[00:00:44] Andrew Michael: Hey Gabe, welcome to the show.

[00:00:46] Gabe Larsen: Hey man, excited to be on.

[00:00:48] Andrew Michael: It's great to have you. For the listeners, Gabe is the chief revenue officer of Atonom, which builds AI employees for businesses. He was previously Chief Marketing Officer at Kustomer and held leadership roles at InsideSales.com with early experience at Gallup and Goldman Sachs. So my first question for you today is how many AI employees do you have on your team?

[00:01:11] Gabe Larsen: That's always the good question. Right now I've got six. mostly go to market AI employees. And some of work really well and some of you know, we're still we're still optimizing.

[00:01:21] Andrew Michael: Nice. And how did you decide which six to have on your team as the the go to market CRO? where did you get started? Obviously this is what you're doing and helping others do, but yeah.

[00:01:29] Gabe Larsen: Yeah, yeah, look, I think this is the debate, It it at first, if I'm honest, if you'd have talked to me, two years ago when we started this, I I I was one of those people that thought you could probably replace a lot of different roles with AI, and I've come to learn there's there's need for humans a across the board. And so that would be a big picture kind of point. But, you know, look, some of the more transactional stuff. So one of our better performing AI employees is. is our our sales development rep. people come to our website, they wanna, they, they wanna call us, they wanna chat with us, they fill out forms and Zoe, our AI employee, is just there 24 hours a day. Just you know, she's pretty good. She can talk to you about any channel you want and book meetings, answer questions, and kind of play that kind of inbound SDR. So that's one of them that's worked really well. one of them that hasn't worked very well is our AI sales rep. You know, if you're selling bigger deals like we are in the more maybe fifty thousand type numbers annual cost. yeah, that didn't work very well. So that I I would call that more of a failure.

[00:02:31] Andrew Michael: Yeah, I think that's also like I I saw an interesting company the other day, I think it's called Zobia, which is now trying to focus on EQ and actually sales being the specific use case because the current models today aren't particularly good at selling themselves. and so being able to understand nuances and conversations and how people are feeling at certain moments. I think it's an interesting one. Yeah.

[00:02:49] Gabe Larsen: Interesting. Yeah, there's a it's always changing, right? Never ne never never a dull day in the world of AI. But the if you're selling a strategic sale, you know, a relation, having an AI employee do that I think right now is not not gonna happen.

[00:03:04] Andrew Michael: It's not gonna happen. Yeah, it's still very difficult. but it it's interesting that you say that the started out that you in the beginning you would have thought that like most roles could be replaced and you're coming back on that now. is that like a constant thought or does your pendulum also swing backwards and forwards between releases of models and the way things advance? Yeah.

[00:03:23] Gabe Larsen: Yeah. look, I think this if I had a crystal ball, you and I probably wouldn't be talking. So y you know, it's just

[00:03:28] Andrew Michael: Yeah.

[00:03:29] Gabe Larsen: I definitely have come to a great appreciation of, you know, how do you find the balance of AI employees and human employees? And and that is a constant question, right? It's like I don't think companies in the past when we when companies grew, it was just people and tools and people and tools, right? One, one, ten, a hundred people, one tool, ten tools, a frank and second tools. In the future, new types of organizations will and are emerging. And I think you're gonna see human employees and AI employees partnering together to grow results, not just head counting, but the question is where? You know, I manage six AI employees right now. I could tell you some of the roles and but I think all organizations are trying to figure out where do humans work best, where do AI employees work best, how do you manage AI employees and how do you build this new type of organization.

[00:04:18] Andrew Michael: Yeah. And how did you so obviously you have six on the team now as well and you're managing them. I think one of the questions in my mind is wha what was the decision and why did you decide to personify them? 'cause I think it's definitely a s a different take. I think there are others out there doing it, but it's generally not like the status quo. I think most people are talking about agents now and you're giving them skills and so what was the idea behind like wanting to personify them?

[00:04:42] Gabe Larsen: Yeah, yeah. a a couple of things. One was you know, my background. I spent some time at Meta in a company called Kustomer and we we were early and doing some of the AI things. And look, I just felt like the the the basic AI is, you know, it's awesome. It's very task based. You know, like prompt something and have it go do something. And you you know, there's maybe a little chat bot, w one one single channel, the memory is very basic and So as we were developing our product and it was, more omnichannel, and you were you could contextually talk across all these different channels and it had memory across interaction and it just started to feel less like a tool and more like a teammate. And so there was just an aha one day where we were like, it was just that line. Somebody in the room is like, guys, this is not really like a tool. This is more like a teammate. And we were like, we should give it a name. And then truthfully, it just got out of hand. Then we started.

[00:05:35] Andrew Michael: Yeah.

[00:05:36] Gabe Larsen: You give it a name, you give it a big five personality, you can give it an avatar, you do a weekly training with it. All of a sudden it got you you hire it, you onboard it. And we've probably taken a little too far, but truthfully, with the advancement, and and you you know this, with the advancement, it it sometimes it feels more like a a a teammate, right? It doesn't just feel like a kind of static.

[00:05:58] Andrew Michael: And no, I think that's also why I asked the question because I've definitely experienced the same. a friend of mine, we build things from time to time. And like one of the things we built at some point was this internal Slack teammate. Uh

[00:06:10] Gabe Larsen: Yeah.

[00:06:11] Andrew Michael: and it was actually a dog. So we gave it like we didn't make it a person, we made it a dog and it had like a big five personality.

[00:06:14] Gabe Larsen: Ha ha ha.

[00:06:16] Andrew Michael: And it was actually it it felt like a much more interesting experience. Like the responses

[00:06:21] Gabe Larsen: Yeah.

[00:06:22] Andrew Michael: and the personality that came back from the agent, like It almost like made us I would say like pay more attention to it than just like a generic uh prompt response. Yeah.

[00:06:29] Gabe Larsen: Yeah, I I love that. I I love that point. Yeah. I look, I'd I'd I'd totally put a exclamation point on that. When when you treat it like a teammate, um,

[00:06:39] Andrew Michael: Yeah.

[00:06:39] Gabe Larsen: it does start to change your paradigm and the way that you interact and the work you do with it. And I think that's probably a overlooked but super important point. That's cool. That's super cool.

[00:06:48] Andrew Michael: Yeah. But yeah, so we we had a lot of fun in the end about it. And my friend was like putting certain ideas in its head and when I was having conversations with it, I was like, This doesn't feel right. Like Yeah. It was called Miles, Miles. Uh so yeah.

[00:06:57] Gabe Larsen: Did you actually did you did you name it or did you just call it dog? so nice, dude. Yeah, I have this one I have this one that, you know, I refer to him as Chad and he's he's like a used car salesman basically. He swears like a sailor. He's got so much attitude. Whenever I'm frustrated about when I'm reading on LinkedIn or something, I'll I'll talk to him and he'll give me give me a straight

[00:07:19] Andrew Michael: That's it. Amazing.

[00:07:20] Gabe Larsen: talk, you know, and stuff so

[00:07:22] Andrew Michael: Yeah, this was similar vibes. we went into some good details around personality. So when you had the conversation you felt like it was like uh there was something, there was substance

[00:07:30] Gabe Larsen: Yeah, totally.

[00:07:32] Andrew Michael: there. But yeah, it makes a big difference. I think 'cause I even see now at work we building at an insane velocity. We're going chatting to Claude all day and to other models, but like what you're getting back it feels very transactional and I think like the experience like the previously mentioned inside Slack, it was like it felt maybe a little bit more genuine, which is weird. It's like that it's a it's a feeling. It's not like any different the results that are coming back if we really look at it. It's just like the personality adds something unexpected. Unexpected.

[00:07:59] Gabe Larsen: It's true. That's right. Yeah. Yeah. A little more flavor. A little more flavor.

[00:08:03] Andrew Michael: So six on the team, which was the first one you chose to get started with? What was the use case?

[00:08:11] Gabe Larsen: the the the first one was again, I'm more in the sales and marketing realm. So the first one I went with was actually a signal-based outbound AI employee. You know, we're a f kind of a fast growing startup and we were getting a decent number of people kind of jumping on the website or checking out the pricing page. And I I rant a ton on LinkedIn. you know, don't connect with me if you don't want to get spammed with random crap. I'm on LinkedIn just a lot and I like to make, annoying comments and annoying posts. And so every once in a while I'll get something that, is gets a few comments, et cetera. And I was just like, man, I'd love to be able to take that audience and do something with it. and so that was the first one. you know, we call him Dex. He's basically like a signal-based AI employee. He just monitors LinkedIn and some of the different signals, like people who spend a decent amount of time on our pricing page. lets me know about those people and he'll reach out to them occasionally depending on, different kind of thresholds and stuff like that. So that was the first one I went with, but it was more of just a pain around some of our outbound prospecting areas.

[00:09:17] Andrew Michael: And does Dex go through performance reviews and w what does your process look like there?

[00:09:21] Gabe Larsen: I I mostly just I I try to have a weekly one-on-one with all my AI employees, and it sounds a little joke jokeish, but if you're kind of like I I I've got I've I've got a handful of real salespeople, for example. I've got a handful of real marketers, for example, and I do a one-on-one with them. And what do I do with my salespeople? You know, I look at their conversations, I look at their results, I look at what they've done and I give them feedback of what they should do different. And so with an AI employee like Dex or Zoe. It's the same thing, It's like, what did you deliver? I probably I might need to go in and give you some feedback. I want you to t attack a different signal. We've slowly added to Dex a couple different signals that he should monitor. And so, you know, call it we call it coaching, but it might be optimization, prompt optimization, you know, bot optimization. So again, maybe we take the personification a little too far, but you can actually talk to Dex and you can give them feedback and have a c so that makes it a little different. It isn't just a static prompt like GPT, I'll often just jump on a call with Zoe and and say, tell me about your performance this week or, let's go through some of your conversations. So you can interact with them on different channels. So maybe it feels a little more personification, but it really is just

[00:10:31] Andrew Michael: Yeah.

[00:10:31] Gabe Larsen: a AI optimization each week for kind of what they're doing in the job they're tasked to perform.

[00:10:37] Andrew Michael: Yeah. Yeah, I I think it's an interesting concept as well because for most people, even though this most of the skills are just MD files and they documents, I think still the way they are closely attached to code and stuff, they can be intimidating. But having a conversation with an AI teammate where they have a skill that can update their skills uh

[00:10:59] Gabe Larsen: Yeah.

[00:10:59] Andrew Michael: and based on your feedback and things like that, I think is like It's a powerful way to improve these systems without needing to be the one like sitting behind

[00:11:08] Gabe Larsen: And

[00:11:09] Andrew Michael: them.

[00:11:09] Gabe Larsen: and that, that's kind that's definitely one of the things I I was trying to help in the market. Like I I'm also not the most technical person. And I think in Claude you can get pretty deep and all of a sudden it feels sometimes overwhelming. And so, you know, I'd say probably 20 to 30 percent of our conversations are where people have built something really cool in Claude that they now want to scale it across. Dumber people and maybe across more people. And so they look to a tool like us and they say, hey, you know, could you take my idea and basically kind of help me make it, you know, secure and safe, SOC2 and HIPAA, but also scale it across maybe 50 sales reps or 100 customer service reps, because we don't wanna, you know, we we don't have the bandwidth and it's getting maybe a little bit intense. And we're gonna have people like you use it and and we want it to be a little simpler, maybe not so complex.

[00:12:03] Andrew Michael: Yeah. and you mentioned the compliance side of things. I think that is also something now that's gonna grow and more and more of a challenge for organizations. 'Cause

[00:12:08] Gabe Larsen: Oh my god, oh man, it's so weird. You know, I'd I feel like when when I talk to people, there's like three or four common questions that I get asked. I'd say number two or three is like, GDPR, HIPAA, SOC2, y you know, like the data

[00:12:23] Andrew Michael: You laugh. Yeah.

[00:12:25] Gabe Larsen: secure. And I think it's mostly because what we read online, you know, like just talking to a a friend the other day and he's like, What do you think about these reports of an AI agent doing basically going rogue, right? And it's like That that that's so out there that I'd say question number two is like, How do we handle security? I can't let everybody's ClaudeBot just hook onto our finan financial system. That feels a little scary, right? And I'm like, I don't know, I probably wouldn't do that. So

[00:12:53] Andrew Michael: Yeah. But I I think at the same time it feels scary, but I think also like a lot of companies today almost feel forced to be relaxed with these things because the competition is. And so like if your teammate is not like if your team is not actively like experimenting and doing things, they're falling behind. And I think there's like

[00:13:11] Gabe Larsen: Yeah.

[00:13:11] Andrew Michael: this weird paradox moment where like everybody knows they want to do the right thing, but they know doing the right thing is gonna like hurt speed and velocity and that everyone's trying to figure this all out in real time.

[00:13:22] Gabe Larsen: And man, that and that that point, I gotta admit, like

[00:13:23] Andrew Michael: And someone yeah, yeah.

[00:13:24] Gabe Larsen: we you know, if I act like on this show that I know what I I mean, I'm I'm just trying to figure stuff out myself. I got scars all over the place of where I failed. I've got a couple of wins, but that that is so true. I mean, we just don't we're we're learning. We're all learning here, man.

[00:13:40] Andrew Michael: Yeah, exactly. And so the You mentioned as well, like you you tried a couple, they didn't work. The one example is like the sales, and I think that makes sense where it's like a personal relationship and what have been some of the other learnings like putting together these AI employees for companies and like where would you see like companies getting the most success

[00:14:03] Gabe Larsen: Yeah.

[00:14:03] Andrew Michael: out of them to begin with?

[00:14:05] Gabe Larsen: Yeah, yeah, yeah. I definitely, we tried to build an outbound kind of prospect, like a cold outbound prospector on the sales side that was pretty a failure. it just was pretty hard. Like the personalization, the AI voice wasn't very good. There's some laws about cold calling people with, you know, electronic communication. We tried to build a customer success AI employee and we found that customer success is big. You know, like it's There's too many roles that you can't build one for customer success. we had to chunk it down into onboarding, support, outreach. So, anyways, I've I've definitely got a couple of failures, you know, from that you mentioned. Those are two that come to mind that, you know, are somewhat freaking. But as far as lessons learned on kind of like where where it's kind of working or where it's not working, you know, definitely the last thing I just said, I I really like when organizations think about chunking. and it is job-based. So customer success was such a good learning for us. If you want it to be like if you want an AI employee to be the CEO of a company, it's just too big. It's there's just what does that even mean? And so one of the things we found is ideally you take a job that's smaller. The second thing was if you don't know the job, AI will not be able to perform the job. So I definitely have had people come and be like, I didn't see one of your templates. but I was wondering if you could like, and then they just go through this like it's always fun. Just w what if an AI employee could like do this, this, and this, and you're like, are are you doing that? Do you know how to do that? And they're like, no, but I thought AI could like figure it out. No, no, no, no. AI can only take what you are doing or what you know and can kind of do it cheaper, faster, better. So that was kind of a a big learning. The third thing I'd maybe just throw out there is certainly transactional versus relational. We mentioned it, but you know, it really is those roles or the the the job functions that are transactional in nature. So if you're a customer experience manager and you're managing three accounts and they're all $10 million accounts, no offense, AI really isn't gonna help you. sorry. Now if you're a customer experience manager and you are trying to manage 40,000 accounts or 10,000 accounts. Woo-hoo, AI has a role and a place to really do it. The fourth thing I just say is AI is a toy versus AI is ROI. One of the things that we now run into every single day is just we call it Minecraft for professionals. We get on with people and they're like, look what I built, man. This is the coolest thing. Like And then once you pressure test it and you're

[00:16:42] Andrew Michael: Mm-hmm.

[00:16:43] Gabe Larsen: like, was there any actual business value here? And they're like, no, cost didn't go down and revenue didn't go up, but look how cool it is, man. I like can do all these things and it can research and it, but did it actually is there real RO? And boy, right now it feels like the ROI game is so important. And so we're really trying to just talk to companies like, I know you can build whatever you want in Claude from a Minecraft standpoint, and your world looks cool. But if you cannot go to the board, I'm t you can't go to the board with the Minecraft world. You need to go to the board with costs down or revenue up. So we really say don't play with projects that are toys anymore. And that's basically in a lot of cases, no offense to Claude users, but so much of that is turned into non-valuable toys. Not to say there isn't, but boy, I I'm like, we gotta get rid of that. We gotta get to our. So, anyways, those would be four things that are kind of just.

[00:17:38] Andrew Michael: Yeah.

[00:17:38] Gabe Larsen: Top of mind we're playing with.

[00:17:40] Andrew Michael: I like that analogy, the Minecraft for work. I definitely see a lot of that happening, a lot of cool cool stuff at the service level, but then Yeah, I'm wielding it's building it.

[00:17:46] Gabe Larsen: And I look I have it too. I'm not I I d I have it. I literally I got it everywhere. It's just it's hard. I've seen board meetings where they try to take these Minecraft products and and the board is like, You save twenty minutes? You

[00:18:00] Andrew Michael: Yeah.

[00:18:00] Gabe Larsen: like that that's that's and and it took you forty hours to build, like ooh. It's a little scary at times though.

[00:18:08] Andrew Michael: Yeah. but yeah, I think there definitely is like people doing exceptional jobs of this. And so I think the ROI like story that was happening like a few months ago, I think there was a paper that went around to say that no company is really seeing ROI from AI at the moment. Think it was a Stanford paper or something like that. I think that's no

[00:18:25] Gabe Larsen: It's hard.

[00:18:26] Andrew Michael: longer the case though. Like I think there is like there's definitely a huge gap between the people that are doing things well and the people that are not. But

[00:18:34] Gabe Larsen: Yeah, yeah.

[00:18:34] Andrew Michael: I don't think it's a question now whether or not like you can get an ROI from leveraging AI effectively in your company. Uh

[00:18:40] Gabe Larsen: Yeah. Yeah. Yeah. But it is kinda to your point. It feels like that's a more even a recent phenomenon. I think for a while we were just playing around with it and now there's been a big shift to just ROI

[00:18:51] Andrew Michael: Yeah.

[00:18:52] Gabe Larsen: positive projects.

[00:18:54] Andrew Michael: Yeah. And and there's more pressure now, I think, that uh the experimentation period is over. Like now you need to start to show like the results and

[00:18:56] Gabe Larsen: That's it. That's the pr it's the pressure. Yeah, yeah. Pro And I feel that. I you know, I was talking to buddies and a year ago it was like, as long as you could say we were playing with AI, everyone was happy and now it feels like the experimentation phase is over. Yeah.

[00:19:11] Andrew Michael: Yeah. What value is it delivering? Yeah. I 'cause you're too you're talking about the board, but I that those conversations were happening like a year ago. It's like, what is our AI strategy? What are we doing? And at that point

[00:19:20] Gabe Larsen: Yep. Yep.

[00:19:21] Andrew Michael: everybody was just scrambling to think I think, but also

[00:19:23] Gabe Larsen: figure something out.

[00:19:24] Andrew Michael: models have significantly got better now. So now it's at the point where it's it's no longer a model gap, it's a skill gap. And if you're not getting the results you need, it's you haven't invested the time to understand the tools well enough to be able to get there. So to your point, I think you mentioned as well like You need to know how to do the task or how to do the thing to be able to effectively wield and get good results from AI. And if you're just randomly trying to get new things out of it that you yourself don't understand well enough. And I think it's the old hiring process where when you're building a company in the beginning, the advice is typically go out and understand the role you're trying to hire for understand the pain points so you know what you're looking for, rather than just randomly hiring people on credentials or whatever without understanding what is the job you need to be done at your company because You can hire a salesperson, but sales is completely different in many different companies and the way you go about

[00:20:11] Gabe Larsen: Yeah.

[00:20:12] Andrew Michael: it and the skills you need to succeed in that environment. And so, yeah, without without doing the work, I think first up front it's hard to get good results at the end.

[00:20:20] Gabe Larsen: And i i you it's funny, as I hear you say it, I'm like, it sounds obvious, but you know, I think we all sometimes get that squirrel syndrome where you see the shiny object and you're like, Ooh, AI could fix that or or I'll just ha have AI do it and and to your point it's just not working. It's not really

[00:20:35] Andrew Michael: Yeah. I saw I I saw a really interesting post today. I think it was it was from like a lead engineer at Anthropic on this idea of ambition with AI. and it's like we th there's like four levels to ambition. I think like he sort of categorized it and I'm probably gonna butcher it now, but was like the first level was like speed, like and I think this is what everybody's doing now. Like can I do more faster? and uh we

[00:21:01] Gabe Larsen: Yeah.

[00:21:01] Andrew Michael: execute and I think that's like The basic level I think where most people are operating today is like can we get more done with less and and less time. and then the the next layer he sort of put was something around like expertise. Like you don't have the expertise, but are you challenging the models and using them in a way to be able to go that level deeper? to not focus just on the task that you want to get done, but focus on what is the outcome and then what other possibilities are that you're maybe not considering. So like being more ambitious in what you go to out to build. and then so he went through these like different layers of ambition, which I found very interesting because I think like most of the market today is probably at level one, which is like how do we get speed and velocity? And like I'm always like talking to the team internally is like we need to go back to to first principles and question like why are we doing the things the way we're doing them and like is there something that's now possible today that wasn't possible six months ago or a year ago and we're just accepting the status quo or checking what the competitors are doing from a feature? And so like I think there is this like reset that needs to happen in most people's minds. particularly in engineering, but I'm sure in other places as well of what we believe to be possible and what's currently possible. I think there's

[00:22:06] Gabe Larsen: Mm.

[00:22:06] Andrew Michael: a gap in perception in the market of what people believe and what is reality.

[00:22:10] Gabe Larsen: yeah, that's interesting. I might have to grab that post from you. it is, man, it's changing every day, but those are some pretty cool insights. Those are fun.

[00:22:16] Andrew Michael: Yeah, it is. nice. Okay. So the the company let's know a little bit about like the company today, where you guys at. I see you have some interesting customers on the site and you mentioned I think forty people at the start, but how things going? Where do you see the next six to twelve months going from here? Without asking for the crystal ball.

[00:22:31] Gabe Larsen: Yeah, look, I mean. man, you know, like I said, I no yeah, no, no, no crystal ball. But I do I mean, I like the concept, you know, we're a little broader in the sense that we decided not to just do one agent. We have kind of a platform where you can build, manage, and measure your AI employee workforce. So again, we have about 20 different AI employee templates right now that you kind of start with and and put into your business. But look, obviously everybody's move in that direction. I think we've been growing, you know, we just we're we're almost crossing a thousand clients here, hopefully in the next month. but

[00:23:05] Andrew Michael: Yes.

[00:23:06] Gabe Larsen: I think, look, I think the the game now, as we were just saying, is it's ROI positive. And so the big focus we've been changing to is, you know, how do you get these employees into a business and then really enact change? And as you were kind of indicating, it it feels like It's still it it it is and always will be people, process, and technology. And so we've been really doubling down on our, you know, Ford deploy engineers who have that business mindset and consulting mindset. Cause yeah, it isn't just throwing AI in there and automatically it'll now prove ROI. You have to understand the entire process, see where the weaknesses are, add in the agent or the employee at a specific point. make sure that it's prompted correctly. Then you have to overlay the humans on top of that. And so I think that is going to kind of that's where we're doubling down on. And then trying to find these roles in customer experience, that's where we've been going deeper. because I think there is so much ROI to be had if you have the right customer experience type of organization.

[00:24:16] Andrew Michael: Yeah, let's double click into as well because you mentioned and wanted to come back to it was that you started on customer success realizing okay, there's just way too many roles and functions in this to be able to reliably have one employee. how long into the journey did you realize this? And then like what was the change of course that you decided and how did you start to think about breaking things up?

[00:24:38] Gabe Larsen: Yeah, yeah. And look, I think this is great great learning for everybody. sometimes you think you can just take a big role like customer success and just be like, let's build an agent for it, and as I mentioned before, we were probably three months into that journey when it was like, Whoa, no, this is just it's it's too much. And so at the moment we've broken down the that post sales experience into about five core AI employees, and we've basically been doubling down on that. So that Most obvious one was support. And that is just, I think it's a we actually did a research study where people are getting the most ROI with AI and customer support came out number two on the list. Number one was engineering, but that one was kind of an obvious one, and we probably should have thought of that first. The second was onboarding, you know, bringing people on or bringing clients on and participating in the onboard experience. the third is We refer to it as customer engagement, but it's like long tail outreach or some version of a signal-based interaction where you've got, people going down, poor health, purchasing some signal or something that you need to basically interact and reach out. And then number three, we call it just an assistant. when we mapped the job, we basically found that a lot of CSMs, for example, spend a ton of time trying to schedule QBRs with their clients, and it's you'd be surprised at how much time goes back and forth with hey, we need to get this on the calendar. We can't do that. We got to include this person. And so so being able to have like a, you know, quote unquote assistant that can help you with some more mundane tasks, QBR decks, scheduling meetings, you know, getting somebody, you know, to do a health check with you. so we kind of broke it into that job. So anyways, that idea of breaking out bigger jobs into smaller jobs. And right now that's kind of where we are. I'm sure there's others that we'll find, but that's just been a big aha. And I think it's really unlocked potential to break it into smaller chunks and then go very ROI positive in onboarding. Like how do you take that job of onboarding, turn it into an AI employee and really support somebody so that they can the human can do more important things besides the same seven things in the first 30 days to get this company onboarded.

[00:26:55] Andrew Michael: Yeah. And then the customers using their AR agent to onboard themselves onto the new product, and then you have agents talking to agents when you're thing and the humans not really doing anything at the end of the day. Uh it's gonna get a little bit better. Like

[00:27:08] Gabe Larsen: it gets a little confus I said, I'm you know, I mean we're we're we're learning. If I act like I've got this all figured out, like I said, I don't. I mean we like I said, we've got some failures, we've got some some wins,

[00:27:17] Andrew Michael: Yeah.

[00:27:18] Gabe Larsen: but I think we're we're learning with the rest of them.

[00:27:21] Andrew Michael: Nice. and w what does retention look like then for these AI employees? Like, are they churning from the companies they join or they tend to be sticky? Like do you specifically see? Yeah.

[00:27:29] Gabe Larsen: Look, it's a g it yeah, like I said, that that's the nice thing with with AI employees, you you hire you know, we typically try to do month-to-month type contracts with these AI employees. so you you you hire for a month, you see if they do the job, and then you fire them if they don't. And truthfully, if I've you know looked at the number of AI employees I've implemented and the number that have been fired, there's been a decent number that have been fired. Like I said, we're we're we're learning here, you guys. And so, but I think that is an important that would be one big advice for people out there. If you are gonna buy an AI agent, I would never lock yourself into an annual contract. That would be crazy. That, you know, I hear about some of these things where, we bought an AI agent for a company and it was $100,000 for a year and I just are you kidding me? You know, like y I think you you you you use it for a month, you hire it for a month, probably should be more like five hundred to two thousand because it's a pilot, and after a month, if it actually is performing, you should you should probably keep it. If it's not, you should fire it. So I do think people have got a little bit caught up in the FOMO and you gotta you you gotta try before you buy, so to say. So

[00:28:43] Andrew Michael: Yeah. Especially with something like this, I think. And the sensitive nature. Nice. I see we're running up of time. So I have a couple of questions to ask every guest. first question is what's one thing you know today about churn and retention that you wish you knew when you got started with your career?

[00:28:59] Gabe Larsen: Yeah. Look, I think whew, that's that is a that that's a good one. I do I would say it churn has to be something that is proactively managed. I don't know. Sometimes I used to think that churn and retention was all about relationships and it was something that just more would naturally come and but I've just found great companies have a proactive strategy. th it's not It's not something, sometimes they have a customer success team and they have a customer experience team and they have a customer support team. I'm not saying you need to have a customer retention team, but boy, you have to have a retention strategy. It can't just be something that you kind of like let happen. And I think now that was if you were to talk to me 10 years ago, I think you I let retention happen and and now I say you proactively manage retention.

[00:29:46] Andrew Michael: Nice, I like that. And then maybe the final question is obviously I think we can take it in the context of building AI employees or agents and like in the context of that, what is one question that you wish more people would ask but they don't?

[00:30:02] Gabe Larsen: I think it would probably be best if more people started to question, you know, where where does AI fit and where do employees fit? I I think we we've gone too fast to the tool and we haven't looked at the job. again, it's that shiny object syndrome. So my board said I needed to use AI, my friends say LinkedIn saying, and so we immediately jump to I well, let's just use AI for customer experience. What does that mean? What does that mean? And you heard it

[00:30:29] Andrew Michael: Yeah.

[00:30:29] Gabe Larsen: from me today. You know, I I got probably caught up in that a little bit myself. I'll raise my hand. But man, if you can break it down into specific jobs and then think strategically about where do human employees and AI employees fit and and then start that people process and technology transformational project, I think you'll be way more successful. So, you know, ask ask first where AI and humans fit and then start to build it out from there. Don't just jump to the technology. Systems were never saviors. They never were, they never will be. AI is just the latest technology. we could go back and talk about IVR technology, the internet. The internet didn't

[00:31:03] Andrew Michael: Mm mm.

[00:31:04] Gabe Larsen: didn't make the internet itself didn't make it better. You had to you had to put a process around it. You had to put people around it. So I I I definitely take one step back, two steps forward.

[00:31:14] Andrew Michael: I see I really like that. I I think a lot of times we see when there's specific problems in organizational job that needs to immediately jump to like software or a product that's gonna solve

[00:31:23] Gabe Larsen: Oh, always.

[00:31:24] Andrew Michael: the problem. But more often than not, like the problem is more systemic and you need to really understand it before you can actually improve it. And I think definitely

[00:31:32] Gabe Larsen: And it's just it.

[00:31:33] Andrew Michael: like one of the areas is like task management. I think like the immediate direction, we need something to manage things and it we need X, Y, or Z, but it's like Most of the time it actually just works up people aren't using the systems the way they should, or there's someone being lazy, or there's and so like no

[00:31:45] Gabe Larsen: my goodness, it

[00:31:45] Andrew Michael: no new tool is really gonna solve the underlying problems unless you understand them well enough. Uh just uh yeah.

[00:31:49] Gabe Larsen: It's so but it's so much easier just to throw a new tool at it, know. But I'll tell you what, I I'm getting older and I'm like over one thousand just throwing tools at problems. So don't don't let AI be your latest mistake. take a step back and I promise one step back, two steps forward.

[00:32:06] Andrew Michael: Very nice, Gabe. so wrapping up today, is there any final thoughts you wanna leave the listeners with?

[00:32:11] Gabe Larsen: No, I love those last two questions. I think those were great to leave it with. Like I said, I think the r retention, you guys, is it can be managed. customer experience can be managed, but it has to be managed. And then as we said, AI is hot. I think AI can play a role in your business, but I'm telling you, if you just you think it's gonna be the way to manage retention or the answer retention or the answer to customer experience, you're you're sorely mistaken. Take one step back and I promise as you look at it more broadly, AI can play a role. I think there's a place for it, but you'll see that place and you'll see where people play and you'll see this process optimization. And that combination will actually get you to where you want to be, which is great customer experience and great retention.

[00:32:55] Andrew Michael: Very nice. Well, thank you so much for joining. For the listeners, we'll make sure everything we discussed today will be in the show notes so you can check that out there. but yeah, I really appreciate the time today, Gabe, and wish you best of luck going forward.

[00:33:05] Gabe Larsen: Okay man, appreciate it. Take care.

[00:33:07] Andrew Michael: Cheers.

‍