Not everybody on the street has a chainsaw. You all have to cut down pieces of tree every once in a while after a storm or just because it's, you know, gotten too big. So you go and you borrow a chainsaw, you don't need it forever. You just need it for a little while. That's where I think people that are really serious about AI right now, and it in that small to mid market range that we kind of talked about, they need to go borrow the chainsaw if they're serious about it and start developing that muscle because they don't have the time and the bandwidth on staff.

What's going on here? Hey, hey, Jim. I am doing good. We are you know in the throes of ⁓ you know middle of summer, family and kids and doing all the things and kids have got

How are you today?

Camp ⁓ which ⁓ today's a Friday for ⁓ for us. So that means that ⁓ it is the end of camp week, which means that there's gonna be some type of program that ⁓ we have to go to.

Well, that's always fun. You know, wa watching somebody else's kids. If it was only my kid, you know, doing the program, I'd be all in. It's just I have to watch everybody else's kids too, and that takes a long time. That's right. Not that I don't like the kids. I like the kids.

Yeah. It is. But you know, like you know, like, ⁓ you wrote a poem. Good. It's gonna be just like Tennyson. It's gonna be fine. It is.

It's gonna be really good. I promise I won't fall asleep. So yeah, we're getting ready to head out and ⁓ fish with my dad for ⁓ a week or so. So my kids are looking forward to that. So we're definitely in the middle of summer and still trying to make a lot of time for AI conversations. I'm having those all over town. So

Well, and so, you know, the a lot of stuff happening this summer with AI with a lot of the companies they're ⁓ releasing new ⁓ versions. We've had ⁓ we've talked a little bit in the past few weeks about like some of the the you know kind of government interventions and things that have been going on with the release of of ⁓ new models. ⁓ and as these new models come out, you know, I think there's a little bit of an arms race that we start seeing with ⁓ These tools where they want people to use them. So, right, they kind of make it a little easier, maybe for a week or two. Anthropic gave some extended credits to people, right? To be able to run longer. ⁓ OpenAI similarly did something ⁓ last week that was different. And I've heard different things about it, whether it was in response to some kind of incident they had or Whatnot, but ⁓ the five hour limit that many of us are used to with running our agent and having a usage window ⁓ of of burn. No more five hour limit.

I know. That's what I heard. And so what do you what do you do at that point? What what did ⁓ you decide to do when they removed all the barriers? Well

Well, I ⁓ I had an idea and ⁓ and this is one of those things that we'll talk about today is the fact that like I I saw a product ⁓ that was kind of like an AI assistant ⁓ that could kind of watch your screen and and give you kind of like in the moment agentic nudges. ⁓ and so you know, that means that it's viewing the screen context, it's using some of the inbound accessibility hooks to kind of determine. ⁓ context and intent and and try to you know give you ⁓ you know what what is your next usable action or you know what does it see you doing and I was like, ⁓ that's kind of an interesting idea. I wonder if I could build that. I you and I talked a little bit about, you know, where where are we going to find a project that's gonna like make the AI grind away for, you know, where people are talking 12 hours, 18 hours, right? And and this was my project. And so I gave it to it, said, you know, come up with A plan, multiple sprints, these are the features I want. These are the constraints I want you to follow. These are the and ⁓ and then ⁓ and then yeah, ⁓ I kept it going. And ⁓ and it was a little bit like the first day or so, it was before the five hour limit was up. ⁓ and so that was interesting because I had a reset. And so, you know, they they gave you a do over button in codex as well. I don't know if ⁓ does Claude Cowork that every once in a while? Like give you like a like a magic reset button?

No, no, they don't have they haven't had that concept that I've that I've seen.

An interesting kind of thing that Codex has. So I'm like, ⁓ I stayed up late one night, you it's 2 a.m. I'm gonna hit the reset button and keep it going. And then I didn't have to hit the reset button anymore. ⁓ and it went for like three days. ⁓ really? Yeah. And ⁓ okay, now, like, you know, truth time. What did I get for three days of compute?

Man, I don't know. What did you get for three days of compute? Did I mean like did you get anything good?

I got

This is all about truth. We talk about practicality and what actually happens.

Right. Right. What what what really happened? Did an app get built? Yes, right. Is it is it is it, you know, running on my Mac? Is it Xcode? It it does it have like, you know, kind of, you know, like app controls and you know, a window. And is it running? And i I part of this whole deal was ⁓ if you're gonna put a agent or something to view your own system, the idea was to keep it all local, right? So ⁓ so I'm running ⁓ the Gemma twelve B. model locally on my computer. ⁓ and so yeah, it it it's wired to it. ⁓ is it really fast? Is it really understanding intent? No, it's trying. I mean, like it's it's it it it definitely like sees my email and then tries to give me a coaching thing of like, ⁓ you could write this response back to this person. Do you want me to do it? ⁓ It's not very smart. and it's not very fast. It's not, you know, I'm like, ⁓ you know, this would be better if I had something maybe running it faster. But you and I have talked about this. ⁓ fast and wrong is just as bad as slow and wrong.

Just as wrong as slow and wrong. I mean, a hundred percent. I I would say this is the thing that is really apparent to me now. I mean, I don't know that I've spent ten thousand hours. I don't know about you, but spent a lot of time. There's my dog. ⁓ But it is really clear that it still requires a lot of work and understanding the software because it, you know, when ChatGPT first came out, there were there was nothing. I mean, you could I don't even remember. You could make it a little bit Customized. I think you had some instructions you could put in there, but

Overall.

It was not very complicated. It was a little box and you, you know, tapped in words and something came out. Usually something ridiculous like a story for your kids or something like that. Or remember that was like the the biggest thing for like the first couple of months was, you know, write something like Shakespeare. You know, it was just kind of like hello world type of things before anybody knew what it was really, really capable of. But then it started then they started adding all these bells and whistles and the settings got deeper. And now when you open up Chad GPT or I should say Codex or Claude, you really have to dig into the settings. Small example, just so we give a little bit of color to what I'm talking about. If you want to do skills or use skills in Claude, you actually need to, especially on a personal account, if you're on a Teams account or enterprise account, your administrators probably already turned them on for you. But if you're in a private account, you actually have to go in. There's a setting that allows you to build, create, use skills. And a lot of people don't, you know, know about that and it gets shut off. And then people are like, ⁓ this stinks. So that's just one little example. But there's a ton of settings in there, especially when you get into more advanced work or if you want to do some things like what you're doing with apps, you really have to learn. There's a there's a learning curve. And then to your point, when you create something, it's not always that good at first. And you have to iterate, iterate, iterate. I talked to somebody yesterday, we were building a skill and the skill in like just doing the first part, like going through the process to show it what we were building towards. We're just trying to create a document that they churn out all the time. It screwed up like fifteen times technically all along the way. And it's gonna need better guidance in the future on how to accomplish that task quicker, smarter, faster. So the human had to be in the loop. The human had to do a lot of work. It wasn't ⁓ I type something in and I have this beautiful everything works app.

No. And I think the other part of this is like you you know, like ⁓ what I did the other day was a little bit of a thought experiment. ⁓ and and ⁓ you know, there there was no skin in the game really for it. It was just kind of like, you right, I'm I'm sitting on on a tank of free gas. So right, I'm like, I'm gonna I'm gonna I'm gonna go on a road trip here, it doesn't matter where. And and you know, to those that might listen and be like, ⁓ you know, like that's that's it. You're doing you're doing the thing that that that we're we're s telling everybody that you shouldn't do. There the part of the the exercise was the the journey aspect of it, right? I got to, you know, work on a app, ⁓ you know, specifications. I got to use this, you know, new model that Chat GPT had put out, ⁓ their 5.6 ⁓ with Seoul. They it has a little extra setting on it called Ultra on top of Max. And ⁓ Ultra spins up subagents on its own. And so that was kind of interesting to watch the harness do the work, start assigning things to agents. It it would name the agents over to the side and give them like little ⁓ names. And then you could click on the agent and see what they were working on and then see them. reporting back to the orchestrator. There was ⁓ for me, there was a learning aspect to this to kind of watch how this played out, see how ⁓ you know, what what could I have done differently from a coaching perspective to make this behave differently? ⁓ you know, the the fact that this wasn't a true product that, you know, I I had a lot of s you know skin in and making it work, you know, did it work the way that I wanted to? Yeah. Yeah. Was it a good learning experience? Absolutely. Cause Seeing the tool work, knowing where I I could have done better. ⁓ and you and I have kind of talked about you know if if you're not realizing that ⁓ you kind of suck at it until you do it a little bit, that you need ⁓ you need to get these reps in and these new tools. Every one of us needs to get extra reps in to really extract the value out of these ⁓ just a tool set, right? And then, you know. Then we can actually talk about delivering value, right? I mean and and like moving to a something w that that actually, you know where we do have skin in the game and it and it's something that we should we should be fixing.

Yeah, I I agree. I think the people that have put in reps are gonna be the ones that you wanna lean on, like just switching the conversation for a second to companies. People that have put in the time and have the reps and are the best at AI, you've got to move those folks into the right place. We've kind of touched on that before. And, you know, we've always talked about training, fluency, actual hands-on keyboard practice as being the basis for everything. If you really want to move towards, you know, using these tools in your business, right? That's you can't really skip that part, or you have to hire for it. It was interesting, Uber's ⁓ CEO, or I believe it was their CTO, pardon me, he mentioned that they have started to develop agentic pods. I thought it was

Tripods.

⁓ no. No. Not a warning label for children on these. This is just a this is a pod of people. and what they've done is they took their best engineers who were also their I should say they took the best engineers, according to them, at AI. So you got software developers, software engineers, and there's five percent of them that are

Are they taste?

extremely good at using AI to improve their work. Hey, you, however many people they had, you now all live in HR or marketing or finance. And so they move them out for two week sprints, pretty much, to go. These are just really smart, accomplished folks. They go, they learn the workflow of, let's say, you know, an HRIS process that takes somebody a long time to do, and they automate it. With AI, thereby seeding AI into all these other places in the company. And I I would imagine probably creating a few power users along the way, people that like their interest is piqued by the fact they've got this expert ish person coming in and and helping them make the company more efficient.

Interesting. Yeah. That I I had I hadn't seen the article. I mean, ⁓ so you're describing this thing and to me, ⁓ you know, I've worked in manufacturing with with companies and and you know, there's a ⁓ Toyota had this process ⁓ that they called it the Kaizen Blitz. ⁓ I think General Motors called it ⁓ or had a version of it called ⁓ Diagonal Slice, but ⁓ and we had some name for it with Six Sigma. But

I thought it was interesting.

⁓ you would cherry pick ⁓ people from different walks of life with with within like the plant, right? You'd get an engineer, you'd get a maintenance person, you'd get a tooling guy, you'd get like a like an assembly line worker, and you'd put them in a little pod. And basically you'd give them like 72 hours and a sheriff's badge. Right. Like, like, you know, and be like, you you can go, you do not have to go ask permission. Like, Go find something to fix that is that, you know, i is just a pain and bureaucracy is is holding us down. And they would they would go in and and you know, and it was interesting, right? Because they would get the people to like, you know, people with the idea, people with the know-how, people with the practical skills, and ⁓ and they do so and and and the idea is that you'd have these like incremental little lifts, right? ⁓ that would be maybe impossible because either it would never rise to the right level to for somebody to authorize it at at a high scale, or it ⁓ you know, it it would it would nobody would actually point it out because people have just gotten used to the fact that this process is kind of, you know, janky and you know, you you gotta pull the lever twice every time, whatever it is. And ⁓ but it feels like like like like it's an old management trick, like the the the Uber story to like Say, all right, well, we're just gonna embed these engineers, give them a sheriff's badge, let them go do something for a week.

Yeah, I I don't disagree. I think that's the best way to do it, but the companies we're talking about are huge. So yeah, like it I I don't believe that most small, you know, small businesses, you know, let's say a hundred to two hundred and fifty million and then mid size two hundred and fifty to two billion. I don't think they have a lot of in a lot of cases the bandwidth. on staff to go do these things, to really just say, ⁓ you don't have to do your job for two weeks. Go learn, teach, do in another department. I don't think that that is practical. One of the things that we talked about the other day was, you know, we had this ⁓ an well, I had this analogy, whether it's good or bad, about, you know, not everybody on the street has a chainsaw. Y'all have to cut down pieces of tree every once in a while after a storm or just because it's, you know, gotten too big. So you go and you borrow a chainsaw, you don't need it forever. You just need it for a little while. You go to Home Depot, you rent it. Whatever. That's what that's where I think people that are really serious about AI right now, and it in that small to mid-market range that we kind of talked about, they need to go borrow the chainsaw if they're serious about it and start developing that muscle ⁓ because they don't have the time and the bandwidth on staff but if they're serious if they need to ex speed up the close from you know let's say they have a 20 day close we just talked to somebody recently who's just which it's like tail as old as time. CFO walks in, there's a 20 day close the mandate is get us to five or seven. We got or 10 would be great. Cut it in half. that That takes you're gonna have to go out and borrow a chainsaw. You need to go find a person who's already good because you don't have I mean, if you've got three people on your team in marketing, you don't have the you just don't have the bandwidth to

Right. And I think one of the things though, right? I mean, like the pushback on that is, you know, people thinking, like if if we if we stick with your chainsaw example, right? Like thinking that they need to go hire a landscaping company to come in and ⁓ you know, do that work, right? And and who's got money for that, right? I mean, who's, you know, we got other bills to pay. ⁓ and you're talking about walking down the street and finding which one of your neighbors nearby has got a chainsaw and ⁓ you know, you'll you'll he'll let you borrow it and and ⁓ maybe you owe ⁓ a bottle of bourbon after you're done. Right.

Right. I mean, I think there's I think there's ways that you can find I you can find some people out there right now that would be willing to come in and teach you. I mean, if you're starting from zero or even twenty-five percent ⁓ you know, knowledge about AI is what I'm talking about. If you're starting from a really low point, it doesn't take a lot. You're gonna get a ton of incremental value even by you know, without hiring an a an entire Cap Gemini. Landscaping crew. You don't need all that. You you could find you could find some smaller ⁓ consultants out there or people that are just interested in it. junior, I mean you could hire somebody junior on staff. You don't even need to go, you know, spend a ton of money. You could just find somebody who's extremely curious about it and bring them on and start seeding it into these different departments.

Well that th and and like you have I mean, like you know, that's not just like advice, right? I mean, like you have that as a as a as a real example, right? And there the you you've you've done that. You know, you've had intentionality was part of that, you know, to some degree.

⁓ a hundred percent. You had to we we created a position. We Had that position not only teach, but also embed themselves in these different departments. So yeah, I've we've done that in real life. It works. I've gotten a lot of feedback from folks that said, ⁓ man, you know, because I was intentional. They said, I really am glad that you're giving me these skills. Now, just so you know, I did tell everyone over and over again, hey, not a lot of companies are going to give you these skills. We're giving you these skills. And teaching you these tools, this is really important for your career. So there was a little bit of advertising involved as well. You know, I stayed very much on message. Here's why we're using this. We're gonna get more efficient, but this is also helping you. This is also helping you. So I think explaining what's in it for the for the people are is is really important. That's a little bit of a tangent, but

Well, I mean, ⁓ y you have to make it important to the people. You have to make it important within the organization. But the other thing I and I think we've and I maybe something that we've tried to to, you know, shout, preach, whatever you wanna say, is that like you actually have to go find a problem, right? I mean, like you have to know your business well enough to know, you know, what are you measuring what it where is the problem and then try to, you know, ⁓ it's not that, you know, AI or anything is just gonna magically fix it, but we we're talking here about having some tools that are pretty game changing. If you can articulate your problem, you can uncover ways to solve your problem that you probably haven't thought of.

Yeah. You need people to do the the same thing that they've done forever, which is create a business case and lay out why it's better than all the other business cases. And then and then you have to have a, you know, as unbiased of a group as you possibly can sit down, look at all of them and say, all right, those are the ones we're going after. That's the that's the key, ⁓ especially if you're doing big things. I think with AI fluency and just that. base, hey, how do we get people using it and comfortable with it? That's kind of table stakes at this point. It really is. If you're if you're doing big projects, especially customer facing projects, you're exactly right. What is the problem we're trying to solve? Test it. The whole thing. I mean, there's nothing different about applying an A doing an AI project, especially if it's a big one.

I think what I've been trying to like tell some folks is like the beginning of a project, if you have the right mindset, is a lot faster and you can get to a much like more like cohesive shape of what the of a what you know option A and option B potential solutions look like. You can get there a lot faster. Right. Now the question is, is do you do something like I did as as just a you know throwaway experiment of like, ⁓ but then I just tossed it at AI and said, you know, see what you can do, right? Here's a and ⁓ yeah, you know, I I think I think the the the proof there was that it didn't do great. It did something. ⁓ you know, where where I asked it to just solve the pro solve the problem, go build it, no oversight. But instead, you know. If we had taken more time because we had skin in the game to, you know, well, what is ex what is the what does this exactly look like? What is the exact one problem, two problems I'm trying to solve? ⁓ I have a client that is in the home services industry and and w I I sent something to their supply chain guy the other day, which was kind of a you've explained to me that you've got these three ⁓ Areas that you you have identified. These are metrics that I want to influence, and these are where the problems are. ⁓ And one of the things that ⁓ I've been coaching folks, but also building out for folks, is ⁓ kind of a robust discovery process that can be driven by a Claude or a codex. ⁓ you know, here's a big prompt that gives it enough context about your business and enough context about. its task, which is to interview you, identify the boundaries of what you're trying to solve, ask you some hard questions about, you know, where, you know, where the ⁓ costs and you know benefits really are, really kind of pushing you to have a little ROI conversation. ⁓ but to to again to kind of sharpen the I have this problem with you know you know going from ⁓ I have this problem with three way matching on on on on you know from from from ⁓ all the way back to the invoice and delivery to no, I really have a problem specifically in this leg of the pro ⁓ of the process. And there is a way to, you know, that's the longest leg and we have a way to maybe reduce it ⁓ from four days to two days. But that that idea of like you you could have me or another consultant come in and and sit with you for multiple days, or you could invest, you know, Forty five minutes, right? Into a a deep conversational thing to get to a much tighter you know, problem set and solution options and then go justify, let's go build that.

Yeah, a hundred percent. I I do think I do think you can do it without a consultant. I mean, you and I have, to the point where we're now consulting people. I think sometimes it's like a personal trainer, you know, you're not going to do the reps unless somebody's watching you and you've carved out time and you have accountability. ⁓ so it just depends on I think it depends on the person and Again, the bandwidth of the company. So you have to try to figure that out and how fast you want to go. You could certainly accelerate it if you bring, you know, a resource in and help identify that one, not the whole thing, like you said, not the whole 48 problems, but the one problem. ⁓ because you know compartmentalizing all of this stuff is very helpful.

I think so. And just kind of getting to a structured answer, right? You know, where we take the, you know, I feel and I think out of the equation, ⁓ I guess, you know, and and turn it into, okay. Yeah. ⁓ and these tools excel at that. And I think, like you said, like, you know, you and I find ourselves in some of our engagements acting like that personal trainer, right? We're we're ⁓ We're we're coaching them through the process. We're helping them build the muscle memory so that they can do it without potentially being watched the fourth or the fifth time. But the first four, right? We have to kinda coach and correct the form and, you know, make sure they don't hurt themselves.

Absolutely. A hundred percent. ⁓ yeah. So I think homework for everyone would be, you know, decide decided, you know, where you wanna I mean, I think it's a strategic question. Like how how fast do you want to move? Are you serious about, you know, putting this into your workflows and ⁓ the way that you do business really? And if you are, pick something that's broken. Find one curious person or bring somebody in and ⁓ write a business case around it, get some controls. If I'm a finance guy, you know, I mean like put some controls in place, get some governance going. Light, light, light, light. But that governance for now can just be, okay, did you do a simple business case on a spreadsheet so that we understand what's gonna what what should happen and why we should invest this money. That's gonna go over a lot better with the CFO and the CEO than than just, hey, I need you know, what do you always say? Ten pounds of A I need ten pounds of AI someplace. It's like, okay, yeah, let's just focus. I think that would be the homework. That's what I would say. If you're if you're curious about AI or want to be the champion for that in your company, I would sit down, take the forty minutes and get the business case done and show ⁓ what is possible.

Yeah. And I mean, this is where we, you know, we plug a little bit of the work that we do, but like, you know, like that's where we help. I mean, we're helping folks do this and kind of get not just fluency and vocabulary, but get a couple of reps in. I think if you're curious and you see it in action, ⁓ you're you're not indebted into a into a consultant ⁓ driven world for forever. ⁓ and I think we're both talking to our ⁓ clients these days, ⁓ Really with this attitude of like how do I, you know, how do I get out of here as fast as possible? I I really just want to help you be a better version of you and then, ⁓ you know, and and and take take take this message somewhere else.

Exactly. Speaking of getting out of here as fast as possible, yeah. I am going to go talk. Well, I'm going to go talk to someone who wants to use ⁓ AI to accelerate some of their workflows at a small business in the finance and accounting department. so this isn't ⁓ I just continue to get lots and lots of calls and inquiries about this.

You've gotta go.

very topic that we touched on. So I think it was good to to cover it off.

Yeah, I think we're both seeing traction with this and I think, you know, folks can reach out to us and you know, find out if ⁓ there's some ways that we can help you or point you in the right direction. 'Cause, you know, you know, spent another episode with Veer and Jim.

That's right. We are the functional executives. Thank you for taking the time to listen. ⁓ please ⁓ follow us, like, subscribe, I don't know, all the things that wherever you're listening to this, do all of that. And ⁓ we're gonna have ⁓ a guest here in a few weeks. So looking forward to having our first guest on the podcast because what we'd like to do is is start talking to some business leaders that are are using this at varying levels. ⁓ some to really really great effect and others. Hopefully we get them to tell us some stories about, you know, where they stubbed their toe and what didn't work because I think that's really important with all the hype around this. All right. I'll talk to you soon.

Yeah, absolutely. Thank you. All right. You take care.