Token Utilization Is the New Timesheet

What if tracking how much AI your team uses tells you more than tracking their hours?

In this episode of KP Unpacked, KP Reddy and Nick reveal a controversial management shift happening at Zero RFI: KP monitors enterprise Claude analytics and reaches out to employees with low token usage, not high spenders. The new performance metric isn't billable hours or output volume. It's curiosity, commitment to learning, and willingness to experiment. Someone burning through credits is building, iterating, testing limits. Someone avoiding the tools is resisting change. And if the CEO isn't in the top third of token usage on their team, they're failing at leadership.

The conversation unpacks Zero RFI's first internal hackathon: seven hours, cross-functional teams pulled out of silos, non-engineers shipping production code by end of day. One team built a preventative maintenance prediction system for a business they knew nothing about. Another deployed a Slack-to-Notion content aggregation engine an hour after presenting. The philosophy? More is better until better is better. Give people space, support, and freedom to build. Then track whether they're actually using it. Nick raises the scar tissue transfer problem: how do senior execs pass decades of decision-making lessons to junior associates without endless meetings? The answer lives in skills files, transcribed Notion calls, and treating Claude as a training partner, not just a task executor.

Key questions answered:

• Should you track employee token usage as the new performance metric?
• What happens when you reach out to low token users instead of high spenders?
• How did Zero RFI's internal hackathon work, and what did people build?
• Why is $30K/month in token spend an easy ROI decision for some CEOs?
• How do you transfer decades of institutional knowledge without one-on-one mentorship?
• What's the difference between using Claude for deliverables vs. training?
• Why are skills files the solution to IP leaving the building when employees quit?
• Should seed-stage CEOs be coding alongside their CTO or delegating?
• Why did PE firms decide San Francisco proximity matters more than New York headquarters?
• How do you codify scar tissue and lessons learned into persistent company memory?
• What should CEOs do if they're in the bottom third of their team's token usage?

If you're managing a team wondering whether to limit AI spend or incentivize experimentation, trying to scale institutional knowledge beyond senior leadership, or questioning what productivity measurement looks like when timesheets become irrelevant, this episode will reframe how you think about performance in an AI-first organization.

Listen now.

Transcript

SPEAKER_01 0:34

Hey Nick, how's it going? What's up, KP? Nicole. Hey Nick, how's it going? How many people do you think are watching the video version of this podcast?

SPEAKER_00 0:44

Um.

SPEAKER_01 0:47

Much fewer than uh the audio, I think. What's that? I think much fewer much fewer than the audio. Otherwise, we'd get a lot of comments on um setups, how we're looking that day. I don't I don't hear a lot of like, hey, you're like last week was really rough. Like, what was going on? You know what's really funny, right?

SPEAKER_00 1:06

Our industry is so weird. Everybody's like stalkers. Like, there's no comments in the YouTube, they can't bother themselves to hit the little thumbs up button, like nothing, right? And then I had one of our LPs, I was like texting him, just checking in with him, right? And he's like, Yeah, man, I'm like on vacation. I'm like, cool, everything going okay. Man, I've been loving yours and Nick's pod, but was just listening to it. Yeah. I'm like, how about you like mash a button or something, man? Like, give us some love. But I get that all give me, give me a signal that you generally like what is going on and the yeah, based on the engagement data, it's like me and you were the only one listening to this.

SPEAKER_01 1:57

I do get random messages from people from time to time, and it's always and usually the messages that I that I get and that I like to receive are like some very specific comment that was made, yeah, in the pod. Yeah, and so it's like, all right, you were like you were listening at minute 41 and you had a pretty good question on on that topic. I actually did, I'll I'll probably save the question for when we get to the unpacked version of the pod, but one of our LPs sent me a follow-up, I think it was an email that you sent, but it was something we covered in the pod last week, which was the when the when the IP leaves the building conversation. And he and he was he just he he sent me like five questions about it. Like there wasn't there was no context. He was just like, Hey, I want to figure this out. Here's five questions that I have about how to solve this problem, and so we should cover that just listener mail.

SPEAKER_00 2:52

Listener mail, that's amazing. Finally, finally, it has I will it's it is kind of interesting as we've grown the the team here at zero. I have a lot more people internally listening than ever before, not just you and Matt. Of course, I don't think Matt ever listened, but Matt definitely doesn't want to hear us. He doesn't want to hear us, he just enough of us live. But it's been interesting, like some of the questions I get, like, hey, you said this. I'm like, wow, you actually listened. And then last night at dinner, my father-in-law brought up like that he had listened. He's like retired, he's not from the industry and he retired. You know, he's retired. He thought the conversation we're having about like optimism around Artemis and all that, he thought that was like really interesting to him. That you know, because I I think for a lot of people in his demographic, they're spending a lot of time on Facebook and on the internet, and it's just really gross out there. If you if that's your only point of reference about what's going on in the world, definitely not a lot of optimism.

A 24-Hour Phone Shutdown

SPEAKER_01 3:58

Totally true. Yeah, totally true. I've for some reason that I think that conversation last week we had around Artemis, and then we talked a little bit about just how much information is flooding us with with social media on previous pods that we've talked about, like the cognitive overload. I actually did a I did a Friday night sundown to Saturday night sundown, no phone. It was the first time I've done that in a while. Like I've done like a night to morning thing, but yeah, completely turned off the phone. I mean that my whole family did it. So Sarah, obviously, oh my two boys didn't do it, they don't have phones, they're four and two. But we did it. We just like were like, okay, we're we're gonna shut it out, shut it off for the next 24 hours. It it was like kind of a it was a trippy experience.

SPEAKER_03 4:47

Yeah.

SPEAKER_01 4:48

And in some ways, like, you know, I mean, I showed up to work on Monday, I felt like my brain was actually fresh because I wasn't just consuming, scrolling, scrolling, scrolling. But I almost looked at not only was I fresh, I looked at I looked at information differently. And I just found myself a lot more focused because I wasn't like it, it it broke the cycle. It broke like the addictive cycle of like, okay, I need to spend the first hour of my day consuming and you know, reading the newspaper, wherever that is. And a lot of the headlines are, of course, are not great. So yeah, that was a that was a unique thing.

Screens In School And Lost Learning

SPEAKER_00 5:24

Yeah, that's pretty cool. You know, it's I mean, it's it's like super interesting. You know, there's a lot of narratives right now about school, you know, taking phones out of the school and that kind of thing. And the school a three-year-old is going to go to in the fall, they actually have a no-phone policy, I think, up until high school. And in high school, you're allowed to bring it, but it has to stay in your locker. You know, so I think it's it's interesting. I was listening to you know, the guy that started Khan Academy today, and he was talking about like how we've we decided back during you know, laptops and internet that it was really important to have computers in schools. So, like so many schools have these, you know, little carts with Chromebooks because we decided that that was most important, and then all the curriculum went on to the laptop, and so now like kids can't even do homework without having a computer, yeah, yeah, right. And so now we're sitting here as parents going, I don't want a computer, I don't want my kid anywhere near computers until they're whatever, and it's not part of the curriculum. It's like, well, then your kid's not gonna be able to do homework if they don't use the Chromebook.

SPEAKER_01 6:32

Yeah, totally. The other the other thing about that dynamic that's interesting is the if the interface where they're learning information is a screen, it's a it's essentially, you know, it's it's a it's a very small two 2D screen that is not does not extend that you can annotate if you have the right pencil, but if you forget it, you can't. Like I just remember like the richness of a textbook and being able to write in it and annotate everywhere and you know, draw connections and you know, and and turn, you know, turn the corners of pages over to remind yourself, like, oh, I need to go back here. I feel like there's gonna be something lost in terms of like the ability to like just learn from a physical thing. And um, I think like I I when I read, I I find it kind of hard to like I definitely would prefer a physical book. I just feel like I learned better.

SPEAKER_00 7:21

Yeah, yeah. I've been I've been buying some of the books from our friends at Strike Press lately. Yeah, they're so good, yeah. Oh, they're so good, right? Like, uh I've got Elon Gill's book, which is great. And then I forgot what's the other one. What's his name? You know him. The Brian Potter? Yeah, yeah, his book.

SPEAKER_01 7:41

Yeah, I've got it. It's like their covers and everything is so crap handcraft, just done well, right?

SPEAKER_00 7:49

Just done well, yeah.

SPEAKER_01 7:51

For sure.

SPEAKER_00 7:52

So no, I mean, I think look, I think I think that's the the stuff everybody's trying to figure out right now. You know, how do we want to raise our kids? And AI is not going anywhere, robots aren't going anywhere, and you know, yeah. But it was really interesting was the Khan Academy guy was saying that they're doing a new AI certification course, and he said very little of it has to do with AI, it has to do with communication skills, human skills. Love that pretty much all the things I wrote about in my book two years ago, by the way.

SPEAKER_01 8:24

The book the book was the was the prophecy, the original prophecy.

SPEAKER_00 8:29

Yeah, wow. I was just like when Jack Dorsey said middle man AI is gonna get rid of middle management. I was like, dude, you should read my book two years ago. I already said that.

SPEAKER_01 8:38

Stole my ideas.

SPEAKER_00 8:40

Yeah, get no attribution from Jack Dorsey.

SPEAKER_01 8:43

You're rich enough, Jack. Just stop pilfering ideas.

SPEAKER_00 8:46

Send me a check. I'm easy to find.

SPEAKER_01 8:48

Royal C fees. Come on.

SPEAKER_00 8:51

Send me a link.

SPEAKER_01 8:54

Um on the AI topic, it's a good segue. I know in the past week you did something new at zero. I think I think this is something you did for the first time, which was an internal hackathon. Yep. So what was the what was the setup? You just basically said, hey, we're not doing any work today. We're gonna fully focus on ideas and concept concepting and prototyping new ideas for for products and internal processes, and basically like, you know, sort yourself into teams and and and go build and at the end of the day, come present.

SPEAKER_00 9:29

Yeah, I mean, Barry kind of Barry put together a pretty good framework, which I think was, you know, you know, your when your CTO is trying to build a framework, it could be too constrictive. He did not do that, he made it pretty wide open, just like guardrails. So it was it left a lot of room. But no, we had everyone take Friday to do the hackathon. It ended up being like seven hours, right? We also required vendors, like consultants and people like that that were spending more than 20 hours a week with us, they had to do it too. And it was everything from my executive assistant to marketing to HR, like everyone had to do it. And they weren't allowed to be on a team in their same functional group. So they all they had all put up leading up to this, everybody posted a bunch of topics on Notion and people picked teams, and of course, like three minutes before everybody moved everything around. It's it's like, you know, but you know, it was everything from building custom Slack bot notifications to Notion automations to writing code and to building code, and I'm still kind of on the hackathon high. You know, it's not even it's it's almost a week now. I'm still kind of jacked up about it because one people build very usable things and they're still working on them because they they we gave them the time and space to work on something in a very focused way. They all had access to the engineering team if they got stuck, right? In fact, it was so successful. Barry had to set up a new GitHub GitHub repo instance just for people, you know, like, hey, let's keep that totally separate. Yeah.

SPEAKER_01 11:22

I was gonna say, I don't know if it was so successful, he might have he might have not created a a new a new code base.

SPEAKER_00 11:30

Yeah, so he set up he set up a new repo for for all the hackathon stuff and all the you know for all the non-engineering people. So I think it was a few things. One is just honestly just giving people the time and space to learn, get support, and work on a project, you know. And I think a couple months ago on my Substack, I said, hey, give give people Friday off every month and run a hackathon. You you'll be you'll be surprised. And that was like me just you know, being me on Substack. I was like, well, let's do it, right? And we did actually do it. And people are still working on their projects, right? It wasn't like, oh, that was fun, you know. The only feedback I got is like, hey, isn't there usually a prize for a win? Didn't this one win? I was like, man.

SPEAKER_01 12:16

Of course, one of the people did you order and and beer delivered to everyone on the team?

SPEAKER_00 12:22

I was like mostly just you know, some skier yogurt and some berries all day.

SPEAKER_01 12:28

I've got packs, packs of athletic greens in the office. I'm handing out.

unknown 12:32

Yeah, there you go.

SPEAKER_00 12:33

It's like full of we're a little different, we're we're just a little we're built different around here. So no, you know, I was willing to be back. Uh you know, hey, isn't there supposed to be a winner?

SPEAKER_01 12:48

Do you think you would you just have the teams present? There was no, there was like no Barry wasn't the judge or anything.

SPEAKER_00 12:53

No, no, they just presented. Barry actually participated.

SPEAKER_01 12:56

Oh, cool.

Building As The Fastest Value

SPEAKER_00 12:57

So no, I mean that's what we had full participation. People played outside of their, you know, you know, we have the company Building Works that does project closeouts and all that, and we had a group that knows nothing about project closeouts, there's not from industry, create a whole preventative maintenance prediction system for building works. I was like, what do you guys even know about this? Right. I mean, because you you kind of feel like people are gonna stick in the realm of what they know, and they're like, Well, we didn't know anything about it. We spent some time with Claude talking about it, and it turns out that this is a thing, and so we built it. So good, right? So even playing out of their swimline, but you know, the fact that this stuff is kind of moving forward, you know, at different paces. A lot of like the actually one of the Slack the Notion, because we have a lot of Slack channels and a lot of content. One of the guys built a aggregation engine in Notion for all of our Slack channels and getting rid of a lot of the noise, right? Just really just getting the content out of it versus the you know, the thumbs up and the yeah, it sounds good, but like just pulling the content parts into it. He had that deployed by the end of the day, by like an hour after Hackathon, he had it in deployed. So we're seeing people moving forward with a lot of these things, and you know, I think the lesson there, you know, what I kind of explained was like the reason we're doing this is to create enterprise value, right? And that that's like the North Star. How do we create enterprise value? And I think people kind of understood that that you know, you can create enterprise value by doing a good job when you come into work, sure, right? But if you're building stuff, you're really building enterprise value, right? When you start to build stuff that does your job for you, that that's your contribution to enterprise value is much more than the time you spend like doing a good job, right? I mean, clearly people have to do a good job, you gotta do your job and do it well. However, when you think about like building enterprise value, you the easiest way to do that is by building.

SPEAKER_01 14:54

Yeah. If if I mean I guess, yeah, I was I was gonna say if if product and software is is where the IP lives, then anytime you can be anytime you can be building and concepting of new ideas and great features that your customers will love, it's obviously where you want to be spending time. But I even I think that's actually that's what I that frame is wrong. Today, even if you're a services company where your core identity is actually just doing is is is giving your customers an outcome and performing the work, still the function should be to build product that supports better outcomes and a better service experience that's gonna be oftentimes through software, potentially maybe maybe hardware is integrated to that. But it like it feels like we're in a new a new era where everyone can claim that spending time building product is actually building enterprise value. That's uh that feels unique.

SPEAKER_00 15:54

Well, I mean, I think you were you'd referenced that Substack thing about like IP leaving the building, which is what private equity people have always said. We don't invest in services companies. I'm sure we probably said it as VCs. We don't invest in services companies because the IP leaves the building.

SPEAKER_01 16:09

We definitely said that as a VC. I can confirm that.

SPEAKER_00 16:12

Okay, we also don't invest in marketplaces. That was the other one, right? But however, with AI, is that really true? And you know, so think about this, right? Like, think about like skills files. You hire a consultant to build your messaging and you tell them, look, the deliverable is not going to be a word doc or a deck, your deliverable is a skills file, right? So it might only be like a I might pay a 10 grand to build this messaging framework, but once you codify it and put it into my system and it can persist and learn, I kind of don't need you anymore. So if you think about that, like one of the biggest things you ever see, you always see in sales, right? Sales is notoriously a high turnover business because it's performance driven, right? The number the numbers don't lie, you know. And every time a salesperson leaves and the next person comes in, the first thing they do is look in the CRM and like, hey, what where are the notes on this prospect? What's the background on this prospect? Now, if you think about how you capture all that data within the CRM, within Claude, within Notion, you don't like you don't need a transition. So the point being is that as people go to contribute to their companies, a lot of that contribution may actually be much more codified in either skills files or in you know software, like whatever it is. And then what does that do to them, right? It's like you're there's not that repeat, oh, send that deck to that to our messaging person to have them review it every time, right? There's no that ongoing maintenance of that skill doesn't exist anymore. So that person has to go do the next thing, right? So I don't need someone to pull an AR report for me or an AP report. Like we don't need any of those things anymore, right? Once it's built out, it's good to go, right? You don't you don't need the person anymore. So what that leaves is that you are actually absorbing the IP, much like we've done always in software and in marketing and strategy, right? But what happens when McKinsey shows up and you pay them their million bucks, right? They show up, they do all their work, but you tell them the deliverable is not a deck. With the deliverable and skills files and the decks.

SPEAKER_01 18:43

Can you can you clarify for people what you mean by skills files?

SPEAKER_00 18:47

Skills files were built, it was basically developed by Anthropic and now it's used everywhere. And you can basically build them based on it's almost like memory for for your AI instance. And we run enterprise, right? So I have everyone's memory right in our system. So, you know, so essentially if you're like, hey, here's here's our messaging framework, you load it into it and you say, like, build me a skills file of this. And what it is is it takes that all that information and basically puts it into a markdown file that persists in memory into the system. So the next time someone says, Hey, I need to write this article, it'll be like, Great, I'll use our messaging framework, right? So it already knows it in memory. So think about it, it's almost like storing a series of prompts in a in a state, right? But it's for everyone, right? It's for everyone to use.

SPEAKER_01 19:45

And it's and I think one of the important things is you don't need you don't need anyone to engineer the skills file. Other it's it's basically knowledge transfer and stored memory in your overall LLM instance with whatever product you're using. If it's open AI, anthropic, you can build an open source model, but it's literally just like a it's a so a markdown file be comparable to a PDF. But it's but the it's a list of instructions that tells you what to do and what not to do when you're when you're performing a certain task. And I think one thing that's like one easy way to use it is whenever you're performing that particular skill, you just basically tell them like it's like, hey, you know, I need to update a pitch deck or you know, our board, our board deck, and um, I need you to just to reference this skill when you do it. And it's basic and it it makes sure it makes sure that um it's gonna follow those instructions.

Token Culture Starts With The CEO

SPEAKER_00 20:43

Right. So think about that. Like we're fanatical because we're AI first, right? So everything has to be in Notion, everybody has to be on our enterprise cloud account, on our enterprise git repo. Like we're very strict about those things, right? Now we give people laptops that are locked down, kind of very IT cyber-y, right? But think about all our AEC friends that are not buying enterprise versions of any of this stuff. And so what their employees are doing is getting their own personal instances of Claude or OpenAI or whatever and paying for it themselves, and the company's like, yeah, it's too expensive, it's$200 a month per head. And it's like, but you're allowing your employees on your time, put the company's IP into their instance, and you're not going to benefit from it. So if that you know, if your entire wastewater department picks up and leaves tomorrow, if they've all been using the tools and you've implemented them well and you've done it the right way, if they walk out, what you're doing is actually cap the department still persists, right? Yep. And and then when the new person comes in to build to build up your wastewater treatment plant, it's like, hey, where do I start? It's like it's just prompt what you need to prompt, right? And so I think companies have to start thinking about that IP angle because they a lot of it they think it's like, oh, well, the IP lays you know lies with the person and their expertise. I'm kind of saying, like, yes, and that expertise can walk out the door unless you keep them on your corporate platforms. So it's it's it's super interesting to get people to understand these things. And you know, our hackathon was like amazing. So we decided for our Q2 event in May, which tends to be more like Mastermind roundtable, we're just gonna do a hackathon. Like I think that's just a new format. I mean, the team building was amazing. People got to know each other. I mean, when do you have time? When you're we're running remote teams, when are you and I ever gonna spend seven hours together remotely, Nick? Never happened. Never, never, right? And so now everybody spent seven hours together working on a project, getting to know each other, right? I mean, massive team building effect, massive knowledge transfer effect. And then we actually created enterprise value in a day, right? So I said, like, hey, our Q2 event, May 5th in Atlanta, let's do a hackathon. And what's funny is I got several people reach out to me. They're like, hey, that's the same week as ENR. I'm like, yeah, that's the same week as ACEC. I'm like, yeah. They're like, why can't come? I'm like, oh, so you want to go see people talk about AI? Is that what you're like? That's cool. Maybe you're you're at that level of development that you're in the level of development to watch people talk about AI and schlep their products, or show up with me and I'll put you on the keyboard, right? You're gonna spend all day banging stuff out. And I think if you're serious and want to understand how this stuff works, you got to be on the keyboard. Yeah, I mean, I actually don't think and there's no reason not to, right? In terms of, you know, as much as like I'm kind of semi-plugging my event, I'm just saying, like, get to come spend a day with technical resources, build a team, meet new people, and walk away with something, right? Walk away with something. And so if you think about that, so we're we're what's we're April 16th. Hopefully, everybody paid their taxes yesterday. That was a very painful day for me. Always is. But ViBathon was end of October last year. I'm not sure that any of those projects got moved forward, right? There was a lot, you know, there was a lot of lovable stuff, you know, quick prototyping. There wasn't really an understanding of path to production with those ideas. Probably most people in that room didn't know what GitHub was, right? So there was just a lot of proof of concept stuff running local. And I think now if you just look at the level of experience and knowledge that people have that are that are spending time on the keyboard with some help, we're actually building stuff that can go into production. If if not in production for your team, in production for yourself, right? So I built this dossier program because I have so much data on so many people and so much in the industry that our sales team can basically right now. They have to ask me, we haven't deployed it yet, but they ask me, like, hey, I haven't meet these meetings this week. Can you pull a dossier on the prospect? And they get like a three-page PDF of everything about them, what projects they've worked on, who they've worked with, what their upcoming projects are, and what's what could be some of the critical challenges around that. They show up very well prepared, but you know, that that's just running on my local machine, that's not in the cloud per se. So, I mean, you can run stuff locally. So I think, which by the way, I've been loving because on some of the stuff I'm building, I'm like on V7. And I think about it and say, like, if V1, if I would have handed it to Barry and team and I would have had to do seven iterations involving them, I mean, I'd be embarrassed. Sorry, guys, here's another version. Hey guys, sorry, you know, so I think that's what's kind of cool because I think you can be a bit of a stream of consciousness building stuff and realizing, oh, why did I do it that way? No, I should have done it this way, right? You learn over time, and saving your engineering team from every thought you have is is a great thing.

SPEAKER_01 26:32

Yeah, love it. Yeah, it's pretty, it's pretty clear even since Vibathon last year, which was what in was that in October? Yeah, it's clear how much better it's six months, man. Yeah, I mean it's just like yeah, kind of kind of night and day in terms of what you can, yeah, how how how much further you can get the product if you're non non-technical. There was a there was definitely a limitation. I think those tools like where they shined, they were great at concepting. They really help, they really helped you get like an initial concept and a prototype out, but a lot of the actual backend engineering work wasn't wasn't great. Yeah, and so they would still and they would stall out. And now the the coding tools at the language models are so good that you can actually get very, very far one or two shots, and then you version it out, keep iterating. There's it's not it's certainly not going to be a complete product that first time, but yeah, you keep you you keep you keep you keep iterating on it, and sure enough, like you will find yourself with a good product that you can send other people.

SPEAKER_00 27:34

Yeah. I mean, you've heard me say this before, like more is better until better is better. And I track on analytics, our cloud account, our cloud enterprise account, and I'll reach out to people if their usage is low. It's like, dude, what's going on? Right? Whereas I think some of some some of our industry people might be like, hey man, you're using too many tokens. Knock it off, right? I mean, it could be that the it there could be a correlation to the curiosity of your team and their commitment for change based on how many tokens they're using. Now it could also mean that they're just like screwing things up, right? I'm not saying that it's perfection, but would you rather have someone overspend and ask for more credits than a person that never even touches their credits?

SPEAKER_01 28:22

Yeah, definitely the the former.

SPEAKER_00 28:24

Yeah. Because we do have one guy that's a little bit he's overspending, right? So, but all that did is flag the engineering team to say, hey, hey man, what are you doing? Maybe maybe there's a maybe don't use the expensive model for everything. Maybe you know, like that, like let's tweak it a little bit, right? Yeah, dude, I'd rather have that problem.

SPEAKER_01 28:46

I mean, look at that, yeah. I think I think like you if the more you can anchor your your entire team and incentivize them to use more tokens just for the behavioral effect itself, yeah. Like worth the additional, you know, hundreds of dollars or thousand, you know, thousand dollar, thousand dollars a month to just to get them using the tools. And I think over time, to your point, yeah, if they're not if you can't, if you can't trust them with their token usage, it's not adding adding value that is measurable, then you can dial them back and put limits on it for sure. But I think I'm uh yeah, I'm 100% in the camp. Like get people using as many tokens as you can right away. Like, don't lose them.

SPEAKER_00 29:26

Who do you think's using the most tokens over here? You yeah, absolutely. Yeah, and I would almost say if you're a CEO listening, this is your job. Because my ability to build stuff and codify thoughts, strategies for other people to use. Yeah, I'll tell you one of one of our people that the other day uh was pretty funny. They wanted me to review something, right? Some analysis. And I've just been busy, I've just been so busy, right? And I'm like, I'll get to it, I'll get to it, I'll get to it. You know what they did? They took it, the analysis, and pointed it to my sub stack and said, read KP Substack and give me feedback on this as if you were KP. And they were blown away. They're like, What? Like, it knows you. And I'm like, I mean, like, I'm pretty predictive, like it, there's no new news with me, really. It's just like you've got a lot of artifacts out there for sure. So many artifacts, right? And then I was like, Well, do you still need to look at it? And they were like, No, yeah, turned out great. So think about that. Like, but if I'm not putting my content, my artifacts out there, then how does the company continue to like execute and scale and think about things, these things? So I think you know, in the world of, and you know, I'm a terrible manager, I'm a good delegator, terrible manager, but on the leadership side, I do okay, right? But I think if you're a CEO and you're talking about leadership, you know, you don't have to be top in the in the in the cloud analytics like me. But you should you should be somewhere. You better not be in the how about this? Everybody listening should commit to not being the bottom third. It's a good goal, yeah. That's fair. Just don't be in the bottom third. And I know most of your IT people, so I'm gonna get a report.

The Real Cost Of AI Speed

SPEAKER_01 31:15

Just like uh yeah, a quick second on the economics of token usage. So there's a a friend of mine is the C C he's the CEO of a software company, and most of their development today is done via coding tools. They have a team of engineers, but they're like a hundred percent of their code is basically being written by Cloud Code and Codex. And I think so he's told me his daily token spend is over a thousand dollars a day, like like extra usage, right? So just do the math on that. So, like some days he's probably going over a thousand, but for round numbers, it's thirty thousand dollars a month just from his own token usage, not anyone else's. And it's like in in his mind, it's it is like the easiest choice, would spend the money not you know, a hundred times out of a hundred, because the product the production and the speed at which they're at which they're able to move due to that to that token usage is just is it's just so so worth it. It would take, you know, yeah, it would have taken months to build the same, you know, this the same amount of product with you know it with additional hires that are difficult to find. And so it's just like the easiest choice in the world.

SPEAKER_00 32:31

And it's but it's a lot of money. No, it's look another, you know. Uh I was giving this talk the other day, and I was like, you know, we we all know that the early days of Uber were subsidized by venture capital. Totally. That five dollar ride, they weren't making money. The VCs were funding that, right? And I'm unsure, you know, it's it's unsure like how much they're making on our token usage right now. But we have the two the two guys, like you know, anthropic and openai. I mean, they're talking about going public. It's a different level of scrutiny around how you, you know, how the investors think about what you should be doing. And if they decide to be wildly profitable, they just bump up the token costs, right? Yep. And so, which that's you know, so so I think what's pretty cool is you know, I think you should track token costs and usage. I think you should also look at how do you you know some of these open source models, right, to to run locally, but you know, not now, right? That I was like right now, more is better, is my advice, right? Once you start getting some understanding of things, then you know, then worry about you know better being better, then start thinking about optimizing and running local models. And you know, I was talking to Barry about like, hey, we should probably create almost like some appliances for open source models that you know, in our offices we can just kind of drop ship it, you know, go plug this into the into the network kind of thing. Because God knows you don't want people like local configing that stuff. So, no, I mean I think, but no, 30 grand a month. I mean, what's the other choice? Either go slow or hire one engineer. Yeah, yeah, I think that's one engineer, yeah.

SPEAKER_01 34:20

Yeah, I think it's like pretty simple labor arbitrage, is what it is.

SPEAKER_00 34:25

So and I think and I think a lot of folks in our industry, like you know, last week, you know, last Friday we saw clawed code, saw a lot of co-work, right? And I think the unlock, I mean, I just feel like last Friday was such an unlock for the team. Yeah, it's one it's one thing to like read the website, hear, hear me talk about our strategy and the future of AI. It's different to like do it.

SPEAKER_01 34:50

I was gonna say, one of the I think one of the things that's interesting day-to-day for technology companies right now is the impulse to go build at all at all times because the tools are so good is just like like it's deafening. That sound, the sound and the pull into into into doing stuff with with coding tools is like, yeah, just it's it's super loud. And yet you still have to do your your your you still have to complete your day to day, right? So like these are like when you're building products in most cases, like you're these are side hustles and you're experimenting on the side and after hours. And I think that but but you know, every day, like I like it's like a video game. You want to play, right? You want to wake up and you wanna go pick up where you left off, and especially if it's like accretive to your your workload and your efficiency, like of course, like you know, you can easily rationalize spending time on it and not getting your core your core stuff done. And what I think is unique about allowing your team members in this, you know, during during this this this phase of existence we're living through to actually you know come together and and and and do that thing and not have to make it like this after hour side hustle. That's a pretty it's just like a it's a it's a strong vote of confidence that one, hey, you're recognizing what they actually want to spend time on and do and make their make their work better. But two, like, yeah, it's a great unifying thing. I think like it's a video game for everyone, like you can build anything.

SPEAKER_00 36:19

Yeah, there's they're using this term called Claude Widows. Because everyone's everyone's dying. No, because everybody's not, you know, uh uh, you know, a lot of these guys are on Claude 24-7. Their wives are like, Great, I guess I'll go do whatever, you know.

SPEAKER_01 36:40

Claude Claude zombies.

SPEAKER_00 36:42

Yeah. No, I look, I I think also like you know, you it's it's very hard for companies to make the space, but I'll tell you, if they do it once, they're not gonna go back. But it'll it's only gonna work. Look, I'll keep saying if the CEO's not in the room, I don't care how big the company, if the CEO is not in the room on the keyboard, it doesn't work.

SPEAKER_01 37:07

It doesn't work. Yeah, like from our from a shadow standpoint, when I when I've had this conversation with some of our founders, you know, they're all really excited and jazzed about it. And usually the impulse for them to do something like this is hey, I've got team, I've got members on my team, even some of the engineering staff, that just they're just not they're they're not embracing, they're not embracing the tools. Like I need to figure out how to get them excited. I've got other people that are jazzed about it and and super efficient, but I've just got this lingering crowd that's resistant and they still like the you know the crap the craftsmanship. So they use it as a as a way to accelerate and basically force the you know the the resistant crowd to to get excited. But to your point, I don't think that habit builds over time unless you do it one, do that hackathon consistently, and the the leader is actively, you know, actively building with the product with the with the with the tools themselves, and is the one like enthusiastically pushing the ball forward every day. If it drops after every hackathon, you know, you're you're not gonna get that same sort of follow-through afterwards.

SPEAKER_00 38:20

Yeah. No, and I think you have to give people, you know, you just gotta keep people you gotta give them space to do this kind of work, focus, time, and then support, right? It's not you can't just say like, hey, go go write some code. Tell me how, right? I think it is definitely like a you know being in the room side by side with your team to do this stuff. I I still get shocked by founders I talked to. I'm like, hey, you're building this. Like, are you guys using you know claude code equivalent to like, yeah, my team is? I'm like, well, why aren't you? Like if you're a startup CEO and you're like pre-seed or seed, unless you are selling every deal, right? Unless you are the sales quotient to the business. And oh, by the way, you should probably be using Claude Code to improve that. I mean, these aren't big teams when they say, Yeah, my CTO is doing that. It's like, what? Like, you're I mean, you and your CTO are doing this, right? And I and I I get shocked by small teams where the CEO acts like, you know, like hey, you know, you're building a tech company, you know, you're building a tech company. So I mean, I don't know. You know, we we've hired a couple of these, it's it's kind of a new term, relatively new term, is these go to go to market engineering. And I don't think I really quite understood it till we hired someone that had that background. And it is night, and I mean it's just different. I just don't think these days you can be in, I don't think there is marketing per se, right? I think there's go to go-to-market engineering. Like, I don't think you can just be like, Well, I do web copy. Like, is that a job anymore? I do web copy.

SPEAKER_01 39:56

I haven't seen cop copywriters in terms of new hires in a long time. I think there is a function for I have seen people hiring for writers, but pure like web copy, that's I mean, that's automated. Yeah, yeah.

unknown 40:10

Yeah.

SPEAKER_00 40:11

So I think like, oh, web copy, that's part of your job as a go-to market engineer.

How To Transfer Scar Tissue Fast

SPEAKER_01 40:16

Yeah, it makes sense. To pull back the topic on when IP leaves the building. The reference I made to the LP, the LP who followed up with us after the last pod. So his the thing he's trying to solve, he says, I'm trying to impart years of my experience in decision-making processes, and ultimately a lot of the scars that I have on my back to to an associate that is junior to me. He worked, he works for me, but I want him to like deeply understand the things that you know I've learned throughout my career. And right now, the like the time element there, it's expensive because it just takes like I can't share every story and and you know, important every lesson that I've had in my entire career in a meeting or a series of meetings or a year, right? It just takes a long time for someone to to to clean that. So it's like he's so he's asking, how do I engineer that? And I think the answer is like a little bit of what we covered, right? It's like it's kind of a skill, it's it's a skills file. And I think like literally, uh the job would be to pull out. I mean, how I would I'm curious to like I'll ask you this in a second, but how I would think about doing it is like go skill by skill that you feel like you need to teach to you know, to your junior associates and create a create a separate file for every single skill. Well, maybe one's decision making, maybe maybe one is acquisitions learning. This guy's a an acquisition. So like specific deal making MA knowledge, that's another, that's another file. Another file is like the relationship side of the of that of acquisitions. Like, how do you think about the setup of meetings and where to, you know, do you do a first call? Do you not do a first, like all of all the basics across the full spectrum of work? Everything is a separate file and a separate skill, but it's your job to document that or use the AI tools to help pull pull it out of you, what what basically the knowledge that you have. And you can dictate, you know, you can dictate and and and speak on the all the lessons that you've learned. You can type it in, you can write it in in a notebook and take a photo of it and put it in and put it into an LM. Like how you actually get it in there is not difficult, and it can help you retrieve all of the lessons and skills that you have.

SPEAKER_00 42:44

No, like I'm I'm a big fan, I think of like I think of our core stack of how we think about things, right? Is quad, notion, Slack. That's kind of the that's kind of the stack, right? We have other stuff, but that's the main stuff. So nowadays when I do calls, I transcribe with Notion, it drops it in. If I think you know, I don't I don't keep everything, right? It drops it into Notion. I'm like, oh, this was a really interesting call. You know, because like right now I'm talking to a bunch of capital allocators for the next round, right? So I'm having a lot of these one-hour calls and 30-minute calls, and it's transcribing it. I go back and edit it, and then I move it from private into the public spaces. So now when someone goes into Claude and says, Hey, how does KP think about this strategy or whatever? Claude talks to Notion and sucks it all in, right? So there's there's a great way to capture data, put it in, put it in the notion, and have it accessible, you know, via Claude and for everyone to use. I think what I would what I would do if I was this person is you know, focus on a couple key areas, right? I think also, you know, I was talking to the team on Slack yesterday, I was like, you know, don't forget if you think about Claude and you say, Claude, go build me something, if you don't know what that indeliverable look is needs to look like and what great looks like, you should not do that.

SPEAKER_03 44:16

Yeah, yeah.

SPEAKER_00 44:17

What you should do is have Claude train you, coach you. You know, I gave the example on Slack. I said, look, you know, if I ask you to go do a go to market strategy and you've never done go to market in your life, do not go into Claude and say, I need to go to market strategy, right? Say, hey, KP asked me to do a go-to-market strategy on X, Y, and Z. I've never done one. What are some questions I should ask him before I move ahead? And then comes up with some great drop those as hey, KP, I'm happy to work on this go-to-market strategy. I need a few questions answered. Yeah, cool. And like so use it as your like training partner. But I think there's too many people just dropping stuff in, asking for an indeliverable that they don't know what great looks like. And if you don't know what great looks like, you probably, I mean, you probably shouldn't do that. But definitely use it as your training partner. And I think in this person's case, right, to have some content or whatever for this younger person, say, hey, like next time I ask you to do something, you know, be sure to ask me questions. Like, do ask, ask Claude, what questions should I ask you? You know, so I I think not everything, how you use the tools can be very different. And I think a lot of people have gotten to just use it for straight up deliverables, and they don't even know what a good deliverable looks like.

SPEAKER_01 45:37

Yeah, totally true. Good advice.

SPEAKER_00 45:42

All right, man.

When Standards Create Mediocrity

SPEAKER_01 45:44

I had there's one post that stood out to me this week. Okay. I think you're yeah, I don't know. Your your posts, I was gonna say, have been very CEO oriented. Your your your language your language, I think, has changed in the last six months, which makes sense because your job has changed. Okay, so this is this is one that stuck out to me. So standards and processes exist to reduce variables and deliver a consistent outcome. For example, quality means consistent consistency and not necessarily a better outcome. This is general, this is generally mediocrity. With AI, we can run a multiple, we can run a multitude of variables and push a better outcome that is in fact not consistent. Case in point, do we actually need master spec? Talk to me about the just that qual the quality trade-off and how things are different now.

SPEAKER_00 46:45

Yes, so I actually wrote about it in my book. I think that the title of the chapter is Deming is Rolling in His Grave. So if you know about Deming and Total Quality Management and the Toyota way, right? It was never about like building a better car, it was about building a consistent outcome, right? Quality was about consistency, not about better, because in the world of if you pull 10 people, nine of them would rather drive a Ferrari that breaks down than a Camry that doesn't break down. Yeah, right. So I think we've in our industry is very complicated, right? It's a very complicated industry. So we've created a lot of these standards because we had to reduce the number of variables. You know, an architectural sheet is an A.1, A dot, like we've we've done this, right? It's all standardized and codified. The reason I was picking on master spec and CSI a little bit, I don't know if you know, they're suing a lot of people, and it's become a big thing. Like, oh, you need to pay us every time you use our it's almost like every time you write a spec, they want a slice.

SPEAKER_03 47:52

Yeah.

SPEAKER_00 47:53

Yeah, I don't know the details exactly. I just know everybody's getting sued over this thing and everybody's mad about it because a lot of people kind of feel like the CSI master spec is almost like open sport open source at this point, like it's been around for so long, right? Yeah, and and I think that's what happens, right? Like if you think about the reason we had to create all these standards and processes and everything, because we didn't have the compute, you know, to look at everything on a custom basis. And now I think if you look at something, we can look at everything on a custom basis, right? You know, I think we were talking about pitch decks, right? Everybody decided that, like, oh, 10 slides. Here's Kane Inventure Standard Dex, you know, like what? So what did that do, right? That gave founders the opportunity to feed us information in the way they wanted us to see it, right? Much like maybe a dating profile, right? It's the best version of you. Maybe not the truest version of you, but it's the best version of you that you're gonna put out in the space, right? I would almost rather have a long farm old school business plan with all the data, not the summary of the background data, not the summary of the TAM. I want the hundred pages because then I'll suck it in, I'll interrogate those hundred pages based on what's important to me, not what's important to you, but what's important to me. And so I think that that's where the world has changed, is like we've tried to get everything down to step one, step two. And I feel like highly documented operational processes is is mediocrity, right? McDonald's consistently delivers a burger.

SPEAKER_01 49:37

Yeah, their goal is standardization, they're very consistent.

SPEAKER_00 49:41

Yeah, yeah. I mean, that doesn't make it a great burger, makes it a consistent burger. And so I think we've done a lot. I mean, we have standard bodies for everything in our industry, yeah. Drawings and specs and contracts and this and that. Maybe, maybe that's what's also holding us back. Is that if you're gonna stick to standard processes, it's it's it's the path to mediocrity.

SPEAKER_01 50:04

That can finally change. We now have a new new form of leverage.

SPEAKER_00 50:08

Yeah, I'm sure plenty of people have great ideas about like I mean, if I had to do specs, here's how I'd do them. Totally. Go do it, go do it. You can don't get stuck. And for a lot of our industry friends that are listening to this, is like, that's insane. Why would we do that? These specs have been around forever. That's great. Tell me how that's been going.

SPEAKER_01 50:30

Good Noton.

SPEAKER_00 50:32

All right, man.