DH 370 Alastair McDermott
===
Introduction and Guest Welcome
---
Jonathan Stark: Hello, and welcome to Ditching Hourly. I'm Jonathan Stark. Today I am joined by guest Alistair McDermott. Alistair, welcome to the show.
Alastair McDermott: Thanks for having me, Jonathan. It's great to be here.
Jonathan Stark: You know, I was thinking before the show that you had been on before, but I think you haven't. I think this is your first time here and, and what I'm remembering is being on one of your shows.
Alastair McDermott: Actually run two of my shows, Jonathan. So
Jonathan Stark: Okay?
Alastair McDermott: we, we've, we've, we've talked a bunch of times, but yeah.
Jonathan Stark: Yes.
Alastair McDermott: great. It's great to be on this show for, um, finally, uh, I've been listening to this for, uh, a long, long time, so
Jonathan Stark: good.
That's great. Well, today we're here to talk about your new book, use AI Stay Human, A Survival Guide for Experts who Wanna Stay Relevant, authentic, and Indispensable in the Age of ai. So I'm really excited to talk about that.
Alistair's Background and Journey
---
Jonathan Stark: But first, could you tell folks a little bit about who you are and what you do?
Alastair McDermott: Uh, yeah. Cool. So, um, background, like, I think a lot of people listen to your show. Uh, I have a techie background. I was software developer for corporation for years. Um, I left them in 2007, started my own business, um, fumbled around for a long time trying to figure out, I, I sold SEO as a service. Um, that was the first thing out the gate. Um, 'cause it was kind of hard for a, a system software engineer to find, um, some way of running a kind of their own business, their own agency. And,
Jonathan Stark: Yeah.
Alastair McDermott: so. I went through the, I went through the whole process of, you know, um, being there, done that, bought the t-shirt in, in terms of making every single business mistake you can possibly make. Um, and, uh, yeah, so I eventually ended up with a business, an agency called Website Doctor, and I kind of knew that I wanted it to be more of a consulting business than an agency. I didn't really want the agency model, I wanted the expert consulting model. so that developed and, and, um, that eventually developed into a brand called the Recognized Authority, which some of your listeners may know, uh, because you've been on the recognized authority, uh, at least once, if not twice. so what I was doing was I was helping people who were invisible experts to become known as a recognized authority in their field, uh, which is something I kind of feel passionately about. And, um, and when all of that was happening. At the same time I saw this kind of incoming, uh, tidal wave of ai and I was thinking if I continue down this road of just talking about building authority, and I, I don't acknowledge the AI piece. Uh, it's just, it's, it's like sipping martinis on the beach with the tidal wave oncoming. So I decided, okay, I've got to understand this. Um, and I've got an advantage in that I have a techie background so I can understand a lot of the, the technical aspects of it as well. And so I rebranded the business completely from Website Doctor and the recognized authority.
All, uh, all are now under Human spark.ai, and that's the, the name of the brand. Um, and what I do is I help people, um, particularly I love helping experts, uh, to use AI and, and to become super productive with it.
Jonathan Stark: Awesome. Okay. And I, I think I saw a. LinkedIn post about the book, which is, and I, I jumped on either that or an email and I jumped on it and I said, Hey, let's get you on the show right away.
The New Book: Use AI, Stay Human
---
Jonathan Stark: So tell folks a little bit about the book.
Alastair McDermott: Yeah, so I think, um. One of the things is that I think there's a lot of people are threatened by ai and, and I think like, it, it's right to feel a little bit of that because, um, like it, it's, it's just disrupting everything. but at the same time, I think that for true experts, AI can be like a force multiplier.
Rather than a threat. And, and I, I think that it's a false choice to say that you should use AI or not use AI, and, and, and if you use ai, you know, you're not being authentic, you're not being credible. I think it's possible to stay a credible expert and still use your own human judgment and your own human relationships. But use this amazing tool that we have, uh, to, uh, I'm, I'm trying to stay away from all the cliched words, like amplify or, know, all of all of these horrible, uh, overused words now, which is part of the, part of the problem that we have with trying to use these, these things, but to use it to, uh. help us to, um, to, to get our word out to be more scalable without becoming a spammer, but actually creating high quality content that's based on our codified expertise, which is something that I think AI can really help us with.
AI as a Tool for Experts
---
Jonathan Stark: Do you mostly recommend using it for marketing reasons or also delivery and creation of ip, those sorts of things? Is it. Of like a holistic approach that that you have for it.
Alastair McDermott: Um, I think marketing is probably one of the weakest things that we can use it for. It's, it's not bad for marketing, but there, there's so much more that we can do with it. And I, I think, um, like I, I, I see kind of four different opportunities that we can use for ai. And one of those is productivity, and that's kind of table stakes.
Everybody talks about using AI for productivity. And like, don't get me wrong, it's really great for productivity. You can get a huge amount from it. Um. I've done some, uh, projects where we saw like 10000% ROI and, and things like that. You know, reducing a two hour process down to a five minute process,
Jonathan Stark: Right.
Alastair McDermott: um, that kind of stuff is possible to do. but there's other things that we can get from ai, like for example, uh, apart from just productivity, if you think about it in terms of capabilities. So. AI can give us extra capabilities that we simply didn't have before. So like as a business, you have the capability maybe to deliver, uh, something that you couldn't deliver before, like, for example, design components or, um, you know, an app or you know, some sort of web-based access to whatever it is that you have.
You've got the capability to deliver new things that you didn't have before. So you, you as a business can now do more. Uh, and then there is, um, smarter decision making. So in part, that could be simply using. AI systems to help us make decisions, um, in part by just, uh, like telling them about decision making frameworks and getting them to walk us through it.
It's very simple and, uh, I built a bunch, bunch of free tools to help me make decisions, like put some of 'em out there. If anybody can go look up my name on, on Chat Boutique and find them. But you can actually get it to walk you through a decision and give it the background and say, Hey, can you help me figure out which, which decision making framework?
'cause there's tons and tons of them that would be best to use here. Uh, and then as well, we can also give it data. So, uh, we can give it specific data and we can say, can you help us make the decision based on this data that we have? So you're making smarter decisions, you've got more capabilities, you're more productive.
And then the last one is accelerated learning. And we can learn things particularly, uh, for, for experts, and I think this is where we get superpowers as experts, subject matter experts using ai. Um, if you're using AI at a, at an expert level as well, you kind of have this superpower. So, yeah, like you can tell I'm really passionate about this stuff because I, I, I, I really love. The, the positive sides of this and like, don't get me wrong, I understand there's tons and tons of negative sides to AI as well, and, and like, I'm, I'm not unrealistic about that, but I think that the positive sides of this and what it can do for us are, are just amazing.
Jonathan Stark: Mm. Well, I I love that your, uh, focus is specifically on experts, which is, you know, the audience for this podcast and you're being one yourself.
Practical Applications of AI
---
Jonathan Stark: How have you used it? Like, what are some examples maybe from, from those list of things where you've used it? I, I, you know, for example, how did it play into writing this book?
I'm sure you used it quite a.
Alastair McDermott: Yeah, so I use it in a very nerdy way. Um, I used VS Code, which is Visual Studio Code, and I used a, a plugin called RU Code for that, which allows you to connect up, uh, AI to Visual Studio. So Video Visual Studio is a coding environment. It's just a way of, of editing text files. But one of the nice things about it is it has built-in versioning. So you can have version control through a tool called gi. Um, I don't wanna get too techy on it, but basically what this allows us to do is it allows us to have, uh, like version one of a file and version two of a file, version three of a file. And now when we can connect that up with ai, we can say, Hey, I want you to, you know, rewrite the, the fourth paragraph in chapter two. And then we can see the before and the after because one of the problems that we have with AI is, is quite often, um, you know, you'll give it some text. You'll ask it to make a change, it'll make that change, but it'll also make another change that you didn't realize, and you know it, that, that goes undersurface
Jonathan Stark: Yeah.
Alastair McDermott: when it's problematic.
Jonathan Stark: Yeah.
Alastair McDermott: um, I really love having version control around ai that, that helps a lot. And then one of the really cool things about this type of system is, uh, like what you can do with it in terms of setting up these kind of agentic systems, which are when you can give. Ai, different personalities or different modes. And so for example, in in, in the system that I have, I have an orchestrator mode, which is like the, the kind of the, the book editor in chief. that orchestrator then, uh, commands these other modes that I have. Like I've got a fact checker mode, which can check for facts. I've got a mode which is just for rewriting into my voice.
Jonathan Stark: Mm-hmm.
Alastair McDermott: a mode which is quite extensive, which is actually a beta reader mode.
Jonathan Stark: Mm-hmm.
Alastair McDermott: So the beta reader mode actually has a file associated with it that defines five personalities of five different beta reader personalities. And what it does is it reads the chapter, chapter one by one from the context of this synthetic beta reader.
Jonathan Stark: Mm.
Alastair McDermott: saying it's as good as a human beta reader. If any of your listeners have ever written a book, trying to get feedback
Jonathan Stark: Yeah.
Alastair McDermott: from beta readers is a nightmare.
Jonathan Stark: Yeah.
Alastair McDermott: actually being able to programmatically get, uh, five different perspectives from five different beta readers, albeit not human, but that doesn't mean that the feedback is not useful. Um, and that allows me then to have a conversation. So I'm able to have a conversation with my editor and say, Hey, you know, I think we really need to upgrade, update this chapter, you know, with, with. Whatever it is that the, the, the point that we're focusing on, uh, because the feedback and, and you know, you can also ignore the feedback if it doesn't feel quite right.
But that allows me to go from having, uh, a concept to a draft chapter, to feedback on that chapter and, and, uh, and back again, you know, in, in hours instead of in months. And, and that
Jonathan Stark: Hmm
Alastair McDermott: a huge difference just in terms of product productivity and everything. Um, but uh, having those systems and thinking about how those systems work together, I think that's a really core skill.
Jonathan Stark: Hmm.
Technical Deep Dive: AI in Writing
---
Jonathan Stark: Okay, so let's get a little bit more techie. It's a lot of software developers listening. So what. How are you orchestrating this? Is it vs code? Is it scripts that you're writing? Are you just, do you have the orchestrator say like, okay, just do this stuff?
Alastair McDermott: Okay, so, so vs. Code is plugged into a system, uh, sorry, has a plugin, which a system called RU Code. Ru code is, uh, like an open source version of Cursor, which a lot of people are using.
Jonathan Stark: Mm-hmm.
Alastair McDermott: uh, what I like about it is it's open source. I've, I've. Given feedback on, on some of the, the bugs and stuff. And they've actually incorporated them already, which is really cool. Um, so I have a system set up where I have, within my book project, I have what's called a, a custom modes file, which is a ru modes, Ando modes. And in there you can define the modes or the different kind of personalities that, that a, um, that the AI can, can take. And so within that, I have an orchestrator mode and the orchestrator.
Is not allowed to write to files. It can only read files and give commands to, to kind of subagents. And the subagents then can, can act. Um, so for example, the, the rewriter which re rewrite something into my voice, um, if it, if it comes across some text that I, that the orchestrator is told it, Hey, go rewrite this.
It's able to do the rewrite into my voice. And it has a specific set of rules. Uh, the beta reader mode is not allowed to write to files except. The beta reader feedback file.
Jonathan Stark: Mm-hmm.
Alastair McDermott: write to that file. And so I can run, like, I can run a, um, run a beta read, for example. Uh, it'll update that file. And then what I can do is I can say, okay, let's triage the feedback that we've got.
Let's prioritize it and let's look at, you know, what we need to do in the context of, because ideally you're gonna want to write for one single reader ultimately.
Jonathan Stark: Mm-hmm.
Alastair McDermott: this feedback from different perspectives is useful. Um, and so I say, okay, so would our ideal reader feel the same way as, so for example, one of the beta readers I, I always use is like a hard noses veteran who's a
Jonathan Stark: Yeah. Mm-hmm.
Alastair McDermott: it's really good to have that perspective reading the book. Um, and so sometimes I ignore what the hard noses veteran will say 'cause they're a little bit too cynical about things. Uh, but sometimes that perspective is really useful as well. So
Jonathan Stark: Hmm.
Alastair McDermott: just, um, kind of, uh, uh, like I can go more detail than that, but I think that gives a, a, a good overview.
Jonathan Stark: Well, let's, let's zoom about a little bit. For people. 'cause a lot of people I talk to in my, uh, orbit probably are gonna write a book or, um, their first book or maybe write another book. They've already got one. And you've written a few books before this.
So what, what did you give it at the very beginning?
Like just a table of contents and you said Start, start this. How did you train it on? How does it know what your voice is? So those, like how much traditional writing did you do before you set the AI agents loose on it?
Alastair McDermott: So, um, you were a contributor to the book, uh, because you were on my podcast, uh, the AI Power Thought Leader. And so I did a podcast, um, about a year and a half ago now,
Jonathan Stark: Mm-hmm.
Alastair McDermott: I, I was kind of examining, like I, I knew that I, okay. I. that point we were very early on in this AI revolution, and, uh, like I thought, like it's not that we don't know the solutions, we don't even know the problems at this point.
So I was trying to figure out, okay, what are the problems that we're, that we're, we're fitting, we're, we're, we're gonna hit here? And, and one of those was like you and I spoke about, you know, we, we may have to redefine the word author because it's gonna mean something different, you
Jonathan Stark: Yeah.
Alastair McDermott: Um, that was something that we talked about. And I think that's still in the book. Like, that's still like one of those things where I say like, we do have to think about that. Um, but I still did take a, a kind of a traditional approach to, um, to gathering the knowledge. I had conversations with real people. And thought about it a lot. Um, I have in the book a photo of my handwritten notes, uh, with a lovely glass of Danish beer because I was in a hotel in Denmark. Uh, my partner was at a conference and I went along as just a, a kind of, um, just a like, okay, I'm gonna have a, like little
Jonathan Stark: A a plus one. Yeah.
Alastair McDermott: Yeah. Plus one. Yeah. So, so I was, um, I was working on this, um, and so a lot of my notes were handwritten at that point.
Jonathan Stark: Mm-hmm.
Alastair McDermott: then I had, I, like, I brought them into, into, um, conversations with ai and when I say conversations, I mean literally voice conversations.
I would, I would be driving from the west of Ireland up to Dublin on the East coast. And so I'd be in the car for three, four hours and I, I would chat with, you know, um. Advanced voice mode or, or you know, uh, regular voice mode back in the day. Um, and, uh, I'd have a conversation, okay, I'm thinking about doing this.
What, like, how would that work? Um, would that, would that be a good flow?
Jonathan Stark: Hmm.
Alastair McDermott: the other thing that I did, I think is super important was I actually codified some of my knowledge, so I have some very specific. Knowledge that I have from talking to 200 people like you, uh, really smart experts in your field.
In fact, I would say that well over 50% of the people that I've spoken to on my podcast have written at least one, if not more than one book.
Jonathan Stark: Hmm.
Alastair McDermott: of knowledge about. Book writing from my own experience and from other people's. So what I did was I started to document that I, I, I, uh, started to document it as like a framework that I could then say to ai, Hey, I wanna make sure that when I'm writing a book that it follows this framework.
So this kind of general structure, uh, I, so I, I, I don't care about books that aren't, um, kind of educational focused, uh, single problem business books. That's the only kind of book I'm probably ever gonna write.
Jonathan Stark: Mm-hmm.
Alastair McDermott: a sci-fi must talk to you about how you got on with your sci-fi.
Jonathan Stark: Yeah.
Alastair McDermott: I, like, I, I think that I'm, I'm, I'm gonna only write those types of books because that's what I'm good at. So I decided to codify my implicit knowledge. And I think that this is something that experts can, can do with AI is we can say, here's all the things that I think about this, or even better. Here's some conversations I've had with other people about this topic. Let's say I was on a podcast. I take all the podcast transcripts where I spoke about that topic. I put that into ai, and I say, can you pull out the implicit knowledge that I have that maybe I don't even realize that I have the
Jonathan Stark: Right.
Alastair McDermott: and insights, and then start to codify that? And so I did that for book writing,
Jonathan Stark: Mm-hmm.
Alastair McDermott: I gave that to ai and I said, now have these notes. I have this framework, and I have this approximate outline. Let's talk about actually turning this into a plan for the book.
Jonathan Stark: Mm-hmm.
Alastair McDermott: of like, one of the things I love doing with AI is, is mashing several sources like that together and saying, what can you come up with? And then using human judgment,
Jonathan Stark: Mm-hmm.
Alastair McDermott: is the, the really irreplaceable thing to actually look at that and say, okay, probably need a little bit more this, a little bit less of that, you know? So, um, yeah, I, I don't wanna, I don't wanna monologue too much childhood, so lemme
Jonathan Stark: No, no, no. This is perfect. So my experience has been, and I use AI every day. I use it a lot.
Organizing and Managing AI Outputs
---
Jonathan Stark: And, uh, in, in a lot of the ways that you've already mentioned, uh, one thing I've noticed is that it can quickly generate a massive amount of stuff to read. So when you, you want to bring that human judgment into it.
I, it's almost like, I wouldn't say more work. It's a different kind of work, but it's still a lot of work to go through and. Sort of review it all, or like I'll find stuff, I'll, I'll, I'll go down a rabbit hole and I'll think like, wow, that was, that was really interesting. Kind of like, I use it a lot for brainstorming.
So like, just check my logic on this. What am I missing? Are there any holes here? And it gives me answers. And sometimes I agree with them, sometimes I don't. Sometimes they're totally stupid and sometimes they're like, oh, that's a really good one I wouldn't have thought of. And then I can't find like, how do you keep it organized?
Like I can't find where. What the insight was. I almost want like a, an AI to go through my che GBT logs and be like, okay, find all the stuff that's garbage and all the stuff that I liked, or, I mean, it gets, it gets unwieldy pretty quickly if you use it a lot. So, but I'm just using, um, default interfaces, so I'm not using any APIs.
I'm just using like Claude and Chett and, and a teeny bit of Gemini, but in the normal chat mode. And I haven't even used like the projects, the project stuff.
Alastair McDermott: Um, okay, so, so here's a couple of tips that might be useful. so first off, I think it's super useful to make some kind of meta knowledge documents about your business. And so I, like, I created, uh, a bunch of, uh, here's my positioning. Here's our ideal target audience. Here's our services, here's our pricing, um, here's our marketing material.
You know, the whole website goes in there. And, um, quite raw, but, but with some, with some meta commentary around it to explain it.
Jonathan Stark: Mm-hmm.
Alastair McDermott: Then I use that. So, so I have one which is called Human Spark Coach. So this is internal, nobody else can access it. But uh, I can talk to it and it already has the context.
'cause one of the most important things is giving it that context.
Jonathan Stark: Right.
Alastair McDermott: actually have that meta information, uh, version controlled in git. Um, so I, I have, um. I, I don't have it on GitHub 'cause I, I didn't want to go that far for, for privacy reasons. But I
Jonathan Stark: Mm.
Alastair McDermott: locally here in Git.
Jonathan Stark: Mm-hmm.
Alastair McDermott: what that means is that I can then update that and I can kind of see it as it, as it progresses over time.
And, you know, quite often there's an evolution of your services, your positioning, things like that. so having that met information is really useful. Um, I also have codified things like. Uh, if I, if I'm writing a LinkedIn post or a, um, a blog post, uh, I have a lot of that process already prebuilt into a, um, a custom GPT. So, for example, I can say, okay, I have this idea about this topic, um, and I'm thinking about writing a blog post on it. Can you just talk to me about that? And, and, uh, quite often I'll use the dictation mode. I'll just be speaking into my iPhone and I'll hit Go and, and it'll process that and then come back to me.
And then I say, okay, that's the outline of the blog post. Now give it to my blog post writer, which uses, writes it in my voice. It puts it in there and, and now it's, it's got what I said, but it's, it's now much tidier. And I say, okay, that looks good. Let's change this. Do that. I say, okay, fine. Now give it to my prepare blog post for publication.
And what that one does is that one actually writes, uh, WordPress friendly HTML, nicely formatted, and it also gives me meta tag suggestions, regular, just the, the comma separated tags. It gives me the, um, SEO meta tags. It gives me the, uh, five suggested titles. It, it. me, I, I have, I have a separate one, which gives me 20 compelling opening hooks.
Jonathan Stark: Mm-hmm.
Alastair McDermott: one suggests three or four compelling opening hooks that might be better.
Jonathan Stark: Mm-hmm.
Alastair McDermott: it just, it just gives me all of that back. And I, I, I eventually built this into a, a single web app, which can go end to end doing all of those things. Uh, I don't tend to use that 'cause I'm quicker just using the chat GPT interface to do it.
Jonathan Stark: Yeah.
Alastair McDermott: but uh, like you can, you know, like you can vibe code a, an interface to do all of those things one by one as well if you want to. Um,
Jonathan Stark: But how are you? You said you have one. Give it to the other one. Like what does that actually mean?
Alastair McDermott: uh, so, so, uh, quite often in chat GP that's just typing like, uh, the at sign and, and the the GPT name
Jonathan Stark: Oh, so these are all, so these are all
GPTs that you created.
Alastair McDermott: of 'em are custom GPTs. Yeah.
Jonathan Stark: Okay.
Alastair McDermott: I, I really like the custom GBT interface and how their gpt work. Uh, I don't always like their models the most. Um, I, I mostly like Claude, uh, but I use it the least because, uh, quite often it's just simply not available. As in it, it, it, um, it, it's totally rate limited or, um, it just throws an error. So I, I actually recently stopped using Claude altogether because of that. Uh, I use Gemini 2.5 Pro for 60, 70% of my work, and then I use a. Um, chat, GBT and I have the team's version of that so I can share it with my assistant and, and we can share the internal, um, the internal GBTs across that. And, uh, then the other one that I use sometimes is Google AI Studio. And I, uh, that does train on your data, so you have to be careful what you put into it. Uh, but with that one, I like that you can set a thinking budget in terms of token numbers and you can also set a temperature. So I like to be able to lower the temperature to get more reliable, consistent output on, on some things.
Jonathan Stark: I've never heard anyone say temperature. What's that?
Alastair McDermott: Okay, so temperature is, is a setting for, uh, language models and, uh, it, it basically controls the, um, level of probability or creativity. So, uh, one is usually the default. Um. if you put it up to two, it's gonna go super creative and start hallucinating all the time. Um, if you're working on something where you don't want it to hallucinate, in fact you, you want it to be really accurate but possibly less creative, then you might wanna put it down to 0.5 or 0.3.
Jonathan Stark: Hmm.
Alastair McDermott: I kind of typically run. With, uh, 0.6, 0.7 for most
Jonathan Stark: Hmm.
Alastair McDermott: If I'm brainstorming, I'll probably want it up at one, 1.2, something like that.
Jonathan Stark: Mm-hmm.
Alastair McDermott: just one of these kind of creativity hallucination, uh, to less creative, less hallucination kind of sliders.
Jonathan Stark: Hmm.
Alastair McDermott: but it's not exposed quite often in. Uh, interface. Uh, you will see it if you're using the API you can,
Jonathan Stark: Mm-hmm.
Alastair McDermott: temperature on that. So it's was one of those things that kind of, um, I think, uh, ai, uh, folks behind the scenes know, but but tend to get exposed to the end user much.
Jonathan Stark: Right. Okay.
Ideal Reader and Learning AI
---
Jonathan Stark: So let's zoom out a little bit so that for the ideal reader of this book, who's, I'm gonna say almost definitely not as deep into the technical aspects of it as you are and, uh, probably aren't gonna start off being that complicated. Maybe, maybe the experience that, well, you tell me who's the ideal reader for the book, what's their level of experience with ai?
Alastair McDermott: Um, I, I think that it's somebody who is a subject matter expert who wants to, who is curious enough to have that kind of, um, mindset of wanting to learn more. Uh, like the analogy that I give people is, um, I, I think it's a bit like learning to drive. Most people should probably learn how to drive a car. Um. But if, if you're a subject matter expert, being an expert in using AI is a superpower. It like it is, got this multiple, uh, multiplying effect. And I, I think that, um, you can go, like, you can go to an extent where you're not just, uh, learning how to drive, but like you're learning how to be to drive a race car. You know, and, and so you, it doesn't mean you have to be able to build the engine, but you've, you, you become, uh, like you're able to master the machine more. And, and so it, it's for people who wanna do that.
Choosing Your AI Setup: DIY vs. Subscription
---
Alastair McDermott: Um, and, and like you can take it to any level you want, like you can install VS code and set up the APIs yourself. Or you can pay, you know, 30 or 40 bucks a month to cursor and not have to worry about the technology part. Um, and just use, use the, the, the basic interface, like it depends on which way you wanna go. There's lots of different kind of levels of, um, levels of, of, of technical expertise that you can choose.
Jonathan Stark: Mm-hmm.
Alastair McDermott: think that this as a skill, able to.
The Importance of AI Expertise
---
Alastair McDermott: Understand and, and use and, and know the limitations as well. But being able to use these AI systems, particularly these agent systems where you've got one passing control to another, to another, I think that level, uh, um, of AI usage and knowing how to orchestrate these systems, I think that it's just one of these things that is, um, is, is potentially revolutionary for a person as an individual and for a business if it's a business leader.
Jonathan Stark: Right. Yeah, like maybe like the idea of just learning how to use a spreadsheet, you know, 30 years ago or something like that. It's like, you should really learn how to do that if you're gonna be dealing, if you're gonna be running a business. You know, that's like a very useful skill, like very useful.
And this, this is like, it's almost a joke to compare this to spreadsheet, to AI and spreadsheets together, but I, I, I tend to agree with you, especially in the expertise space.
Um, let me ask you this. If the ais are so good, what do we need experts for anyway?
Alastair McDermott: Because AI make mistakes and ai, it's not that they can make mistakes, which they always say is they definitely will make mistakes and, and like this is the, this is where the expert has the advantage. I mean, people who are junior. Who don't know when it's giving them the right answer or the wrong answer.
They're like, they're at a serious disadvantage now. Um, whereas people who are actually experts in, in their field and, and, and know instinctively when they see something, oh, that's right or that's wrong, um, that, that's, that gives us something to build on. Um. And, uh, so yeah, it does, it makes mistakes, uh, you know, like there's problems and around, um, bias, there's problems around power usage and things like that as well. Absolutely. You know, you gotta think about those things too. but what I'll say is that, um, as, as an expert, your expertise and your judgment is the thing that, that, that gives you the, the, the credibility, um, your relationships, all of those things are things that AI can't do. And, um, you know, it, it can fake empathy.
It can't do real empathy, but unfortunately, neither can a lot of humans.
AI's Role in Business and Personal Productivity
---
Alastair McDermott: Uh, so, so, um, but, uh, you know, I, I do like, there are certain things that AI can, can allow us to do that we simply could not have done otherwise. And, uh, and, and, you know, it, it, if you use it properly, it can make your work much higher quality and, uh, and it can give you those extra capabilities that you simply wouldn't have had before.
Jonathan Stark: Yeah, that's been my experience as well where I've got. You know, hundreds of students and people who contact me and regularly, and a lot of them, you know, it's, it's a, it's a polarizing topic. Some of 'em are super anti ai. Other ones are super pro ai. Most people are kind of in the middle. I've used it for certain things.
They, you know, they use it on their Zoom calls ever. Seems like everybody's doing that now. Uh, it seems really good. Hardly ever comes up with something that. It hardly ever hallucinates or makes things up and you were just on the call so you can read it and be like, oh, this is right. You know? So there's certain things that it's, you know, certain very specific use cases.
It's really good for, I've had a couple, it's happened a couple of times to me where.
Challenges and Limitations of AI
---
Jonathan Stark: Someone would say, Hey, I was asking Chad TVT about something, you know, some posi like how to position my business. Something like that. Something that I spend a lot of time, I've been doing for years. Something that they haven't been spending a lot of time doing, and, and it'll give them this response and they'll say, what do you think?
And I'll be like, and, and I'm reading it and I'm like, it's, this is great. This is like so good. But the person who actually did it can't tell.
Alastair McDermott: Yeah. Yeah.
Jonathan Stark: like, I'm, I'm, there was another one. It was like, how would I come up with a marketing plan for, for a business, here's my website. How would I come up with a marketing plan for this business?
And it was a long marketing plan. There was literally nothing I could think of to improve on it. And there was nothing wrong. There was no hallucination, there was no bad advice. I was like. I was, I was like, I couldn't, I can't improve on this. But the, the takeaway here is that the person got the information, in my opinion, an a plus marketing plan for their business and couldn't even see it.
Alastair McDermott: Yeah.
Jonathan Stark: 'cause they didn't have the, because they're like, they. They're like a, well, I don't wanna say junior and play. They're just not an expert. And it's, and it takes that judgment to be able to say like, oh, this is good or bad. This is, this is insane. This is totally wrong. But especially as a soloist, I find it really useful to.
Have something to pitch ideas to that is a reasonable facsimile of sort of, it, it just allows me to like, think out loud almost, but with my fingers instead of my mouth. You know, it's, it's extremely useful for me for certain types of things, but I'm always bringing that judgment in and saying like, like I can tell the difference between what's probably a good, you know, what has a high probability of success and what is absolutely.
Stupid. And most of the times, you know, most of the times it's giving you middle of the road pretty good. It's usually pretty good. It's usually not horribly bad. It's not usually amazing. So I find it really useful as a soloist. But you can't, you can't, the judgment has to be there, otherwise you're just gonna use it.
And it, you know, like, like imagine if I, if I said, oh, gimme a, gimme a report on warn peace. I never read Warren Peace. It just gives me a report. I wouldn't know if it was right or wrong. It could completely have made something up, you know? So it's just like without the expertise there, you don't know how much weight to give the thing that it gives back to you.
Alastair McDermott: So one of the, one of the issues there, so.
Effective Prompting Techniques
---
Alastair McDermott: So one thing that you can do to inoculate against that is you can crosscheck.
Jonathan Stark: Mm-hmm.
Alastair McDermott: So if you say, Hey, this is a summary, uh, that Claude gave me, Gemini, can you tell me if it, if it looks right of, of war and peace, right? You can do that and, and you're, you're like, you're dramatically increasing.
I'm not saying it's gonna be a hundred percent, but you're gonna be like five nines probably that this is correct. Um, that's one thing that you can do. Um, the other thing is. more that you use it, the more you start to understand when it is and is not going to make a mistake. and I, I think that like the only way is to put the reps in, in terms of using it. so you, you start to a, again, it's this kind of instinctive knowledge thing. It's like as you use it, okay, I probably haven't given it enough background context here.
Jonathan Stark: Mm-hmm.
Alastair McDermott: so it's probably gonna misunderstand something. Maybe I should have worded that better, or maybe I should have written a more structured prompt.
Jonathan Stark: Mm-hmm.
Alastair McDermott: instead of giving it like a very conversational prompt, particularly if you're, if you're, um, doing stuff over and over again, maybe you want to the conversational prompts and actually turn them into structured prompts and maybe add some knowledge documents as context,
Jonathan Stark: Mm-hmm.
Alastair McDermott: it'll do, um, whatever you need to do in a more repeatable way. So, um. That's something that you can do that, that can help with that problem. Um, I'm sorry, then I lost my train of thought.
Jonathan Stark: Well, I've got a question for you based on that though.
Using AI for Coaching and Workshops
---
Jonathan Stark: So, uh, I've had, so like I'll have like a, a really long coaching call like maybe last week I did like a real deep dive with somebody on the productivity systems and it took four hours. We were on the phone for four hours and I got a transcript of it and I wanted to.
I basically did a one person workshop with this person, and I was like, I want to pull out all, I wanna pull out the frameworks. Like what did I, how did I walk through that?
Uh, it made sense to me at the time, but I couldn't, I couldn't like recreate it in my mind, you know, like it would've, you know, so I was like, go through there, pull out the system and see, you know, I just wanted to see if there was, see how interesting it was boiled down and I was two, two things happened that were, uh, unimpressive.
One was, it didn't pull, it like, didn't, it did a terrible job. Like it, it, it was co completely wrong and incomplete. And I said, well, you know, this is wrong and incomplete. Like where, why is this wrong and incomplete? And then it said, oh, well, the transcript was too long. I could only read the first quarter of it.
I'm like, well, that, that would've been useful to know, you know, before I started interacting with it.
And I'm like, how should I have asked you to do it so that you read the entire transcript? And it was like, oh, well I'll do it. You know, I'll reprocess it now and I'll do it like this and go through it.
And then that is super frustrating that it silently fails.
Alastair McDermott: Uh, well, I, I tell you what might be more frustrating is it'll, it'll tell you that it's gonna do it right the next time and then it doesn't.
Jonathan Stark: Yeah. So, okay, let's do, let's have some horror stories here. Yeah. Keep going.
Alastair McDermott: Yeah. Like what this, like, what these machines are doing. Is it's artificial intelligence, it's not real intelligence. And it's doing is it's giving us the most statistically plausible answer to the request that we give it,
Jonathan Stark: Mm-hmm.
Alastair McDermott: what that request is. And if that request is, can you tell me why it's gonna apologize? Because that's statistically what happens, is it apologize, and then it says, oh, I'll try and do it again. Because, and, and, and it doesn't have the mechanism to do it again necessarily. It might, you know, it may have or. But it probably doesn't. So like this is one of the, the super frustrating things. It's baked into the technology, like the, like they are stochastic parrots. Some people call them,
Jonathan Stark: Mm-hmm.
Alastair McDermott: now just because. All it is, is it's a probability token machine. Just because that's all it is doesn't mean it's not useful,
Jonathan Stark: Yeah.
Alastair McDermott: it does depend on knowing how to use it.
Right. You know, um, it, it's, I I I, I, I liken this to somebody saying, you know, this, this drill is terrible. All it does is it drills holes. It's like, well, that's,
Jonathan Stark: That's what it's spelled for.
Alastair McDermott: Yeah. So, so if you need, if you need a hole, then it's great. Um, but yeah, so, so the, there's a couple issues here. So one is. A four hour transcript is gonna be a lot of tokens,
Jonathan Stark: Yeah, it was about 5,000, 50,000 words.
Alastair McDermott: Yeah. So 50,000 words. Yeah. So if you weren't, if you were using Claude, you might just about have got away with that, but probably not. If you are using Gemini 2.5, you might have got away with that and it might have worked in, in Gemini. Do you, do you remember which one it was you were using?
Jonathan Stark: I think it was Chad, TPT.
Alastair McDermott: Yeah. Yeah, so I think that the token limits in chat, GBT won't cover that. If you tried that same thing in Gemini, it'll probably work better. Now.
Advanced AI Usage Tips
---
Alastair McDermott: The other thing I would say is that if you give it a more structured prompt, rather than, you know, just typing in whatever is off the top of your head at the time, if you give it a more structured, prompt and structured instructions, um. will help. So I mean by that is, so we like, I tend to think of like three, three really, um, basic categories of prompts. One is a simple one, like, you know, summarize this text or translate this to French. That's a really simple one. Then there's conversational, which is, uh, and I, I. You do conversational all the time in dictation mode. What I like about conversational is you can ramble away for two or three minutes and give it loads of context, but that is, Hey, chat gt. I was just on a coaching call and I'd really love for you to go through and pull out all the key insights. Oh yeah. And also can you see if there's any frameworks and I, I would literally say it like that.
Jonathan Stark: I do. That's how I say it. Yeah.
Alastair McDermott: Yeah. Um, now that's good. But if you are doing this in a, in a repeatable way, like let's say, let's say you want to transcribe every single coaching call that you have, and always have it give the same format, and then maybe it feeds it into an email, uh, system where it emails automatically to the client with your, with your generated notes. You might want to have a, a, a more thought go into that. Now, one of the great things is we can actually use AI to structure our prompts better.
Jonathan Stark: Mm-hmm.
Alastair McDermott: I, I've a free, I've, I've about 70, 80 free tools, by the way, in the GPT store. If you look up, uh, human spark. Um, but one of those is a prompt creator and you can say, Hey, here's my prompt.
Can you help me to make this better?
Jonathan Stark: Hmm.
Alastair McDermott: will just take your prompt and, and restructure it and reflow it a bit to, to work a bit better. Now, one, one of the things that we. Learned recently from, um, a research study from, I think it was early this year, was that, uh, there seems to be two types of, of ways that the, um, that the neural network, uh, large language models work. And one of those is based on instructions, and the other is based on, uh, what they call demonstrations in the paper. Basically means examples. So if you give it not just instructions, but also examples, you are like almost doubling the amount of brainpower it's able to give to your task. So if you
Jonathan Stark: Okay.
Alastair McDermott: it instructions as well as examples that will really help, um, and that, that will, that will, you'll get better results out of it.
There's lots and lots of little tweaks like that, that you can do, uh, to help you get better results. Um, in, in this case, I think the biggest thing is the context window, which is the amount of, um, the amount of, of memory wor viable memory that it has. You probably would be best off in, uh, Gemini for that task, just for that.
Jonathan Stark: Well, so before when you were talking about, you know, giving these different, you know, like the operator, I think you said the, the editor, GPT, you give it like a lot of context and context is really good, but doesn't it send all that context every single time that you are. Like, how do you, how do you do that in a way that doesn't, uh, overwhelm, what did you call it?
The token window or whatever it's called, like it's memory.
Alastair McDermott: okay, so there's, there's a few different things. So like, I would start a new chat for a new task. If it's not related to something previously, then I would start a new chat.
Jonathan Stark: Mm-hmm.
Alastair McDermott: the other thing that is really useful to do, some people call it, um, compressing context, which is where it, uh, it takes the context and it, it, it basically summarizes it and starts a new chat based on that. Um, I have a, I have a, a prompt. Like it's a prompt template, uh, which I call project sync.
Jonathan Stark: Hmm.
Alastair McDermott: it does is it actually does it in a, in, um, it, it takes that context and it summarizes it in a particular way. I'll tell you what, I'll, I'll, I'll just give you a, a quick flavor of what it does 'cause it might be useful.
Jonathan Stark: Mm-hmm.
Alastair McDermott: It says, uh, generate a comprehensive project or chat summary that can be easily referenced in a new project, um, that clearly explains all aspects of this project and conversation to my business partner, who was also a software engineer. The document should be enough to thoroughly bring, bring them up to speed and enable them to contribute effectively. Now there's lots more, but. But I just wanna give you that, that kind of
Jonathan Stark: Yep.
Alastair McDermott: uh, I don't have a business partner who's also a software engineer. It's just that I wanted to capture the business aspects and the, and the engineering aspects. 'cause
Jonathan Stark: Right.
Alastair McDermott: aspects of, of the conversation that I, I want to capture.
Jonathan Stark: That's, this is really smart. Yeah. And then you take that and you start a new chat with that.
Okay.
Alastair McDermott: usually has enough context at that point.
Jonathan Stark: That's really good. That's really good. 'cause I find myself, I'll go, I'll ask, I'll originally start by, I'll put in like this, you know, like, oh, I hit a kickoff call with a new coaching student. Like, pull out all of the, the, these things, this list of things like what, what are potential positioning statements for this person?
What's their product ladder now? Or what could it be? What to dos, uh, and for whom, you know, what are the next steps? And then format it in this way.
Then it'll say something like, it'll be like, oh, do you want me to make a marketing plan for the person? Something like that. I'll be like, sure. And then, and then like the thing that I really, all I really wanted to do was take the transcript and like have this summary document, like a detailed specific kind of summary document.
Alastair McDermott: Yeah.
Jonathan Stark: And then I'll go down this rabbit hole. And, and I kind of do want to stay there because it has the context from the phone call. And then, uh, but then I'm like, uh, now I'm kind of off in a different direction and, and I'm like down a rabbit hole and,
Practical AI Tools and Resources
---
Alastair McDermott: is one, by the way, where Google AI Studio is really useful. and the, the, the issue here is, uh, Google AI Studio trains on your data. See, can't put in anything confidential
Jonathan Stark: mm-hmm. Yep.
Alastair McDermott: it does give you the ability to branch from a conversation. also gives you the ability to delete previous context.
Jonathan Stark: Oh, that'd be so good. Yeah.
Alastair McDermott: it's really nice for that and that's why I use it for some stuff where I don't care about the confidentiality stuff and it's, find it really useful for that. Um, like I put in my white papers, I've, I've got a bunch of white papers on, on, um, ai, um, adoption, and so I put those in and I use those to, to kind of what, like when I'm working with corporate clients, uh, those are, it's really useful for me to have conversations with those. And then I might just delete that or branch it out to, into a new chat. So, um, yeah, that, that's really useful. Um, I put the project sync into the chat here in, in the system that we're using to chat. By the way, John.
Jonathan Stark: Okay.
Oh, I see.
Alastair McDermott: it again. Um, send it, send it to people. Um. I've, I've got some
Jonathan Stark: Cool.
Alastair McDermott: like one that I always like to say is, is there anything surprising or insightful you can tell me about approaching this problem,
Jonathan Stark: Right.
Alastair McDermott: not have thought of already?
That's a really nice one. another one I like to use is please ask me a series of focus questions about this topic, one at a time to gather the information we need
Jonathan Stark: Yes.
Alastair McDermott: And,
Jonathan Stark: Right.
Alastair McDermott: where Helpful. Offer a numbered list of realistic answer options for quicker responses. Otherwise allow for open-ended replies.
So what it'll do is say, okay, we could go forward. Here's three options. You like any of these, or, tell me what your option is. And you could just go through this numbered menu
Jonathan Stark: Yep.
Alastair McDermott: this. So those are, are like really simple things, but they'll just get you in. Um, like, uh, they'll, they'll get you moving on something.
Jonathan Stark: Mm-hmm.
Alastair McDermott: a, a series of questions like I did. Um, I did really, really solid positioning statement for my business, uh, about a year and a half ago now. Um, I did it on a live stream in 15 minutes. said, I want you to ask me a, a series of questions about my business and my positioning in order to create a really good positioning statement in this format.
Jonathan Stark: Mm-hmm.
Alastair McDermott: solve this problem by doing this so they can do that.
Jonathan Stark: Mm-hmm.
Alastair McDermott: just gave it that. And when, now, here's the thing. I knew the positioning statement format that I wanted to use, and this is
Jonathan Stark: Yeah.
Alastair McDermott: expertise comes in.
Jonathan Stark: Yeah.
Alastair McDermott: know that, then that's where the, the, um, that's where the expert has the advantage here. Because you know how to ask the right questions. You know, what looks right, what looks wrong. Um, yeah. So there's a bunch of ones like that that I find really useful. Um, now I have a very simple way of, uh, of, of documenting my, my prompts, the ones that I use all the time. I, I turn into a custom GPT, which is basically just, um, uh, uh, a structured prompt with a name and some knowledge documents, like some, uh, like some text documents or, or something like that attached to it.
So it has some context.
Jonathan Stark: And that's in you? That is in Chatt PT specifically.
Alastair McDermott: Yeah, I use those in chatt. Bt you can also do it in Claude as projects, and you can do it in, uh, you can do it in Google Gemini as gems Chatt. BT has the best kind of interface for using those and, and they seem to be the smartest about using the context.
Jonathan Stark: Hmm. And they're local, you said?
Alastair McDermott: yeah, and then I, I just have a simple, uh, shared spreadsheet that I have, um, for myself and my assistant and, and, uh, so, um, we can both copy paste.
So that's, that's where I copy pasted this one, uh, for you out of,
Jonathan Stark: Mm.
Alastair McDermott: Um, so I just have, it's just called, uh, my AI prompts and I just click on that. I can go find something so I have a bunch more. I'm not gonna give you them all, but I have a bunch more in there. Um. One of those, uh, I I like to use is say, I want you to put on the persona of, of our ideal target client. Um, and I want you to picture the, the project that we're doing. Would this person say, hell yes or meh or you about this and whatever it is that we're working on. And so, you know that, um, uh, sorry, I've forgotten the name of the guy now. Uh, CD baby. Isn't that who, who wrote
Jonathan Stark: Oh, Derek Sives. Yeah.
Alastair McDermott: Yeah, I think it was, it was hell, yes.
Uh, or no, I think he, he basically talked about that. Yeah. So, so like, this is one of the things I would use for quality control of the book. I, I created a tool, by the way, on chat, chat G bt, which is a free, if you look up hell yes. Uh, on chat GBT you'll find it. Uh, but basically you can put in any, any, like who's, who's the target audience, what would that target audience persona think of this. And, and so I run a chapter of the book by it, or a subtitle or a title or a design and say, would they like this? It'll say, hell yeah. Or it'll say, no, they wouldn't like this. Here's why.
Jonathan Stark: Hmm.
Alastair McDermott: be, that can be really helpful as well.
Jonathan Stark: Wild.
Alastair McDermott: I can mono here. I get very passionate about this stuff.
Jonathan Stark: Well, I'm looking at the clock. I know we have to wrap up.
Final Thoughts and Resources
---
Jonathan Stark: So, um, where can people go to find out more about what you're doing? It sounds like you have a ton of tools on the website or built into Chet BT. So where, where's the best place for somebody to perhaps take advantage of all of this time that you've put in making yourself an expert in this area?
Alastair McDermott: Uh, go to Human spark.ai and look up my name Alistair McDermott on LinkedIn. I'm very active on LinkedIn. Love to connect with anybody from your audience.
Jonathan Stark: Cool. I mean, I, I, you gave me a bunch of great tips and, and like, duh, I'm just kind of smack myself in the head. I have a keyboard shortcut thing that I use for like tons and tons of things and I don't have any prompts in there,
which is stupid 'cause I have tons of prompts that I use all the time. So
just immediately after this I'm thrown in a section for AI prompts and they'll be probably pull, pull a dozen that I use on a regular basis.
Alastair McDermott: I'm gonna give you one more, Jonathan,
Jonathan Stark: Yes, yes.
Alastair McDermott: um, when you're doing this or anything, uh, talk to Chachi PT and say, I want you to create a hyper-efficient workflow that uses the 80 20 rule to do this thing that I wanna do, and it will help you shortcut all the parts that you don't even need to do.
Jonathan Stark: Nice. That's right up my alley.
Alastair McDermott: Yeah. Cool.
Jonathan Stark: Cool. Yeah. Well, thanks so much for coming on, Alistair.
Alastair McDermott: Yeah, it's, it's been a pleasure.
Cheers.
Jonathan Stark: Alright folks, that's it for this week. I'm Jonathan Stark and I hope you join me again next time for Ditching Hourly. Bye.
