AI Just Took Over 80% Of Our Tasks (Here's How)

I think most businesses, the vast majority of small businesses, could double their profit, not their revenue, but double their profit just by using a whole bunch of automation in the business.

Mike, thank you for making some time heading down to the Peninsula.

Absolute pleasure.

You are— there's only a handful of people that are considered Systemology royalty. You, without doubt, are one of those. Had such a tremendous impact on the way that I work, the way that I think about things. I know we got to share an office together. You graciously let me share the big wall of systems story that I, I talk a lot about in the book. But there's just so many lessons that I've learned from you, and you're kind of an original E-Myth systems guy, so from day one. So we used to always geek out about that. So when we started talking about what we're going to talk about on this episode, I thought you were going to be so good at this because you, you see into the future. I feel like I can see into the future more than most, but you can see beyond that horizon. You're always thinking 6, 12 months ahead of me when I feel like I'm looking down the line. Thanks. And It was a pleasure to have you at the Business Systems Summit late last year. Good day. And you were talking about Claude Code then, and you were telling me probably 6 months before that, Dave, you've got to get on Claude Code. This is amazing.

I've never seen anything like it.

I know. And it wasn't until the start of this year when I have hopped on it and oh my God, without a doubt, it is the single greatest tool I've ever used in my life. It feels like—

That's why you send me a text every Friday going, oh my God, Claude Code is so good.

I do. I just don't know who else to send it to. And sometimes they don't even need a reply. Like it's just me just going, oh my God, this is amazing.

Which you know yourself.

And I thought, let's geek out for an hour on Claude Code. So maybe even just to start, I'd love to get your thoughts on What do you think makes this technology different from other things that you've used and seen and that have come before it?

I think there's a really important distinction between the model and the stuff behind the model. Technical word is harness. You can think of it as the scaffolding that tells the model what to do in certain situations. So what makes it so good? Well, there's that, and it kind of has emotional intelligence. The way it replies to you, the way it talks, it kind of somehow matches your level. Like, it doesn't talk to me in super uber geeky software development terms because it knows I don't understand that. I'm not a software developer. It's more business terms and just plain English, but it's all that harness behind the model that enables it to work so well on my computer. So particularly for the stuff that I'm doing, that members in my community are doing, it's basically knowledge work stuff.

Yeah.

Anything that you do in front of a computer, you can have something like Claude Code, and we'll talk about alternatives as well. You can have it help you with that because it's un— it knows how to use your computer. Weird, weird case in point, right? I am— we just moved house, and I'm trying to get my printer to work. And it's coming up in green. And it says, yep, printer connected. Every time you get a print, it just sits there and can't find a printer. It's right there. And I just gave up and said, Claude, can you sort this out? And it goes, oh yeah, but the IP address was the thing and the data there. And I sort of changed this and like, there's a PDF sitting on my desktop that I really need to print in black and white. Can you do that? It goes, yeah, sure. And it outcomes. Okay, I'm never going into settings again. I'm never going to touch all of that stuff. The computer quit. I was on a podcast last week, Riverside, and it just quit after an email. Shit. Find the link, go back in, apologize to the host. Don't know what happened there. He said, it's okay, we've still got the recording. Good. After the podcast, I said to Claude, why did that happen? Can you go back through? I know you've got access to all the logs and stuff. Can you figure out why that happened. It's, oh yeah, this, this and this. Can we make sure that never happens again? I have a dashboard that I have on one of my screens open all day. Calendar, Gmail inbox, bunch of stuff that's happening, some research, YouTube channels. We now have a card at the top of that that shows in green. And if something needs to be changed, it pops open and sits there in orange asking me to do it. It also knows if I have a podcast recording on my calendar for the day, Now it's red and saying, you need to restart this or do this or whatever it is that it tells me to do. I just do it, shut this down, restart the computer so that that problem doesn't happen again. Because it just, it can do all of that stuff on a computer. If you've ever sat behind a geek and they type stuff in on that big black screen, a terminal, and they type these big long strings of characters. That was what always stopped me learning to code. Way back in the day, all this MKDIR CD into this. And it's just, I don't know, my dyslexia, my OCD. I couldn't cope.

To be fair, you're also, you've been legendary with spreadsheets. I remember your—

I do love a good spreadsheet.

Macros and formulas. So I feel like it's in there.

Yeah. I think part of that is a way of thinking that is just really, really once again, I'm incredibly lucky. I've had that sort of training of Google Apps Scripts and systems thinking and all the Gerber stuff. And it's sort of a way of thinking in terms of input, process, output. And you just put those Lego blocks together. I think a big part of— there's a few parts to this. Gosh, I'm wandering off on multiple tangents at once here. But yeah, the like, having done a little bit of computing stuff in the past, thinking that trains you to think logically and break big problems down into small problems. And if that small problem is still too big and something isn't working, you break it down into smaller problems and you just work your way through it.

Which I just use a big term to kind of explain lots of different things. I think that's systems thinking, right? Like you're thinking of the system and then breaking it into pieces and then where's the problem isolated and what are those inputs and what are the outputs and why is it not correct and what's going on in the middle? Like, this just feels like a systems thinking brain is a competitive advantage with this new tech.

Competitive advantage. And I think there's two others. There's curiosity and there's your willingness to endure pain. The people that are getting the most out of AI right now are the ones that are just willing to get slapped around a little bit and like, but it worked yesterday.

Yeah.

And not just give up at that point, but just, I know, I know I can get it to work. It's what I've always loved about maths and coding is that I know that if I just keep going, there is an answer.

Yeah.

That's why I was crap at English at school because there are multiple answers and I always picked the wrong one. Like, yeah, I just couldn't write for garbage. But maths, if I keep trying, I know there is an answer. I know there's something I can get to. And it's kind of the same with this. I know if I just keep pushing it, I can get it to do the thing that I want it to do. I've just got to find the right way to ask it.

It does feel like the Iron Man, you know, and Jarvis.

Yes.

We are at that situation where we've got this computer brain that has visibility into everything that you're doing. On the computer and it can do it. Like you, and you just have to train yourself to ask the question, like the printer question. Oh, can you fix this? You wouldn't think of that normally on, oh yeah, that's what I'm going to ask ChatGPT to do or something like that if you're in the browser. Like, but you're now going, every question that I've got, if I'm touching a computer, I can now run it through Do you have a name for it? Is it chaff?

I call it 80/20 brain.

Yes.

So that's what I give to members of my community. I give them their own brain.

Yeah.

Let's call it brain. So 80/20 brain doesn't really roll off the tongue, but I spend all of my day in the brain, which is actually, yes, a big scary black terminal window with multiple conversations going on at once. Whenever I leave that to go into another app, I'm pointing here because they're usually on different screens. Why I'm now asking myself, why am I going into this screen? What am I trying to do here? What information do I need? And yeah, I've trained myself to ask the brain, hey, can you get that? Can you go grab the new emails from Gmail? Can you see what's on my calendar this afternoon? Because once you play with those Lego blocks, then I call it the adjacent possible. Let's wander down that path because I think it's a good frame, a good mental model. You're standing in front— AI, scary. You're standing in front of this big scary haunted house. And the front door is open and the first room is lit and the rest is all dark. And you walk into that first room. And there's a few little Lego pieces lying on the floor. And you're like, weird. Nothing else in this room. No doors or windows. Few Lego pieces. You play with those.

Oh, I get it.

That goes there and that goes— you look up and there's 3 doors have opened up out of the middle of nowhere. Yeah, there are 3 adjacent rooms. And you pick a direction, you wander in there, same thing, a few Lego pieces on the floor, no doors or windows. You pick them up. Yeah. What the hell is this? Oh, this is a— and when you look up, there are 3 more doors and you just keep on working your way through the house from room to room. You don't get to jump from room 1 to room 50. You have to earn your way to room 50 by playing with some Lego blocks along the way, putting them together and going, ah, once you get, ah, then you can move on to the next room. And that's the process. All these little things start to fit together. You try it, you see it working, you're like, oh, well, if we can do that, then, oh, well, if I can do that, if I can connect it to my calendar, then it can tell me what's coming up on my calendar.

Yeah.

And if I can do that, and I know from before it can connect to all of my meeting notes. Well, I've got a meeting with Dave this afternoon, so it should be able to give me some meeting prep. And summarize what happened in the last 3 meetings. Tell me what my action items were. Tell me what your action items were. It can look at my Asana and tell me if those action items got done, what's still outstanding. If it's a podcast, if you're hosting a podcast, oh, you're meeting with so-and-so. These are some questions you might want to ask. If you're the person on the podcast, these are some questions he's probably going to ask. And maybe you want to, on the drive down here, prepare some of this and think about these. And And it's just, you start adding more and more Lego blocks to that. And once you get that, then you go, well, that was easy.

Can you do that?

And you're just asking more and more and more and pushing the limit. I'm struggling to find the limit. I'm struggling to find the boundaries. That said, it'll be really good at one thing. And then the next thing that you do seemingly really similar, it'll be terrible.

Yeah.

And then the next thing, oh, it must be terrible at that too. It's really good at that. And it's, it's jagged is what Ethan Mollick calls it. Jagged intelligence. It's spiky. It's, it doesn't make much sense. It should, if it's good at this, it should be good at that other thing that's really similar, but it's not. But then this other completely different thing, it can be good at some of that and terrible at some of that. So it's just, you have to play with it, which is back to the curiosity thing.

Yeah.

You just have to experiment and play knowing that it is going to be a bit painful sometimes. That it's going to go wrong, that it's going to break. But if you keep on pushing, you get out the other side and it's marvelous.

When we walk into that first room in this haunted house and we look at the first 3 Lego pieces, like, what are the first couple of Lego pieces that someone needs to get their head around that kind of really set the stage? I know you've got 80/20 brain or the brain.

Yeah.

And getting full context, knowledge of your situation. As you know, we can touch on skills or connectors or— and I know some of these things are further down the pipe. One thing we always start with is context and getting your head around the fact that AI will give you significantly greater output if it understands more about your situation, position, what you're trying to accomplish. Yeah. So, I mean, for the longest time we've been saying, well, what gives AI great context? Well, it's SOPs, it's your policies, it's your position descriptions, it's your client avatar, it's your company goals. Like, it's how do you collect all of that? Is that what you think of as the brain or is that part of the brain?

That's definitely part of the brain. I don't think that's room one for most people because unless you're using something like System Hub and you've spent the time Yeah, to document that stuff. Most businesses don't have that documented. And you can't say, trust me, if you just go off and document all of this stuff, it'll take you somewhere between 10 hours and 6 months. It'll be really, really good. Like, they just have to see that there's potential, and then probably add that context bit by bit. And then, because once you see it, you can't unsee it. Yeah, you just get this visceral moment of Oh my God, AI can do that. It's not just telling dad jokes and writing Dr. Seuss poems.

Yeah.

And replying to an email. It can do that. Wow. Okay. I'm in. And then everything changes. So what is that moment? I think there's a couple of ways I can answer that. One is pick something that's really personal to you. I think for a lot of members in my community, it's been something outside of work that they've just said, surely it can't do that. Like, yeah, physicians will hook it up to some of their equipment and surely it can't do— oh my God, I can do it. Yeah. Because you've got more curiosity in that part of your life to go through the pain. If it's, if you're an employee somewhere and maybe not particularly engaged with your work, you're sort of not going to push through that barrier of, oh, does that work? So there's that. For a business, places where you're copying and pasting information or data between different apps, you know, a lot of us will be taking stuff out of Gmail and sticking it in Xero or taking something out of Xero and putting it in a Google Sheet. That's a really good signpost. Maybe not a day one activity.

It sounds like, because I always default into step one, what is the system? Which, which I'm imagining it's almost like identifying a problem that you want to solve.

Start there. Yeah. Yeah. Fall in love with the problem.

Yeah.

Pick a problem. The tricky thing with that, people say, but I don't know what scale of problem. Doesn't matter. Start really, really small. Pick something little, something that maybe you spend 5 hours a week, 5 minutes a week on. Just something little just to go, oh, oh, that was easy.

It did that.

Yeah. Okay. I feel a bit smarter now. I can see potential. Oh, if you can do that, then you— and then you iterate. Yeah. So you don't— there's two ways to build software for the last 50 years, right? What the old way was, go to the whiteboard on the wall, and everybody sit around for a few months, work out the perfect plan. And we've got this big waterfall chart, and everybody knows what they're doing. And that took 12 months to figure out. Now everybody, off you go. Yeah. And as soon as you start, everything starts to break and fall apart because you can't anticipate all of those things. Or there's the, you know, roughly where we're heading. Yeah, let's just start and we'll figure it out along the way. And I much prefer that version, which is the, uh, beg for forgiveness, don't ask for permission sort of approach to work. That works really well with this because you kind of need to see it. You don't know what's going to happen along the way. Our knowledge is the lowest at the beginning. We just dive in and start and and create version 1, version 0.1, and then, oh, version 2, version 3, oh, I've got this great idea, version 4.

And it's that speed of going from idea to impact now is so short, sometimes a minute or two, which I think that's the big unlock and why this is the greatest tool for the business owner or the visionary, because the visionary, someone with ideas, sees things and then has trouble usually with implementation, but it's the shrinking from idea to execution.

Instead of going on Upwork and spending 3 months and $20,000 on a team or external providers, you can just— and it's, it's, there's no handoff.

Yeah.

There's no, nothing being lost in communication because you are just talking to your computer, literally talking. I barely type anything anymore, talking to the computer and just like you would talking to a human being. It's not human. It can't feel. It doesn't have a conscience. I think AI should be alien intelligence. It's human-like, but it's also very weird. So pretend it's a little alien, pretend it's an alien junior employee, and you're asking it to help you with stuff in your work.

I reckon that's a really good way to think of it as a junior employee, because some people come with an expectation of it's going to nail it first go. They miss all of that iterative thing. But if you approach it as this is a junior apprentice that needs to learn the context, it's their first day of work.

They teach it stuff, right? You don't— somebody starts working with you and you don't say, oh, reply, build the proposal for Bob.

Yeah.

Who's Bob? What's he buying? What do we sell? Who are we? What are our values?

Yeah.

What products do we sell and what price point are they? They have to teach someone all of that before they can build a proposal. Exactly the same with AI. It's only as good as you said before, the context. You give it. It's just that giving it all of that context before you see what's possible feels a little bit, oh yeah, but I've got to build the trust first. I've got to know that this is going to give me something before I tip in all of that work.

I think, uh, and it's interesting to hear you say that and see your members who might not have the SOP or the System Hub account full of this stuff because I didn't realize that's one of my biases. I look at our best clients that are absolutely crushing it with AI, and there's a lady, she's literally up the road, Shannon Schmidt. Yeah, she's on Main Street here, and she's my gold standard for basically taking and implementing AI. And it all started with a System Hub account full of processes and how they do what they do. And it was very easy for them to just say, Hey, this is everything that we're doing. AI, how do we do this faster, cheaper, and better? Because it takes an incredible amount of time, but that's not everybody.

No, it's not. Most people don't have SOPs documented. I come from a digital marketing background the last 20 years, and while we built lots of SOPs, they would go out of date very quickly because that industry changes so fast. My GM used to say, if you wanted a nice slow industry, you should have gone into insurance because nothing changes there every 10 years. System Hub members absolutely have this huge advantage because AI is phenomenal. It's spiky, but it's phenomenal. All these things, but it's not a mind reader. Yeah, you've got to get it out of your head. And we had this conversation 2 years ago about like helping people get that stuff down on paper. One of the easiest ways to do that, if you don't yet have a System Hub account, you should, is Loom. I love Loom. Yeah, pick a task. Pick something small. It's not some huge project like multi-day thing. It's a task that you do in one sitting at the computer. It might be 5 minutes, it might be 30 minutes, probably in that range. And just sit down and just record yourself doing it and talk out loud while you're doing it. Yeah, I do this, I'm clicking over here because da da da da da. Once you've finished, 30 seconds later, the transcript's ready. Drop the transcript into an AI and say, how can you help me with this task?

Yeah.

Are there bits of this you can help me automate? Are there bits of this that I'm not seeing? How could this be better? Did you say cheaper, faster, better?

Yeah. Yeah. I feel like, uh, if step 1 is to identify the problem, step 2 is to understand how you're currently doing things, which is like you said, then the recording piece, whether that's a Loom, another one that we teach is just the voice memo on your phone can be enough. You're in the car driving. And you talk through that particular process. If you've got an iPhone, it'll automatically give you the transcript, copy and paste straight into System Hub, uh, that will then convert that into an SOP, um, which then also, like, tremendous wins can be gained for a business anyway, just by making something repeatable and following the existing process and transferable. And then it's very easy to then go to the next step.

We used to have this rule, you've just reminded me, back in the agency of human automation before automation.

Love it.

So if one of my Google Ads account managers wanted to automate something, they can't go straight to my developer and ask for a script that's going to do this, this, this, and this because they found it very hard to articulate the, the end state and the process along the way and very hard then for my developer to understand all of this kind of human stuff and codify that. The step in the middle that solved everything was, okay, first you need to teach a junior, could be a VA, could be someone in a different country, could be the person sitting next to you, but teach them and give them the checklist. You can't put everything into a checklist. It might be sort of step 4 is stop and talk to someone more senior and ask for help. 5, 6, 7, 8, stop and talk to someone. But get the checklist done because once you have that, and you can sit with someone and go through it, me, we, you, yep, then you find the gaps in the checklist, checklist gets a bit better, how you can take that to a developer, or now an AI and say, let's automate some of this. And I think we've been, yeah, banging on the drum for 10 years of, I think most businesses, the vast majority of small businesses could be I'll say double, double their profit, not their revenue, but double their profit just by using a whole bunch of automation in the business. There's so much wasted time, miscommunication, handoffs, things getting dropped. The problem with that in the past was that required a particular type of person who sat in the corner with noise-canceling headphones on and didn't talk to anybody. And they were expensive. And they were always had a big list of things to do. And it was always, yeah, I'll get to that in 2 or 3 months' time.

They're big, oftentimes black boxes. They'll create something, but you don't really know what they understand.

But you have no idea.

And then if it breaks, it's like, okay, who can fix it?

And I had exactly the same thing. I had a wonderful dev who worked for me for 8 years. Charlene, shout out, wandered off to her dream job at Bupa. Awesome, awesome developer sitting in the corner with noise-canceling headphones on. And of course, a week after she left, stuff started breaking.

Yeah.

And I had to go in and fix it because everybody else was crazy busy. I had the curiosity. I'm like, I want to learn that. So I teach myself to code again and I'll go learn stuff. This is a couple of years before ChatGPT. Yeah. But yeah, inevitably stuff breaks. It's not well documented because that person that's really good at building that stuff is not the person that documents all the stuff really, really well. But now, yeah, we have this incredibly flexible, versatile tool that can write the writing bits of that, that can write the codey bits of that, and get on to skills in a bit. But it's that magical combination of AI and code. It's just, yeah, how computers have been built for 80 years, the deterministic bit, if this, then that. We can take that and mix it with the AI bit. We don't want a task to be fully AI. We don't want the AI on a Monday morning going, How would Dave like me to manage his inbox this week? Ah, let me think. There's certain rules we want in place. There's certain tools and scripts and codey stuff that it can have already written. So it does that thing the same way every time. It creates the report in exactly the same format every single time. But the writing that goes into that report, that's what you need AI for. So it's this wonderful combination of AI and automation. Mush that together and a business can do fun things.

Ah, hey, I know you're tuned in right now, so I'll keep this super quick, but you like awesome YouTube videos, don't you? Well, there's not really that many YouTube channels that create great content about how to systemize your small business, so I want to make sure that you catch every video that I publish. So if you haven't already, firstly, give this video the thumbs up and then go ahead and subscribe to my channel. You can even hit the bell notification to make sure that you're the first to know when I publish anything. So go ahead, do it right now. I'll wait a sec and then I'll transfer you back to the video. I remember having a discussion with you where I was like, ah, I'm gonna learn N8n to do some of this automation stuff. And you're like, Why? Like, just skip that whole tool, go straight to Claude Code, which I ended up doing, and it's an absolute dream to sit there and, like you said, use something like Whisper Flow to be able to talk instead of type it out, have a conversation with the AI and say, this is what we're currently doing. I'd love to set up an automation that does this or that. You know, ask me some questions if you're not clear on something. Here's the link to the Asana task. I want you to jump in and review everything, and then I want you to come up with a plan on what you think we should do. Like, um, it's a particular process, uh, that we're going through now. Um, but that, that was a big unlock for me, just going, why just go straight to Claude?

Let's unpack that because there's so many good tips in what you've just said there. Um, voice, grill me, outcome. Let's go there. Voice is such a big unlock because it's 4 times faster than typing. But it's, it's this, as I'm very good at, as I'm demonstrating, rambling, going off on tangents, and just talking. That is context. That's such rich context. When we type, because of all of school and being a good student and all that stuff, we self-edit a lot when we start to type. And we, we think much faster than we type. So then we're, oh, do I put— oh, don't worry about formatting and grammar. When when you talk, you just— it just all comes out at the speed of thought. Bill Gates wrote a book years ago called Business at the Speed of Thought. Crappy book, but it was a great title. And I've always loved that idea of just business at the speed of thought. That's what it feels like now.

Yeah.

The hell else did I say? Outcomes. The biggest unlock is, I guess, pretty much how we were taught to manage those junior employees we talked about before. You read all the management books, You don't tell someone how. You don't stand over their shoulder and you don't micromanage. You describe the outcome.

Yep.

The why, what does done and done well look like? And you get out the way. Same with AI, especially now. There was a step change back in December where all of the AIs just got better. And they're better able at understanding that outcome that we want. So if you can describe the outcome and say, I don't know the best way to get there, suggest me 5 options. Yeah. Recommend the best one. Sounds good. No idea what you're talking about. Carry on. Yeah, let's do that. And then you also said planning. So having it build a plan of what it's going to do and forcing it to think— it's not really thinking, but let's just call it thinking— forcing the model to think before it does all of the stuff. And there were lots of little tricks to do that over the past couple of years, but now it's sort of baked into— we talked before about the harness— it's baked into the model and all the stuff around the model forces it to sort of think in the background before it shows you an answer, that makes it so much better as well. So you combine all of that stuff, it's just—

I can imagine. So, uh, I see different people using AI for different things and they'll go, oh, I've got this AI agent that sends me a daily briefing that summarizes the search results and tells me the top 3 news stories. And I'm like, oh, how is that useful? It's just adding more noise. Like going straight back to step 1, which is what is the problem we're trying to solve?

Yep.

Uh, one thing that's working really well for us, it kind of links on to what you're saying with this automation, and one of the reasons I was looking at N8N, but now I'm just going straight to Claude Code, is inside our Asana, we use that for our project management. We've got these different recurring tasks that have happened on a set cadence and rhythm. They're usually assigned to a team member, and it might be, sometimes it is just copying and pasting or logging into HubSpot to do a series of maintenance tasks. So once you get your Claude code set up, you give it great context, and we'll talk a little bit about connectors in a moment. I then say, I literally give it the link to the Asana task and say, hey, have a look at this, read through everything, follow any links that you need to get a full understanding of what this team member is doing. Then let me know which parts you think can be automated. It'll go do its research, come back and say, okay, well, steps 2, 3, and 4 are very low risk. I can do that. And it's going to make their job easier because I'll get 80% of the work done and just leave the 20% for the human. And then I say, great, can you build that, run that, do it once manually and post the result into that chat thread and what the exceptions were and where things broke so that the team member can then comment and say, ah, that was good. It missed the mark here. I think it missed it because of this. And then I'll go back a little bit later. I'll have the chat window and I'll say, great. I've got an update from Jill and Jill told me, you know, go read the comments on what she said. Can you see any patterns there? Oh yes, I can see the pattern. Let me adjust the filters and I'll run it again. Yeah. And then it runs it again and posts it in there. We do that once or twice until we got it just right. And then I say, great. Can we now turn this into into a job that you then run, you know, every week or whatever the case may be at a certain time. And what I want you to do is you'll run that on a set cadence. I want you to, in the description for the task, put a bit right down the bottom that says, you know, this is the GitHub Action that I'm running and, you know, a little bit of context. If ever you need to come back and want, you know, instant, I understand what this task is about. And then I want you to Uh, every time you run the action, you post the comment inside the Asana. Uh, so here's what I've done, here's what I've completed, so that then the human who's overseeing it, they're still assigned to the task and they're the owner of that particular task, but now 80% of that task gets done by the robot. Human then looks over the exceptions and double checks, and then it's just handled in automation. I've systematically gone through now our Asana and we're just picking tasks off like this. And that for me has just been a game changer.

Magical, isn't it? It's magical. There's a couple of things in there to pull apart as well, I think. The trust piece. So lots of people will say, somebody listening to this will say, yeah, but it's crap at this. It hallucinates. It does this. You can work your way through that and get it to the point where it doesn't. By providing it with better information, like you said, by running it, be we, you, by doing that with it, looking, checking, tweaking, iterating. But what you're doing with all of that is building trust, just like you do with the junior employee. Yeah, that when you do this task, you do it, you follow the SOP and you do it this way. And after you've done it perfectly 5 times, maybe with AI 10 times, you go, I don't need to watch you anymore. Crack on. Yeah, thank you very much. So you build that. The other bit is observability. And I think more and more we'll hear about this over the next few months because we're moving from— in the past, you had to learn the violin and struggle for years and years and master the violin. And if you wanted to build an orchestra, oh God, now I've got to go learn the drums. Now I've got to go play a different instrument over here. More and more, we're going to be the conductor of the orchestra. We don't need to learn how to play all of the instruments. In that orchestra, but we are going to be the conductor and we need to see that everybody is doing their bit and playing their part and being able to see all the stuff you just said. Like, I want you to write your comments and put them here where the human can observe, spot mistakes early. Hang on, why has it gone rogue? Or did it actually forget to do that thing it shouldn't? Because that's code. Where you've set it up is perfect. You've got code saying, do this at Monday morning at 4 AM every single Monday. GitHub Action you mentioned, there's loads of different ways that you can tell it to do the same thing every time. But that's code. It's not, "Should I do it at 4? Maybe 5. Maybe I'll skip it and I'll do it on Tuesday." Because if you give the model a prompt text, it's a strongly worded suggestion.

Yeah.

And it might decide to do it Tuesday. But if you turn that into code, it's going to do it the same way every time. So build trust in it. Get it to the point where you don't need to watch every single thing that it does. But then having that sort of overarching— I've got dashboards up the wazoo now that I'm slowly pulling into one overarching dashboard because I'm starting to build trust in different places. And I just want to know— it's management by exception. Just tell me when something goes wrong. Tell me flashing orange when I need to go in and make the change.

Yeah, I'd love to pick up another Lego piece, which I think just before we hit record, you said, look, I think this comes before connectors, which is skills. And I feel like, again, you were one of the first to spot this for me. As soon as Claude released it, I remember you shooting me a text and saying, look, you really got to look into skills because skill is an SOP for computers. And there's more to it than that, but this Lego piece. I think we kind of picked up one Lego piece. Maybe there's around the context, and it'll build its own context as we go. Maybe it's not the first thing, but it's an important piece to understand the way that the AI is working, which is we just got to give it the data to know everything about you. And that data, some of it you're going to provide, and we'll talk a little bit about some connectors because it will pull some of the data elsewhere. But this this idea of skills. I'd love to get your take on that.

I think skills are one of the most important things for any business to learn in 2026. Because you'll build on this, it, it, all these Lego pieces start to compound. And it's exactly like you go bankrupt. I'm told Hemingway tells us it's gradually and then suddenly.

Yeah.

And so it, you feel like Nothing's really happening. I'm playing with all these different Lego pieces. I've made my way to room 10 in the haunted house. When am I going to get the benefit out of this, Mike? It's like, it's coming. And there'll be this moment where it's just, oh, it's like that. And then boom. And then all these pieces start to compound and you put the Lego pieces together in different ways and it takes off. Skills are one of those things. So I've built a website to show the difference of using AI versus using AI with skills. Yeah, 8020skill.com. But what are skills? Skills are at their most basic, really just a saved prompt. Everything is a prompt at the end of the day, but people think they have to be a prompt engineer, whatever the hell that means, and write these 14-paragraph things into ChatGPT in order to get something wonderful out. What a skill does is it takes the the knowledge and the wisdom of someone that knows how to solve the problem. We started with problems. How do you solve this problem? It doesn't require everybody to have to figure out how to solve the problem. One person can create the skill that reliably solves that problem in pretty much the same way every time. And again, it's a mix of AI and automation. So the AI bit is basically the prompt that says, this is what you're doing. This is the problem we're solving. This is, you know, these are the various pieces involved. And this is how I want you to put them together, combined with the automation piece, which might be some different scripts. Yeah, so I have a Gmail skill. And it's got a bunch of scripts in there that it's written once, but now it can reuse every time it uses the skill to attach an attachment to an email, to read an attachment email, to create a draft, to add a label, to mean all these little things that you do kind of without thinking about it when you're using a tool like Gmail. But it's code. I didn't write any of that code. It's figured it out. First time it bumped up against the wall and went, oh, I need to use the AI bit, the thinking bit to go, oh, I need to figure out a way past this problem. I know, I'll read the documentation. I'll write a little bit of code, see if that works. No, not quite, almost. Oh, read it. Da da da da—

ah! Got it? Yeah.

And you can see it there. This pops up in Claude, you can see it thinking, and it tells you all this stuff it's doing. And once you've got it figured out, basically say awesome on that now save just like bottle it. Yeah, let's bottle it put that in Amber bang yeah, and then over time of course you can always improve that. It'll improve itself if Gmail changes and the script stops working. It'll go read the documentation and improve the script. But also we will change it and say, ah, actually, given that you can do that and that and that, I now want to do this and I want you to combine this thing with this other thing and these other 3 skills over here and put them together in one big mega skill so that you just do all of that stuff for me now. And it will go, yeah, okay. You start to chain these skills together. So a skill is just a reusable, not even a Lego brick now. So now we're building like a little mini Lego model. Yeah, with a few different Lego bricks and saving that and going, right, one of those, please. One of those, please. I'll have a number 14.

And I've seen like in the early days before skills were around, people would stitch together 10 different prompts or they would write a mega prompt to try and capture all of these steps. That was kind of like version 1 of this, whereas now a skill has evolved that thinking and skills can reference other skills.

Yes.

Uh, and yeah, I feel like, again, I could keep— this is maybe just because I think in terms of the SOPs, I hope the description for a lot of my audience, it'll make sense. It is an SOP for a computer.

Yeah.

It's predefined, uh, problem solving.

But it's, it's, it's that and it's flexible. So yes, yeah, part of that SOP will say here are the 13 steps and step 5 is run this script. So we know it does exactly the same thing every time. And step 8 is use this template so that you use my colors, my fonts, my heading style when creating the PDF or the dashboard spreadsheet, whatever it is. But then you're able to mix in that bit where, where we used to have to be the glue between these bits of automation.

It's the glue. I love—

used to have to be us.

Yeah.

We'd have a Lego brick, which was this piece of automation, and it would kick it back up to us and we'd go, uh, Well, yeah, obviously, and now do this with that. But there was a bit of thinking. It was the stuff that you couldn't fit into an Excel formula.

I feel like this is what's made this technology the game changer because historically we would use things like N810 where we would have step 1, step 2, step 3, step 4, step 5. Oh, it's broken in step number 5. Or when your developer disappears and it's like, oh, we got to do that. And why did it break? Well, it broke because the answer it gave back, you know, one line had changed or maybe the button had now moved up here.

It was a subjective decision at step 5. So it did the automation, nothing broke, but then it's, well, you couldn't code the next bit. It wasn't if this, then that, and if that, then this. It was, well, it depends if given this and with a bit of that and a sprinkling of this, then it should wander off down this path. But you needed a human to make that judgment call. Halfway through the automation, so it would just stop.

And now all the human is doing is making a judgment based on the context it's given it. Whereas if we're giving it all of the context and it can chew through so much more context and remain, you know, keep it in its memory, it oftentimes can make better decisions than a human in certain parts because it's got access to that context and it can go, based on everything that I know, I think this is the best next correct decision. Let's continue on our SOP.

And it's the speed that that gives you because, okay, I'm going to try and describe this by waving my arms around. You'll be familiar with this because you've read the blog post, but I have this blog post called 10/80/10. The idea was that you would do the 10 bits and somebody else would do the 80%. Very, very similar. I'm not saying that he used inspiration from this, but very similar to Dan Sullivan's wonderful book, Who Not How. Yeah, because it's the, the who, the person doing most of the work. Yeah. Our jobs as leaders or a project manager or the person who wants to get stuff off your plate, our job is— I'll do it the other way around. So this is the start and then the process and this is the end. At the start, it's what we've talked about. It's defining the outcome. What is done and done well? In order to do this, these are the resources available to you. This is— these are the tools we have. This is your budget. This is what you can use. And then something would go do all of the stuff in the middle, which would usually be a human being working for you. And they might have human beings working for them. And it was really kind of fractal, but you sent this thing off into this process. They would occasionally bounce stuff back to you because you'd want that. You're like, okay, this is a multi-month project. I need you to kick back up every 3 weeks, tell me where things are at, course correct a bit, now carry on. And off the humans would go. And eventually we get to the end. When somebody presents you with something. And we've all hired somebody where a thing gets presented to us and we get the big red pen out and go, well, this would have been faster to do myself. But instead of doing that, you give feedback and you train that human of how you want the thing done better next time, or how that could be improved so that they're able to— because it doesn't start 10-80-10, it starts let's call it 30, 20, 50, and somebody else is doing the middle bit, but it's certainly not 80% of the work. And then you spend loads of time at the end correcting stuff, training, feedback. Now we're moving to this world of 198, 1, where we can describe the thing that we want. And then instead of loads of different humans down here, it's kind of AI automation, AI automation, And the automation is able to run, kick something back up but not all the way back up to us, but kick it back up to the AI to make the judgment call, the subjective thing, write a bit, think a bit, decide something, next bit of automation, boom-boom-boom-boom-boom-boom-boom-boom, then kick it out at the end. And most of the time we go holy shit! That's really good, carry on. And occasionally we'll go actually, next time you do this I want you to do this bit different. And that's that little bit of feedback, that little bit of training that improves skill.

Yep.

So it gets better next time. And it just, that's the iterative bit. It just keeps getting better. Version 1, version 2, version 3.

This goes back to something we talk about a lot. Just when we think about systems thinking and how systems compound, skills compound, using this AI compounds.

And the crazy bit, and this is the bit that really surprised me, is your ambition compounds. So as your skills compound, as you use this stuff more, as the trust compounds, your ambition starts to compound. You start to ask it for more and more stuff. You start to push it to go, surely it won't be able to do this. Aha, I've got one for you, brain. Oh, it did it.

Yeah.

In 5 minutes.

Yeah.

Okay. And then because it can do that, you go, all right then, let's see if you can do this. And you just start to push it more and more and it gradually, and then suddenly, and all of a sudden, you're kind of like, what can't this thing do? Yeah, let's, let's really push it. Let's find the way out corners of, let's throw a ton of stuff at this. And just, and my prompts now, because it has all of that context, my brain has something like 12,000 files in it. A lot of that's research and YouTube transcripts, but a lot of it are documents that the brain has written. It's its own notes on stuff, its own memories and so on. And I'm just, I will now say, right, hold down the key on my keyboard. I want you to do that with a bit of this. Then when you've done that, go over to there, do this, update that, make sure the documentation over here is in line with all of that. Then push all of that and da da da. And yeah, it's a 5-minute ramble with about 16 different tasks in there. It does it flawlessly.

Yeah.

Most of the time.

I almost feel like now's the time to pick up another one of these Lego pieces because you said it right there where you'll give it some instructions and then it'll go off. And it has connectors.

Yes.

And that's where it can draw more information from sitting in other tools and maybe even comes back to something that you mentioned earlier. Also, at some point we'll get to this idea of, you know, like headless software where you just go straight to it and ask it to do its thing in it. Goes and does that in the background. But the connectors was another one of these big unlocks with this. You know, I think the context for me was huge, understanding, you know, the AI is this glue. Then where can we pull more context from? And there's this whole world that opens up when you start to look at MCP servers and it doesn't even need that always. I mean, if they've got great APIs and plugging in, computer can then talk to other computers.

Let's make it even easier there, because if you're listening to this and going, yeah, CP, API, what the hell are they talking about? I would start with a tool like Claude Cowork. Yeah, so there's a desktop app from Claude. ChatGPT have one as well called Codex app. But I really, really like Cowork. I think, I think it's where a lot of small businesses will be doing their AI stuff by the end of the year. Because it looks like ChatGPT, which is what most people think AI is. It's far, far more powerful than the free version of ChatGPT, almost as powerful as the geeky tool that we're playing with, Claude Code. It's, it's, it's maybe like the baby brother of Claude Code, but just a much easier, more familiar place to play. And there are all of these connectors built in. So you literally just choose from a list, tick a box, Maybe you have to give some sort of authorization, you go over to Asana, log in and say, yes, Cowork can look at this stuff. And again, you build trust, you, you don't start on day one saying you can delete anything you like, you can do anything, you can be super admin, and you can kick people out, you start with, you can read everything. And that's it. You can read all of my tasks, you can read my emails, you can look at my calendar, but you can't delete events from my calendar. Because we're going to build trust. We've got to see that it can do the thing and then give it a little bit more, just like the junior employee.

Someone used this metaphor and I thought that is bang on. A tool like Cowork is like giving a chainsaw to a toddler. Like, you— a chainsaw is an incredibly powerful tool if you know what you're doing, or it can be deadly and dangerous because you're exactly right. I mean, if you don't know what you're doing, you could connected up to your website or your email and just— it's happened in one fell swoop. It just deletes everything because you told it and you didn't quite understand or your instructions weren't clear. I think that's, um, yeah, something we need to build up.

It's the fear of that that holds people back from connecting up these tools. And until you connect the tools, you don't see the huge real power that they can. And so there's this, there is a bit of tension there and I've got a lot of customers in Europe. We won't name countries by name, but there are certain countries in Europe, very eins, zwei, eins, zwei. They're good, but like to do things just perfectly and dot the i's and cross the t's first. And then like, show me, tell me what it's going to do. Well, I can't tell you, it's a bit of an experiment. We don't know.

We've got to play.

Yeah. And then I think the Aussie culture is a bit more, mate, she'll be right. Let's just pile in. And if something goes wrong, we'll fix it later. And I think that second approach is a much better option It's— you've also got to weigh the risks. Yes, there was a guy on Twitter the other day that he was an actual developer and the AI deleted his database and he lost 3 months of work. He was able to find a backup, backup, backup that actually worked and kind of get back to where he was. But it was, it was a very painful process. People will see those stories and go, see, AI bad. It hallucinates, it makes mistakes, it deletes files it's not meant to. But it's probably a 1 in a million at this point. Yes, it can happen, but you've got to weigh the probability.

I feel like there's this risk curve. You can go so far out on the risk curve if you do want to do it, and you can be playing with OpenClaw, and you can give it access to your bank account, and you can be doing— like, that's right out at one extreme.

You don't have to go there.

No, but yeah, we can come back closer. I think Cowork is a great example. And then using one of their out-of-the-box connectors. And this is like when the internet was launched and everybody was like, oh, I don't want to enter my credit card details online because I don't know what they're going to do with it. But now everybody does it. We save our credit card on our Apple Pay. Yeah.

I remember arguing with, with business owners when I was on stage starting my, like, speaking career, 2005, 2006. And they refused to use something like Dropbox, maybe 2007, '08. I'm like, but the benefits and getting rid of that box in the, in your closet. Yeah. And the IT geeks that you pay 80 grand a year to back then. And it's like you've got to weigh the, the risk versus the gain. Most of this is huge gain. I also worry for those businesses that put their head in the sand and go, nope, not worth taking the risk. This whole thing's a fad, it's going to blow over. This is the most transformational technology of our generation, possibly of our lives. And it is going to change everything. We don't know by what magnitude yet, we don't know exactly how, but that shouldn't stop us experimenting with it and figuring out for you, how can you use this to enhance you to make you better at whether it's your job, whether it's your business, whether it's your home life. I've got members using it for estate planning and divorce proceedings and baby planning and all these different things in their lives that are making them so much more productive and happy and, you know, playing with music and classifying 16,000 music tracks because the guy's a DJ and wants to catalog his stuff that's been sitting on a to-do list for years. And you can now build a little app or have it just, have it just organize your hard drive. One of my members, one of the first things he did with it, he said, I've got 420 wedding photos that have been sitting here for over a year with like, you know, IMG_96735. I haven't— where do you even start? So he just had his AI write a better file name, go in, look at the picture and write the file name of what the picture actually is.

I know what I'm doing tonight.

And then tidy those up, get rid of the duplicates, put them them in different folders, and now I can actually go create the book of all the photos that I was meant to. Happy wedding anniversary for Monday.

Thank you. May the fourth be with you.

Very easy to remember.

I, um, I have that exact problem and it's great to know that AI can solve it. I think, I mean, that's the hardest bit now. Once you understand this technology is focus and where do you put your attention? Because it feels limitless on what it can do.

Well, it's back to where we started. It's figure out which problem you want it to help you with. What's a big enough problem where it's actually going to make a difference in your life? Because it feels silly to use AI and then to go spend 2 hours with AI to solve something that used to take you 10 minutes. It feels like I'm going backwards. I don't have time for this. I'm running at 110%. My day is full. I've got a little gap in between meetings. And it's taking the time and evenings and weekends to begin with to, to automate that little task. And if it took 2 hours, what's that? 12 lots of 10 minutes. And maybe this is something that you do twice a week. So it's going to take you 12 lots of that, 6 weeks before you're getting a return on investment of that time. But all the weeks after that, there's this extra 10 minutes that you now have to go build the next little thing or to organize the wedding photos or to write a bit of context. Don't write the context, just talk the context. It's much, much easier. This is our business. These are the things we sell. These are who are important to us. These are the clients. This is the team I work with. This is me. This is how I like to work. Just as you said, talk all of that stuff into your phone, drop that transcript into an AI, even if it's Claude Cowork. Yeah, start with a super geeky version and start there and see the magic. Can we just talk about 80/20 Skill briefly? Yeah, I built this to show people the difference between just using AI, just writing a prompt and maybe dropping in a CSV file or a couple of documents and saying, do this thing, versus using a skill. So I've kind of gamified it. There's 3 levels. And level 1 is just very, very simple, couple of quick demos, you can whiz through those. Level 2, I give you a skill to do a particular thing based on your job title. There's a few to choose from. Because I want to match the, the use case, the example to your daily work, because the closer those are, the more useful it is. If I give everybody a recording a podcast example, for most of the audience watching, that isn't very useful. And so therefore AI is rubbish. But if I can match it to your life and your work, then it becomes, oh, that, I need that.

Yeah.

And so that's what 80Skills is designed to do. Second level is a skill. Now go find a file. Oh, that's a bit of a pain. I'll give you a file that you can use just for the example. And then the prompt that you kind of say, do this with this, it's just a single line. It's a single sentence because I don't want people to feel like Oh, that's a really complicated prompt. I wouldn't have known how to do that. I wouldn't have known the words to say. So the prompt is little on purpose because it's the skill that's doing the work. So you download the skill, you drop that into Cowork. You can download the little example file and drop that in. And then that one line copy paste. So if you can drag and drop and copy paste, you're good. Drop that into Cowork. And then it shows you the difference between using AI normally and using AI with a skill.

Yeah.

And it's night and day difference. So you finish that and there's gamification in there and I'll let you play and find the Super Mario level if you can. And then level 3 is the connectors. So then sort of like, did you know that AI can talk to your Gmail and your calendar and all the stuff we've talked about? Asana, Monday, ClickUp, doesn't matter which task management you use. Yeah, it's in there. Excel, Google Sheets.

Zero, history.

Yeah. And don't give it your bank details on day one, but Westpac, you can start by, yeah, you know, you go to your bank, you download a CSV of all your transactions, you can drop that in and say, categorize my expenses for the last year. Or if it's the business CSV, what software are we paying for that maybe we can get rid of? And it'll say, well, you got 3 versions of this thing. And you Is anybody in your business actually using this? I don't know. Log into it and see who logged in last. And oh, what do you know, 80% of the business don't actually use this thing that we're paying $50 a month for. Yeah, okay. Let's get rid of that. And it can just help you in so many different ways. But what's the problem? A CFO will have very different problems. Yeah, the CEO will have very different problems to the person working in an agency or working in a podcast studio. That they've got different problems, which is where you started.

Yeah. And like one thing that when I learned about this skills thing, it reminded me of The Matrix and there's a scene. Yeah.

You know the one?

And then it's like, uh, he plugs in and then he goes, now I know kung fu. And it feels like that with a skill. I, uh, now there's also like a community. People use GitHub and they're sharing skills and packaging up all of this knowledge. And you can just point your, uh, you have— with caution, with caution, uh, and say, hey, I want you to learn this thing. Tell me if there's anything of value. And there are some skills that are packaged up by Anthropic, the people who make Claude.

Like, those are safe.

Got like a PowerPoint, uh, skill that gives a great understanding of how to build out PowerPoints. And then I match that with a little bit of context. I say, Here's my brand style guide. Here's the transcript of a call I had with a prospect. And here is a template of a proposal document. Can you mishmash it all together and make a version using this skill of this particular slide deck for the client, identifying all of the pain points? 5 minutes, done. Extraordinary PowerPoint. Like, but again, it's, it's one of those things you generally come to AI and ask it questions that you know, or the area of expertise that you're really great at. It's, that's the skill we've got to start to learn is how do we ask things? Because it, it knows areas that you know nothing about.

So I have a, a Grill Me skill. I've been using a version of this for 3 years, I want to say. So instead of you prompting the AI, the AI starts asking you questions.

Yeah.

And the wording is essentially you can do this with a prompt. You don't need the skill, but it's essentially ask me a whole bunch of questions until you're 95% sure that we're on the same page, because that's the goal of having the chat before you do something right. Again, back to the junior employee. Let's make sure that we understand what this problem is, what the task is, what the job to be done is, what resources do we have available, If it was a human in your business, you'd be saying, how much time do I want you to spend on this? Because we've probably all been in that situation where we've expected someone to go off and spend 2, 3 hours. And they've wandered off for 4 days. And they're like, I didn't know, I didn't— oh goodness. Next time, just 2 hours, please. But same sort of thing, or a budget. I'm going to give you $1,000, you can't spend more than $1,000 to go solve this you're putting on an event, people are going to come into the office, we're going to have drinks and nibbles, 5 grand, don't go over that.

Yeah.

You know, there's some sort of, there are these guidelines and constraints that, that's the hardest bit. It's just getting those out of your head and making sure that it's verbalized, it's written down somewhere, it's in the transcript that the person or the AI understands those guidelines and constraints so that it can work within those. And then Off you go. You gotta do the thing.

In the tail end, because one of your skills being the visionary that you are and you can see where things are going, I'd like to ask almost like a little bit of a selfish question around like the direction I see that we're heading with System Hub, you know, and you were the one that were like, hey, you've got to get MCP server on System Hub, which we are like, I'm beta testing. Now. And then, I mean, that's the first thing. It's like, how do you get everything into System Hub in the easiest possible way? And then I think of, you know, what, what is the purpose and where does System Hub fit in this AI world? And I realize a lot of the work that I do is making the invisible visible. It's trapped in people's brains or it's trapped. And even when we're working with AI and we ask it to go do some things, you know, even the way I approach it with Asana, and I'm getting it to try and do things visibly. And that's really— yes, I see System Hub as, uh, the, the central shareable location that we just like, you know, it's like a portable business brain that anyone on your staff can plug into and connect to their AI to give it the predefined context. And there's other ways to package it up and share it. I've done it with different team members and then we find, you know, 6 months down the track, their version of their brain, because they're building their context, is vastly different than mine and there's no central knowledge component. So then I'm starting to think, you know, this is position descriptions, SOPs, policies. We're even starting to build out skill packages now, which live inside System Hub. So it's almost like someone can just plug in and then say, hey, I'm starting to use Claude Cowork. And the first connection we make is over to System Hub here and pull down a lot of that knowledge as the starting point. And, you know, keep referencing back to that because that will then become the information that updates. But I, I, I bounce some of these ideas around with you often and it shapes the direction that we head. And so I'd love— I don't know if there's anything that you can see further down the path or important bits to solve. Like, I try and solve things simply. System Hub has always been about simple. Yes, there's the more complex solution, but I find simple is always the best, most scalable.

You said making the invisible visible. I think part of that is System Hub is Speeding up like time to trust. Yeah, it's getting people to trust in those SOPs. You can allow employees to then improve those SOPs over time because you trust them. And there is this central place of knowledge. What I'm working on right now is taking the brain, which has historically been a single-player thing. Yeah, because I think the way Silicon Valley thinks is I'm an ICP, I'm this geek, and I'm going to have agents working for me. Yes. Dozens, maybe hundreds, maybe thousands one day. They're all going to work for me and it's me at the top. Like, yep. In the, in the real world, we all work with other humans. Yes. Maybe not. Maybe it's a business of one. I'm a business of one now. I've got to 35 people and I don't feel the need to have a business of 35 people again. Because right now it's just so easy delegating, no loss of communication, delegating stuff to different AIs. Most businesses are lots of people and will be lots of agents.

Yes.

And so making the brain multiplayer, yes, is the big thing that I'm working on right now. But you need that central store because if someone changes the business context and says, oh, Bob's business is now doing this. Everybody's brain needs to understand that. Some will be working in co-work. By the way, you don't have to use Claude either. I should have said that before. Yeah, these are just text files, essentially, when, when all said and done. So you can use OpenAI, you can use your own, like, you can bring your own model to the party. I was just playing with some of the different models on Hugging Face the other day. We won't go down that route yet, but there are freely available models out there that you can run on your own computer. So all of this stuff can be free. Yeah, it's not very good yet. And it's very slow compared to the other stuff. But it is possible and it's getting better by the hour. That was a tangent, right? You need this central place so that as context changes, and as somebody on your team creates a skill and solves a problem, yeah, everybody else can now solve that problem. It's like Tesla, right? Tesla will drive and hit a patch of ice and go, oh, edge case. And somehow that edge case gets fixed and the Tesla now knows how to deal with that next time. But all of the other Teslas in the world know how to deal with that problem now. And so it's the same thing. One employee figures out a thing, solves a problem. Oh, I've got a great way of doing that. And now everybody can make use of that. Everybody can hear the violin without having to learn how to play the violin. And so everybody's making everybody else better.

Yeah.

And there is finally a way of doing stuff. You don't have 16 different people in your organization wandering off on 17 different tangents, all doing it slightly differently, which over time just gets worse and worse and worse.

Yeah.

Now everybody's using the same thing, singing off the same songbook, and That alignment is what makes businesses go further, faster.

I mean, this is, uh, your 11 herbs and secret spices. It's figuring out your way. I feel like when you package up this knowledge and you figure out how you and your team does something, like that's a big part of what I think about when I think about systems. We're just oftentimes the, uh, business owner, their DNA is in the business on how they do things, how they solve problems, how they treat clients. And the system side of things is a codified version of that. It's a written-down version. It's documented habits.

It's what Gerber said all those years ago. This is how we do it here.

Yeah.

And the culture is baked into that. Habits are baked into that place. So yes, the way you deal with clients, it's the bit that makes you unique. Yeah. And the more that we can, and let's bring it back to the humans then. The more that we can lean on, not necessarily AI, because people still get freaked out by that, but an automation that does that thing on a Monday morning at 4 AM exactly the same way every time, gets it right every single time, so that we can spend more time on the human stuff, on the relationship stuff, on actually thinking. Because I do want to make the point that we have not talked about you know, all of the media saying that AI is making us dumber. We're not trying to replace how we think. We're doing things, we might use it as an aid for some of the thinking, do a bunch of research, all this stuff in, summarize that, tell me how that YouTube video that I don't want to watch because it's 2 hours long, but tell me how it applies to the 3 big projects I have at the moment. Which bits of the video should I watch? Is it worth me watching? I have skill that does that overnight every night for the 25 channels that I follow. And it highlights which things might be worth watching and how it applies. So I already know going into it, so I'm not wasting as much time. But it enables us as humans to do more stuff. And that is, I think, so, so important. We still do the thinking. Yeah, it can then just— I was listening to a podcast on the way down. And the use case was, you know, all of the information is there in a whole bunch of files on a computer because you've had meetings about the thing and you've dropped some notes in Slack about the thing. And there's 17 Markdown documents sitting somewhere about the thing, but just pulling all of that together and writing that strategy document, final report, whatever it might be. For many of us in the past, that's been an all-nighter. Yeah. Or an all-weekender of just pulling all that together and then formatting the thing and getting it and double-check. And now it's bang, use my template, use my design. And it's really, really good at that sort of stuff, the synthesizing, the summarizing. And yes, of course, we're still going to put our eye over that. We still have to stand by everything that gets sent out. You cannot send out a document, someone comes back and says page 7, blah, oh, I didn't know that was in there. No, like you have to stand behind what you're putting out. You, the human has to be the one accountable for it, but we don't have to do all the grunt work to create the thing.

I think that friction that was previously there, I mean, I see it with creating SOPs and capturing what we do. That was the biggest friction point. Like, I mean, everybody loved to create the CCF and go through the steps and all that, but when it came to the documentation, which, you know, was a key part. But, but now that friction is gone, just that synthesis.

But your genius insight there was that the owner should not be the person creating the SOPs. And you created human grill me skills. Yes. I remember when you and one of your team came in and grilled my team and it was essentially two questions on a loop, right?

Yeah.

Why are you doing that? Then what do you do? And why are you doing that? And then what did you do? And walking them through, but to help them pull that out of their heads. Or we can use an AI to help with that. And now what are we going to do with the thing? What's the problem that we're solving here?

Well, I feel like that's the full loop there. We're back to the problem, which that's where it all starts, which is, you know, what problem are you looking to solve?

And do you have problems? Hang on. Oh yeah. If you have a pulse, then you have problems. Yeah. Part of the human experience.

And then AI can help. So we're going to have a part 2, 3, 4, and many more to come. I hope so. Appreciate all of your generosity and sharing over the years. And this equally was no exception. This is a great, great episode. I know a lot of people are going to get a lot from this. So thank you.

Thank you. Thanks. It's great to be back. And Lovely to see the new studio and thanks for having me.

Where is the best place for people to keep an eye on what you're doing?

If you haven't played with skills at all, play with 8020skill.com. It's completely free. There's a whole course in there on using Cowork as well.

Yep.

Um, because you just need that visceral moment of, oh, I can do that. And then once you see that, you can't unsee it. And then that gives you the the motivation to spend— I almost said waste— to spend, invest a few evenings and weekends playing with it, experimenting. I think you need to invest somewhere in the region of 10 to 20 hours. Yeah. Knowing that it's going to be gradual, gradual, gradual, and then suddenly it's useful. But that first 15, 20 hours might feel like wasted time, especially when we're all so busy.. But there is a point somewhere in that 10 to 20 hours where it's like, oh, I get it. Oh my God, I can't unsee it. Great. The other place is adds2ai.com. And that's my community where I— it's the only way to get a copy of The Brain. You have to be a community member. I don't sell The Brain separately. And what's coming is Team Brain and then Client Brain, a version of The Brain for small businesses. That's, that's what I'm working on at the moment. And that'll be ready soon.

Absolutely. I'll pop a link through to everything there, and next time we'll dive into some of those topics as well. So thank you.