to actually understand the workflows and the data and how people are supposed to communicate. So we realized that we had a huge gap in understanding. My, my co founders also, we're all around the same age and so we knew that we didn't really understand what enterprises and mid market businesses tended to look like and operate. And so we needed to gain that information ourselves. So that was, that was the first thing. Yeah. The second thing was that we also understood just being AI practitioners ourselves, that the typical way that SaaS type businesses like software style businesses were built, scaled and sold in the past 10 years was not going to be the same formula that was going to work in this new age. Because AI requires a ton more professional services almost to actually get the results from that application. Because it's not so like out of the box. Right. I can't just like spend a bunch of time making a bunch of UIs and plug it into something and then it AI just works naturally for them. You have to go in, make sure that the data is correct, that the context is being fed.
- [05:24]
A
Right.
- [05:24]
B
And interconnect it within that organization. So we knew that like the actual delivery model was changing in the AI ecosystem. And the final third thing was just that like we, we knew that the big incumbent players had some advantages. It wasn't clear exactly what those advantages were when we started, but we definitely didn't want to go and sink ourselves into another 10 year bet when everything was so primordial. So I'd say for like those three reasons we were like services make sense, we can get cash now, we can watch the world and we can stay on the forefront of what's possible, possible today. And ideally take those learnings like just maximize surface area of information and then from that, you know, start to build something longer term, more scalable.
- [06:05]
A
How'd you get your first customer?
- [06:09]
B
Honestly, the story of our first customer and even up till now has been largely the same. It's, it's relationships and networking. I don't, I don't think like service businesses are, you know, that, that different. Especially when they start they're largely like relationship based.
- [06:25]
A
I think first relationship.
- [06:27]
B
Somebody that somebody that we worked with in, in the past like reached back out to us. They knew us from our first company. One of my partners had worked for them previously. They were like, hey, we have some like extra project work. And at that time we were like ready to just kind of take whatever came our way and that would, it was, I would say the first six months was anybody with a enough to pay us, a small team of three to like go and do some work. Like we were taking it. And then we started to really like say start saying no after we had a bit of a base going.
- [06:56]
A
Former employer is a big way that People get customers, right? Go back to your boss and say, you like my work, I'm going to do more of it. But now I've got a team. Do you, would you hire me? Is that what it was?
- [07:07]
B
Yeah, I, yes and no. Like, we kind of really like the relationship was one of my partners and his former employer. And they just knew. But yeah, I mean it's, it's exactly what you said. They, they knew that he did great work and he was like, I have this other vehicle for it. And that kind of gave us our first kind of dollars in the door. But.
- [07:28]
A
Why do you need three partners?
- [07:31]
B
So in my last company, I was a cto and these guys were my first two, like founding engineers. The, the real answer is that I realized ultimately, I think like with this group of people, like, first of all, I enjoyed working with them, which I think is like actually really important. And I also kind of understood that, hey, look, I think as long as, like we're in this game, we're going to make a lot of money, we're going to be successful. I'd rather my story be written with them than without it. And so I think that's, that's, that's fine. Like, let's go do that.
- [08:04]
A
You don't want to just take more of the money for yourself. You can hire people, you can just do it even as a one or two person operation, can't you?
- [08:12]
B
Yeah, I, we, I mean, it definitely made it a lot easier, especially with our particular skill sets. Like we, we, we worked really, really well together. I'd say, like, I have more BD tendencies and I think like my area that I'm really good at is speaking to a customer and taking something really complicated technically, which, you know, my background in AI computer science is there, and then actually just translating that for people and helping them get excited about it and understand it. Whereas their skills are like architecture and AI respectively. And so like just the combination of the three, I think like that, that equated to a much higher, you know, multiple than just like any, any of those some parts.
- [08:50]
A
You've told me a lot then that since this is your role to bring in customers and close them, a lot of it has been through referrals. Do you do anything to juice that up? A lot of it has been through old relationships. What's your process for turning old people into old relationships? Into customers?
- [09:04]
B
Yeah, actually, no, I would say only like less than 10% of this business. Less than, yeah, less than 10%. Let's call it, let's just make it, be generous with it. Less than 10% came from, like, old relationships. I'd say 90% were people that I've met and interacted with since starting the business. So you can treat that as net new though they are all like network driven, right? I think for me, like, my biggest secret is not always trying to go out there and like, hunt for sales. I just like naturally talk about what we're doing. I also think I want to like, preface this entire conversation or just like add an asterisk on it by saying we are in such a big tailwind space where like, there's so much force behind the AI movement that I think like our lanes were always greased and they are greased today. You know, it's easier to get. It's easier to talk to somebody about AI because everyone wants to be in it, everyone wants to be using it. And so like, I naturally think that's like, helped me start conversations. But from there it's like, how do we show the value in their business? How do we actually relate our technology and what's possible? Like, show them the art of the possible. Bring them along that journey along with us.
- [10:13]
A
Start a conversation. How. Where are you starting these conversations?
- [10:17]
B
Oh, at like networking events or people.
- [10:19]
A
You still go to networking events?
- [10:20]
B
Oh, yeah, 100%. I'm a little bit more selective now. There's, yeah, there's like meetups in town. Like, for example, I like still on like the Bridgepoint event most of the time. Nowadays it's somebody in my network knows about an event and will to it. Or there's a second way more recently, which is I just host my own events now and I tell people that I like and that I have found a lot of value for like yourself to come and bring a friend or bring two friends. And I meet people that way as well.
- [10:46]
A
I see. So you've gone out to these like, tech events here in Austin. And I'm sure when people say what do you do? Which is part of the conversation, you say that you've got an AI dev shop or AI automation. We can talk about how you present it. And then people naturally get into a conversation and then they might say, yeah, you know, we can use it. Or you start a relationship where they'll introduce you to someone. That's the approach.
- [11:05]
B
Yeah, exactly. We've also, I. I have a lot of. I've spent a lot of time in like forming relationships with VC and PE people, mostly because I'm interested in finance and investing almost personally myself too. And that's Also been like a really great, I wouldn't say it's like a conscious pipeline, but it's been a great pipeline in the sense that they're like, oh, I have like this portfolio company that could use your guys.
- [11:27]
A
How do you get them to take, to take your call or to get in a conversation with you? Because everybody wants them. Everyone wants them because they think if I could get them to start introducing me to their portfolio, I'm in. And I've got a whole bunch of customers at once. How did you get them to pay attention to you?
- [11:41]
B
I, I, I don't go into these conversations with an agenda typically. Like, genuinely, if you look, if I were to like, dissect and reflect on my conversations with these people, it's like 80% me just really excited about AI and them talking about that with them. And I'm like, this is all the cool stuff that's going on. This is how it can work. Like, this is what we're doing over here. And like, I'm not even telling you, telling them how I could help their business, but they get, they walk away from that conversation. I think I'm speaking on their behalf. Maybe they, they might have a different answer. Like, I think they walk away from it being like, that guy understands AI. He's excited about it. He would be somebody, like, who, if I would introduce, like, he would have a good conversation with them about AI and that's the worst case that could happen. And I think that's, that's the impression I'm hoping that I leave with people.
- [12:29]
A
How many nights a week would you say that you went out in the beginning when you were doing your early networking?
- [12:35]
B
Not as, not as much as many other people I've met. I've met some serious networkers. Yeah, maybe, maybe like once a week. Once every other week. Like, not too many. I'm, I'm, I'm also a big believer in, like, you go to the right spots, like, you'll, you find this stuff a lot more readily. I would say, like, I was pretty particular about where and when I was spending my time. And then the other thing, the other side of that, rather than going and just trying to meet a volume, like a net volume of new people. I was really, I was, I spent a lot of time, like, culturing and work, like, culturing my cultivating, sorry, my, my relationships that I've already formed. Right. And it's really easy when those people are also very interesting and they're working on interesting things in their businesses or in their industries, like because that's how I get to learn. And that's like the best part, I think about what I get to do is I'm talking to people that are experts in their own field and I get to translate that information back. And I, I like, I. I just like to learn about all these different things. And so it's easy for me to be like, yeah, let's get, let's get dinner, let's get drinks, let's catch up, let's see what's happening. And, and, and that's like translated really well.
- [13:43]
A
You know who's a big networker. I can't think of his last name right now, but Chris, the guy who hosts events here.
- [13:49]
B
Who is it? Chris Beeman. Yeah. I think I'm doing like Fireside Chat for, For his astronomy thing, like next week.
- [13:56]
A
Yeah, he's out so much.
- [13:59]
B
I could not. I, I'm not, I'm not that type of person. I know people are. I am not. In fact, most of the time at these networking events, I go into it with a lot of like, dread and I'm like, I don't want to meet all these new people. I end up always having a good time, but it's taxing for me. I'm not the type of person. I'm much more of like a one on one person. Like, I genuinely liked meeting people for coffee. And I always get a lot more from that than trying to stick my hand in everyone's face and hope that something pans out.
- [14:27]
A
Yeah, we had a good dinner together. I think you and I are both good at one on one, though. I also enjoy the big ones, the big events. Yeah. By the way, I've got to tell you. So we went out for drinks and then we met people, which I love to do when I go out. I like to spend time one on one with someone, but I inevitably will also expand it and then contract like we did. So the two of us went out for coffee after that. I looked at my credit card. The next day I go, I guess we had a much better time. Or maybe Tarun drank more than I expected. Okay. It is what it is. And it didn't occur to me till later we were talking to this group of people who I think they thought that we. I don't think they were taking advantage, but I think they thought that we were buying a round. And so their round at this place, which was expensive, was on ours. And that also explains why the guy goes to me. I'm. I'm having a guy deliver coke here. Do you want any? Like, he was being, like, exceptionally generous, and I think it was his way of returning the favor of us getting him drinks.
- [15:27]
B
I know exactly what you're talking about. Or I. I know exactly about what you're talking. 100. That's what was happening.
- [15:33]
A
It did not occur to me until later on. I thought he just liked me and want, and I'm sure he did, but I guess we.
- [15:39]
B
No, I mean, you. You have a. You have an insane, in the best way possible, like, energy about you when you're out that I was like, so happy to be a part of in that moment. I love that night, actually. I had a lot of fun.
- [15:54]
A
Yeah, we got to do that again. And so for you, that's the way that you like doing things. Maybe a little more low key than we did it, but essentially it's that.
- [16:01]
B
And then.
- [16:02]
A
And then eventually some business comes from that. You mentioned that you did an event. I went to the event. Did you get any business from it?
- [16:12]
B
Yeah, yeah, definitely. Same way that it happened that we're talking about this time, which is somebody at the event knew somebody that was looking, or maybe they were an investor and had a portco, like, same type of thing. What I find with those events is like, it's. It's almost never direct in the sense that somebody comes to that event is like, I want to hire you tomorrow. Like that. That almost never happens. The though it's. You know, it has. It's. It's mostly just increasing tarun surface area, like in the world and more. And, you know, I don't. I think we are very fortunate in the sense that I actually don't think that there's very many firms that can do what we do technically and with the level of specification that we've had, like, we've been in the generative AI space. Like January 23rd. ChatGPT came out November 20th or November 30th, 2022. So, like, basically a month after we were in the game, and we've been in the game since then. So, like, our volume of work and our ability to speak about it and the breadth of projects that we have I think just like placed us in a very special category. And so when I get to meet people, you know, I'm not just like another marketing agency or, you know, a PR firm or anything like that where there's a lot more competition. We've definitely specialized very heavily.
- [17:26]
A
You did this at the proper hotel in downtown Austin. Beautiful layout. Gave everyone nice notebooks. You had, like, open bar. What did that Run you?
- [17:38]
B
I think like all in. We maybe spent somewhere between like 20 and 25k and like we had sponsors and stuff that helped out with a lot of that. But I'd say like we spent easily like 15.
- [17:50]
A
15 of that was you maybe 10 from the sponsors to cover something like that. You had. I was sitting next to someone who was man, what was the name of the car company? It wasn't. It might have been Maserati or something. She was sitting there and I said, why are you here? She says, I want to understand how we can include AI in our business. And she didn't really know you. You just had connected at one time and she came in. That seemed typical.
- [18:13]
B
Ish.
- [18:14]
A
I guess there were definitely a group of people who are like that. How did you fill up the place?
- [18:20]
B
So I, I did it. The credit credits due with one of my friends. And also we had like a. A joint venture together. Rodeo Turtle. One of my friends, Logan, he. He has a great Austin network. I have a great Austin network. We came together and we're like, we're very particular about again because we're not just like volume people, both of us. We prefer these one on one things. I was like, I have probably about 40 people that I could invite. You have about a 40 people. Let's do that. You know, out of that, maybe 50 to 60 showed up and. And we asked them to like bring along one person that you know, was. We thought would be a good fit. And then we probably.
- [18:52]
A
We did.
- [18:53]
B
We did some cold outreach as well. Like we found other founders in the area that, you know, were we respected or had heard about and asked them to come and that. That's pretty much how we did it.
- [19:03]
A
Shot them an email.
- [19:05]
B
Yeah. Like LinkedIn or email. One of those two. Yep. Wow.
- [19:08]
A
All right. Did you make back the 15,000 from the one deal that you got from a friend of a friend?
- [19:15]
B
Yeah.
- [19:15]
A
You did? Yeah. That's fantastic. So why don't you do more of them?
- [19:20]
B
I think we are. Yeah, we're doing. We. We are working on a secret sort of event project right now. I think that the little teaser, or I'll just give you the title is like AI World Fair. And we're going to do that closer to the summer and then we're probably going to do sprinkling of smaller events from now till then. Um, I think when I, When I did the event it was never. It's gonna sound weird. I know. It's a big investment too. My goal was always just like, like from that event was to cement ourselves as like a real player in the Austin space. Like, get people to know myself and my team, my company's name. And that was really the goal. The BD stuff was like, honestly, a. A bonus. And that continues to be it. Like, I, I don't think that Austin really has like a powerful AI event scene and AI event network almost. And I don't see any reason, like, why we can't own that space. And so I'm going to keep you. I enjoy doing those types of things, so that's like a more personal thing for me.
- [20:24]
A
I'm surprised because it is a lot of work and then if it doesn't work out, then you've got the space. And it look, you look foolish if people don't show up. And when you're offering free tickets is a big reason why people won't show up. They'll say yes because they want to hold on to the ticket. And so for me, it's nerve wracking. But here's the upside. Even if you don't get customers, you are sitting up on stage with somebody from Dell, with someone from Atlassian, with these big companies here locally. It gives you an opportunity to really talk to them, start a relationship with them. And then everyone else sees you on stage with them, which also adds to the halo effect. Right. It just rubs off from them to you. And then you end up with all these people who you invited, who feel grateful that you invited them and maybe are nicer reaching out so that they might get invited to the next one. They also know you as a guy who hosts, hosted an AI event. So maybe they'll ask you if you know of somebody else, a big, big upside to doing that. All right.
- [21:13]
B
I think so.
- [21:13]
A
So first network of people who you had worked with, then you started reaching out beyond. You went to events, you then hosted your own event. You're going to continue doing more events. Beyond that, what else worked for you? For getting customers?
- [21:29]
B
Yeah, I mean, I did. So I think when we launched, so we built like a SaaS product, like we wanted to get more in the private equity space. I had a big vision probably about like two years ago now, which was like, hey, like, if we can form relationships and provide value to private equity firms, AI is naturally a fantastic addition to their playbook. Right. When you think about their portfolio and like, what they're trying to do with those businesses that they're acquiring, and this is, you know, Taylor's oldest time, people always want to break into this stuff again. I had like, first mover advantage here. So I went to conference. I literally. But I didn't know anyone in private equity at that time. I literally. I am. None of us are from finance background. And so I just. I think I called outreach to somebody here in town who works at like, bridgepoint. He pointed me at this conference. I went to the conference and from that landed, like, our first and what eventually became our. A couple other pilot customers for our SaaS platform and just got started from there and like, went to more conferences. Ask them. I'm big on, like, asking people who they know that could also, like, be interesting to speak to. And. Yeah, and the whole time I'm really just like, how can I make this conversation valuable to them and, like, help them out even if I don't get anything from it? And so I'd say only like maybe 1 out of 10 conversations proves itself out for me. And I'm perfectly okay doing that because I like to talk about AI. Luckily.
- [22:55]
A
How did you describe what the company does? Because on LinkedIn you say custom AI and automation agency. I don't know if that really does it.
- [23:03]
B
You know, I've gone back and forth on the description of it many times. Most of the time when I'm talking, you know, more informally in person, I just say I run an AI engineering firm, which I think is, you know, I used to say consultancy. And people then thought that we were mostly doing advising work and strategy work, which is actually like less than 10% of our business. Like, a majority of our. Almost 80% of my team are engineers. And so I typically am. Like, we're an AI engineering firm. We specialize in generative AI apps. And if you. All the stuff that you Read about on LinkedIn and Twitter, as it pertains to LLMs, agents and stuff, we build that for businesses.
- [23:46]
A
Let me take a moment and tell you about my sponsor, Zapier. You're familiar with them, right?
- [23:50]
B
Of course.
- [23:51]
A
Here's. Here's what I've been discovering. A lot of people are now creating whole agencies based on Zapier. Like, what they will do is they'll go into a company and they'll say, I think you need this thing. You're probably doing it over and over again. How about if I create an automation for you that doesn't? And then on the back end, they end up having Zap Zapier do the work for them.
- [24:08]
B
And it's.
- [24:10]
A
And it's everything from internal trainings and feedback to actually, what was it that I saw? I saw one person who created a voice agent that did outbound calls for his customers using an automation. You've seen these people too, of course. I think it's interesting, I imagine that you're laughing to yourself saying, I don't think that this is going to be a long term success. It has to be developed like you do from scratch for the software to really hold up. Am I right about that?
- [24:38]
B
I, you know, that's, that's my thesis, that there is a class of problems and a class of businesses that are associated with those problems that require more custom solutions. But at the end of the day, who, who really cares? Like, who really cares about what it's made in or if it's custom or if it's Zapier? Like, the thing that you're should be buying is the outcome for the business. And if people are able to deliver that value with a Zapier automation, great. If that's the, if that's genuinely the easiest and most efficient way of generating that business outcome and it grows and scale, go do that. Like, I think that's, I think that's perfectly okay. I know, Yeah, I work the other way, which is like, who needs the custom stuff, right? And like, who, who can't work with like just a zapier? And then that's where we fit in.
- [25:26]
A
There's a whole class of agencies that are doing this. I've got to start interviewing some of them because in the back end, many of them aren't even saying that they're using Zapier in order to do this. They're just saying, I can do this. And most customers don't really care. I, I happen to be a nudnik who gets on and says, show me how you're doing it. Walk me through it. And to me it matters because I'm curious. But most people just care about the results. Whether you're doing this for yourself or for someone else. If you need AI automation, I urge you to go check out Zapier, my sponsor, and I'll tell more of the stories of how they've done this and how their customers are doing it as these podcasts continue. All right, what happened to, like, this, AI transformation? Nobody's saying that anymore. It used to be the companies like yours would sell AI transformation. I think it was like 10x Alex Lieberman's company that started out that way. No more.
- [26:09]
B
Yeah, we used to, we used to say it all the time as well. You did? I think, I think. It was a great, it's a great buzzword for sure. And it is technically true. Like I do think that most businesses need to transform into their AI native variants or whatever. The problem that that term lacks is all of the context and even the how behind how you do that transfer. Like how is that transformation accomplished? Right. And ultimately what I think people have come to understand, they've kind of gotten burnt out by that term because it's kind of a big black box with no instructions and no clear outcome for the business. And I think that's what people are starting to really get at. I mean you can see this even in our sales cycles today. Like when we first started, people were like, I want this chat bot. You know, I want AI, I want transformation, I want all this stuff. Like they, it was the buzzword era for the first like year and a half, two years I would say in the last year. People are now sharpening up their budgets, they're sharpening their pencils and they're like, okay, we know AI is effective. We know it's a hundred percent going to be transformative. How, like what are the specific workflows, outcomes, ROI, KPIs that we can look at and how can we justify this price tag for this implementation? Not just like another piece of AI slop is a popular term as well, like that one, like how do we, how do we avoid that? And so I, I imagine that most services based companies are not selling that because there is no one size fits all playbook when it comes to that. And it's really difficult for a customer to understand that ROI because it's so broad sweeping over an organization. Right.
- [27:52]
A
I've heard you say Also we build AI for you, we'll automate your AP and AR function, save you 50%. It's stuff like that. You're. Why are you smiling at that?
- [28:00]
B
Yeah, yeah. I mean I've tried, I've, I've just said so many things over the, you know, trying to figure out what works and what doesn't work. It's such a fast moving space. Both the company itself is fast moving and the AI space is fast moving. That I think the marketing term, like the marketing phrases that we're using is always evolving as well, very quickly.
- [28:18]
A
At the heart of it, it feels like you're a dev shop that uses AI to build and naturally you're going to include AI in your solutions because everything is going to get better with AI. Am I right?
- [28:31]
B
No, yeah, yeah. I mean I think that's fair. The. But I would, I would have just like reversed the order probably is like we are specifically looking like just to be Clear. We are specifically looking for what are the solutions that provide the most value and the outcomes to the customer. And we also happen to have this like beautiful flywheel where the solutions we're building have AI in them and we are also using AI to power the building of said solutions. So I would have just flipped the order. Small nuance.
- [28:59]
A
Give me an example. Like what's a way that you're able to build faster with AI and what's an outcome that you're able to give because AI is part of it? Be specific.
- [29:15]
B
Yeah, I mean they're not necessarily the same, the same answer. So I'll break it into like two parts. Like the latter question can be answered by what are the solutions that are now possible using large language models and this layer of sort of generalized intelligence that these large frontier labs have created that were not possible before. A good example of that could be feedback driven and like natural. So like for example, we worked with a he really big one of the biggest, like family law offices here in Texas. And you know, they wanted to build something similar to like a Harvey, but very specifically around their data and their sort of SME feedback, guiding a chat based system that they could use both internally as a training agent as well as externally to help new prospects and customers answer questions or new clients, I should say, answer questions about their divorce proceedings, for example. That's something that requires a high amount of reasoning, right. In order to provide the correct answer. And that's the type of knowledge work that was previously gated. There was no computer application system that could really help with that. Because the corpus, the possibility, the world of possible questions is so massive. Right. Like there's no way to, to hard code all those answers out. So that's something that we were able to build. That's a brand new technology and it's a very simple chatbot. The cool parts about it are how it collects feedback from the lawyers, how it improves over time, and how it can actually like grow with the firm as they expand the number of cases that they're doing and stuff like that. Internally we use all of the coding agents underneath the sun and we've created a really beautiful system in my opinion.
- [31:02]
A
Describe it. What did you create?
- [31:06]
B
What? What we essentially realized was that obviously coding agents and coding tools and AI was helping us produce more with an individual resource than ever before. Because the actual generation of that code, which was the previous thing that took the most amount of time, was now significantly sped up. Right. You can go into something like a cursor or a Claude and as long as you have the technical know hows about how that system should work and should operate and what the endpoints and everything should be, you can describe it and it can generate the code. And so really what's happened is that it took a coding and I think this is actually a pretty general principle. The time spent on coding was probably 80% generation, 20% review. It's flipped that into 80% review, 20% generation. But the key is, is that the actual amount of like the raw time has dropped by like 10x when you make that kind of switch, right. And so for us we have a, a bunch of agents that are specialized for some of the tasks that we do. So for example, we build a lot of chat front ends, right? And so we have a collection of small agents that are really specialized into building those front ends the way that we like to do them. And so when we say hey, for this application we want like a new, a new ui. It's using all of our historical context and things that we've built up over time, the rules and the guidance that we've given it to generate it in the style and the format and the structure and the layout that we like. And so that gives us a significant leg up every single time we start like a new project or a new engagement and we get a lot of delta from that.
- [32:39]
A
Isn't that what everyone's doing though?
- [32:42]
B
Oh, they should be.
- [32:43]
A
If they, if they what's, what's unique and special, then here essentially you're doing what everyone else is doing internal, like.
- [32:49]
B
The operations of my business. Look, first of all, I think, I think we, we have like a more advanced application than, than the typical firms have picked up on, at least in my experience. But that's a short lived win, right? In the sense that eventually everyone's going to have the same sort of like technical architecture behind it. The key for us though is that that's just helping my business run more efficiently. Ultimately if I don't create outcomes for the customer, then it doesn't really matter. And that's, that's really what like your question I think is like more relevant if we're considering and like a sale of my business, right? Like how would you value press w versus the other agencies in the world? Maybe over time that gap that we have, all of this accrual of skills and knowledge and agents might shrink because every firm will have it. So maybe for the right now my, my company could be deemed more valuable than the others that will get shrunk. But if the Outcomes that we are generating for our customers are much more specific. That's where there's differences in the businesses. Right. So like what I'm trying to do as a business owner right now is not stay so broad, not answer every type of AI question and requests that we get, and instead focus on very, very specific workflows that we know when we solve, have tangible outcomes for our customers. And I mean ultimately I think that's, that's what every business is going to do. Our specialization difference is going to be like how well we execute on that outcome. And that's just like a standard, you know, that's just competition.
- [34:10]
A
What's the focus for us?
- [34:13]
B
We perform really, really well in like highly regulated industries. So we have a ton of, just like our case studies and our examples are a lot in financial services, health care, legal, things like that. And so we've started to like niche down in there and then even to, to layer on another niche on top of that. We are really, really good at document intelligent document processing workflows where you might need a lot of escalation handling or you might need a lot of context about the way that that particular work workflow exists in that industry. And so that's, that's where our specialization comes in, is by layering those kind.
- [34:46]
A
Of factors together, I think I'm getting where this is. Do you think there's room for other people to do what you're doing now?
- [34:53]
B
Yeah, definitely. I think it's actually a pretty blue ocean space. I, I, I probably talk to a competitor maybe once every other week. Once a week, you know, once every. And you know, I just think, I think it's, I think it's really blue ocean like the original thing that we talked about, which is like AI transformation is true. I think that every business that exists today will have AI in their business moving forward. So I actually think that the opportunity space is expanding actually over time because Today most businesses have 0, 0 or 1 agent or AI integrated into their business. If that in the future I'm sure that they are going to have hundreds of agents running within their business. And that unlocks an entirely new class of problems.
- [35:38]
A
Wouldn't they be doing this themselves?
- [35:41]
B
Many of them, many of them can if they build up, if they build the infrastructure correctly. There is a, there's a tremendous amount of work when it comes to like the actual data management, setup collection, how that whole architecture works. But, and that's actually why, you know, if you see like the MIT report, 95% of things didn't 95% of like AI pilots never made it to production. I think it's because a lot of the people lacked the expertise on actually setting up the infrastructure correctly. From an agent perspective. Yeah, that gap will shrink. I think a lot of them can do it themselves over time.
- [36:16]
A
All right, I think I see where this is going. I think I've got a full picture here. Let me close out with this. Give me one specific example of what you're doing that shows your unique way of operating. Like the whole sales process, there's a whole automation that happens after somebody signs up. Walk me through that.
- [36:34]
B
Okay, so this is actually the part of my business that I'm perhaps the most excited about. There was a blessing in the sense of us, there was a blessing and a curse of us entering a services based business three years ago. None of us are consultants, none of us worked at Big four. None of us even come from that. We were all software people. 80% of my team is still software. In fact, I don't have any Big four consultants still to this day. I don't have any formal consulting type service people in my business. What I have is a bunch of like product people. And actually most of them are former entrepreneurs themselves. And so what that, what that gave us is it made us, it made it hard to build a services business because we didn't know how they should look and operate from like a professionalization perspective. But it gave us a new way of thinking about how a services based business could run.
- [37:21]
A
Because you're not trying to reproduce what you know, you're just creating whatever makes sense.
- [37:26]
B
Now we're just, we're just problem solving. We're like product people. We're like, okay, we have to run this business. What's the best way we know how to run this? Like we, we know how to build product and stuff. So like to, to, to answer your question directly for us, the way that it works is like we, we are building this like AI native. We, we call it Atlas internally. Unfortunately, ChatGPT stole that name and now that's their browser. But we have, we have something called Atlas. What, what, what this is is when you think about data and people in an organization, historically, data and people live right on top of each other. Meaning that if you want, if I wanted to ask you a question about bootstrap giants or anything that you're working on, you're probably going to go dig into some dashboard somewhere, you're going to go dig into an Excel spreadsheet somewhere or email or whatever and you live right on Top of that data, anytime that something needs to happen, humans need to go interact with data, pull it out, generate insights from there. AI is entering as this layer in between. And so what we do is, for example, when I have a sales call with a prospect, that transcript gets taken and automatically put into a series of agents that are taking all of the insights that I. That we've talked about on the call. Yeah, it knows how we like to generate our proposals and our statements of work. It will generate that with me often in the loop, guiding it and giving it different areas to focus on. It also knows what our rate cards look like, and it also knows what our current allocation looks like as a team capacity. Right. So as it moves closer to a contract, we can start to actually, our agents are the ones that are driving who needs to be assigned to this work, who has the best skill set for this work, et cetera, et cetera. Once the contract gets signed. And all of this is like basically being led through agents. All of the tickets for all of the work, for the project itself get created. All of that lives in a centralized kind of microsite that shares the documentation and the code together. So all of our coding agents have full contact. Our coding agents don't just have context about the project and the code. They have context about the very first customer conversation that ever happened all the way through to every playback that we're doing with them. Sprint playback, where we're checking in every sync, every email. They understand exactly what's going on in the project, and that allows them to plan, strategize, and execute far, far better than anything else. So, I mean, what I like to say is, technically speaking, I think we are approaching the point where our business could go from one sales call to a finished project with very, very little human in the loop. Basically just prompting and nudging and tell.
- [39:49]
A
Me how you're doing all this. How are you getting it all in there together, all that data, making it usable.
- [39:54]
B
This is the complex part. Like, so we have, we have. We've built some, like, custom kind of connectors that suck up the data from our Fireflies, which is what we use to record calls to notion, which is where we store kind of proposals and things all the way to, like, GitHub, where, you know, so there's different data repositories within the business. We've built small connectors that are able to suck that out as well as put it back in when it makes sense.
- [40:17]
A
Okay.
- [40:18]
B
And then we orchestrated all in our own little application that is largely built on top of Claude and Claude code both together. And yeah, all those agents are like, largely skills and things like that that exist in that sort of repository. But, yeah, it's definitely like the operating system. It knows all of the data, and we're just interacting with the agents to get the sort of outcomes that we want.
- [40:43]
A
All right, why is it called Press W?
- [40:45]
B
When me and my two partners started the company, we. We were all sitting in a. We were trying to think about a name, and we had just gone through a very painful process of renaming our last company right before our. Our acquisition. And so we wanted to just find something that worked. And the three of us came from playing computer games. Like, the reason that we all got into computers in the first place was. Was gaming. And on a keyboard, WASD is typically a lot of games move around like that way. W is always the forward key. So press W means to move forward.
- [41:15]
A
That makes sense. All right, turn right on. Thanks.
- [41:18]
B
All right, thank you, Andrew.