Martin Moore (1:58)

Them into rogue prompt cowboys, and how.

Martin Moore (2:00)

To use AI to dramatically increase your speed and impact without handing your brain.

Martin Moore (2:05)

Over to the machine.

Martin Moore (2:07)

If you're a leader who wants to use AI to amplify your edge rather.

Martin Moore (2:10)

Than erase it, this episode is for you. So let's get into it. James Killick, welcome to no Bullshit Leadership, mate. It's great to have you here.

James Killick (2:20)

Thanks. Thank you so much for having me.

Martin Moore (2:22)

And it's very rare that I actually bring someone else into the podcast, but when I need an expert in a particular area, I'll sometimes invite them on. So I haven't had an expert for a while, and I believe that you've been working with AI for almost 10 years.

James Killick (2:34)

That's right.

Martin Moore (2:35)

Can you just tell me a little bit about how you've been working with it.

James Killick (2:38)

Yeah, sure. So originally it was through products. So we've helped people launch apps, web apps, we've even done a bit of crypto. And throughout that time I've helped people pop AI into their apps. Essentially, the first one was actually facial recognition, helping ensure that the right cleaners went on to the right jobs. I've also done things.

Martin Moore (3:00)

So the right cleaners on the right jobs, you're talking about like biometric security?

James Killick (3:04)

Yes, exactly. Fantastic cleaning company. And they had some high profile clients and they had to do a certain level of compliance to be able to clean those sites. So they had to ensure the right cleaners that were, that had the right identification were going on to the sites.

Martin Moore (3:19)

Right, fantastic. That's pretty cool. There was a time when I remember we were sweeping our boardroom for bugs because we thought we were so important someone might try and record our conversations at a particular company. I won't say which one it was. So I get that, that's fantastic. And look, as I look at AI now, I see that there's been so much movement and so much hype, I feel as though I've got a reasonable handle on the large language model type of AI, you know, ChatGPT, Claude, you know, co pilot and so forth. But I'm really interested in how powerful it's going to become as well with agentic AI. So when I think about AI as it stands today, it's a fantastic tool. I use it all the time, but I don't use it for anything I really need to rely upon. So I still feel as though there's this hole in AI that is, they call it hallucinations, but where I can quite often be wrong. Even though it has all this information at its disposal. Do you just want to talk a little bit about how reliable AI is now and whether that's going to improve and at what speed?

James Killick (4:23)

Yeah, I think the, the really critical aspect of this is you don't know when it's going to hallucinate. Right. So, you know, you could be spending hours on AI getting it perfect, perfect, perfect, and then the next minute it hallucinates. So if you're not on the ball, if you're not keeping a human in the loop, it is actually scary what it can do. Story for you. Quick one, please.

Martin Moore (4:42)

Yeah, yeah, go for it.

James Killick (4:43)

So I did a bit of research on you today.

Martin Moore (4:45)

Right.

James Killick (4:45)

So I've got this thing, this setup with Claude. I use something called an MCP LinkedIn to a LinkedIn API and that allows me to check your profile. It also does a web search on you and it found. Am I right in thinking that you once did law?

Martin Moore (5:00)

I did.

James Killick (5:01)

And then dropped out?

Martin Moore (5:02)

Failed law student. Absolutely. From the 1980s.

James Killick (5:04)

Okay, cool. So it gave me awesome research on you, but then at the end it said you've walked the same path. You are also a law dropout. And that's just a complete lie.

Martin Moore (5:15)

Right.

James Killick (5:15)

Okay, I never dropped out.

Martin Moore (5:16)

Well, it lies all the time.

James Killick (5:17)

Yeah, it does.

Martin Moore (5:18)

And. And there's times where I'll ask something where I sort of half know the answer.

James Killick (5:22)

Yeah.

Martin Moore (5:22)

So for example, the other day I'm watching an American football game and I say, oh, this is driving me crazy. So I get onto ChatGPT and I say, chad, because my daughter Olivia told me this is. You call it Chad and it likes you. It's more personal. Chad, please tell me, mate, how many fumbles have the Patriots had in the first half? And he came back and said, the Patriots have had three fumbles. And I said to it, well, hang on a minute, I'm pretty sure I've seen at least four. And it comes back and says, you're right, there were four fumbles. And I go, well, which is it? Is it three or is it four? Are you just trying to make me happy? Cause sometimes I feel as though my AI is just trying to make me happy and please them. Is that true? Yeah. Why does that work? It's not a person. It's not actually Chad. So why does that actually happen?

James Killick (6:07)

I think it's just with the programming that's gone into these lands, right? They're, they're language based, so large language model, and they're pulling in all of this information and there's lots of, lots of contrary information. So, you know, it could have been that there were stats out there from a few minutes ago when there was three. Now there's some stats out there where there's four.

Martin Moore (6:27)

Right.

James Killick (6:27)

And so it doesn't really know what to pull from. And then when you say no, surely that's not right, it's just going to agree with you.

Martin Moore (6:34)

Yes. Okay. Okay, fair enough. That's super interesting. Right? So as I see it, and the way I have in my head, this mental picture of AI, I think is important because the way I see it at the moment, I don't think it's replacing many jobs right now. And let me just explain. So I used to use Google, of course, my searches, I type in a string and it would come back and Give me a list of websites that satisfied that search criteria, ranked by popularity, frequency, sometimes sponsorships, whatever else. But it basically would just point me to a website with ChatGPT, the LLMs. It actually goes and crawls through the Internet through all of the data and extracts data from a range of different sources and presents it in a coherent summary. That's how I see it. Is that right? Is that if I got my head right around that.

James Killick (7:25)

Oh, it's an interesting perspective. I mean, I see AI is like an overeager intern.

Martin Moore (7:31)

Right.

James Killick (7:31)

So the problem is that it does go deep. Definitely. Right, right. So when you ask it to search, it's going to not just pull up the homepage as it's going to go deep, it's going to go through the linkage. But you can, you can spoof AI, you can put up lots of different websites, you can try and put yourself forward as the authority and you can trick AI at the moment. So.

Martin Moore (7:51)

Right.

James Killick (7:52)

Geo generative search is actually a little bit, a little bit dodgy in my opinion. Right, right. People know how to black hat it.

Martin Moore (7:58)

Right.

James Killick (8:00)

I think as well.

Martin Moore (8:01)

Well, sorry, just, just to be clear, when you say people know how to black hat it, they can game the system.

James Killick (8:05)

Yes.

Martin Moore (8:05)

So they get preferential presentation of what they're putting out there on the Internet as opposed to the other information that might be available.

James Killick (8:11)

Exactly.

Martin Moore (8:11)

Right. Okay.

James Killick (8:12)

In a lot of ways I actually still think Google search is better because the ranking that's used within the algorithm for SEO is taking a lot of different things into consideration and it's grown maturity over lots and lots of years.

Martin Moore (8:26)

Right.

James Killick (8:27)

Geo agentic search is very, very new.

Martin Moore (8:31)

Right. Okay. Well it's sort of interesting because quite often on a Google search I'll go to a site and I'll say, yeah, that's not exactly what I'm after. So I go back and I go to number two in the search list.

James Killick (8:40)

Yeah.

Martin Moore (8:41)

So it takes me a lot more time. Whereas I find that when I'm prompted by AI, if I prompt it well and it comes back with an answer, I can drill deeper then and say, actually I want to drill down on this piece here. So this part of the answer. Yeah. Which one do you think is more effective and more likely to get me.

James Killick (8:56)

The right result if you're prompting. Well, AI for sure.

Martin Moore (9:00)

Okay.

James Killick (9:01)

I was going to say if you're using the right context, you're telling it, you know, the kind of output that you're looking for and maybe even giving it some examples of companies or sites that you like then you're going to get a much better search result than Google. But the average person uses AI like they use Google and in that sense they're better off in Google.

Martin Moore (9:18)

Okay, right, okay, that's, that's good. Okay. You've given me a bit of comfort there, James. I think I'm sort of on the right track. So. So there's been a lot of hype about companies being able to replace thousands of employees with AI now. And there's also been some research as quite often tends to come out. I'm not sure how self serving the research is, but it basically says that the promise of AI so far for the companies who are reducing jobs hasn't paid off. No one's actually having these AI projects that are wildly successful. Do you have a view on that?

James Killick (9:53)

Yeah, the stat is crazy actually. I think last year it was 80% of AI initiatives fell. I think this year it's probably going to be 90%.

Martin Moore (10:01)

Right.

James Killick (10:03)

The problem is that people are not taking a business mindset, a business first mindset. So they are having, getting the solution and saying, how can I put this into my business? Rather than looking for the problem.

Martin Moore (10:15)

Right.

James Killick (10:15)

And that's why they're mostly failing. You know, they're looking at influencers that are saying, clone your voice. Why? You know, so you can put it on social media. But I don't do anything on social media. You know, realistically I should be amplifying what you already do. Well, Right. Or solving a bottleneck. It shouldn't be pulling you in a completely different direction.

Martin Moore (10:35)

No, yeah, I get that. That's really good. So with the use of AI for answering any question, it's okay. But when it comes to using to specifically address a business need, how much work do I have to do to make the existing AI tools perform at a specific business function level? Now, I know you've done some work for our business in terms of, you know, writing code in AIs to get them to do a certain thing for us in terms of our marketing.

James Killick (11:05)

Yeah.

Martin Moore (11:06)

And this has been really valuable to us and we know how much productivity it gives us and how much further cut through and reach it gives us in our marketing efforts. But how much work's involved in that and why are companies getting it so wrong?

James Killick (11:17)

First off, just on search. And so I tick everything off.

Martin Moore (11:21)

You're very thorough and methodical. I like that, James. Thank you.

James Killick (11:25)

I just don't want to bullshit your listeners.

Martin Moore (11:27)

I would not let you go for it.

James Killick (11:29)

So with search, there's two ways you can do it. Right. So if you put something into ChatGPT without enabling the web search, it's going to go through its database.

Martin Moore (11:37)

Right.

James Killick (11:38)

That's like, you know, whatever it used to be 16 years, it's more now worth of text, books, literature.

Martin Moore (11:44)

Right.

James Killick (11:45)

So there's a high likelihood there's bad text in there.

Martin Moore (11:48)

Sure.

James Killick (11:48)

And so you need to, need to be careful of the answer it gives, especially if you want something that's up to date. Right. So if you enable that web search, or in my opinion, use Perplexity, I think Perplexity is the best AI search tool.

Martin Moore (12:00)

Right.

James Killick (12:00)

And you say, I want you just to search documents or websites from 2025, you're going to get much better search results.

Martin Moore (12:07)

Okay, great.

James Killick (12:08)

So that's the first thing. So then in terms of actually deploying that AI for a function, and that really depends on the function. So I find that marketing and sales are much easy early win than going into admin and operations.

Martin Moore (12:28)

Right.

James Killick (12:28)

That's because, you know, sales and marketing is, is more cookie cutter.

Martin Moore (12:32)

All right. I thought you were going to say fluffy bullshit, but I'm glad you didn't. It's going. Sorry, Tash, sorry. It's. It's cookie cutter. I got it. I'm with you. So it's easy to template.

James Killick (12:41)

Exactly.

Martin Moore (12:42)

Good, good.

James Killick (12:43)

You know, the good things that are working are often not too dissimilar from company to company. But you know, there's such a wide range of different operational activities that companies do from different industries and just even within the same industry.

Martin Moore (12:55)

Right, right. Okay. So. So I'm a leader in a company. Right. Let's say I'm in a mid level role somewhere and I've got executives above me who are being fed the dream of how good AI is going to be and how it's going to absolutely revolutionize what they're doing. So they want to get a head start on it.

James Killick (13:13)

Yeah.

Martin Moore (13:13)

So they deploy one or two AI tools and they say, all right, we're going to cut a thousand staff, some of whom are the middle managers. Yeah. But if I'm trying to manage that and I have the reality of operationally delivering what I need to to deliver on our customer promise, and I'm being told to do it with less people because of this wonderful AI tool, which isn't actually going to help me in the first instance, how should a leader look at that?

James Killick (13:38)

So I think you have to take the mindset that I shouldn't be replacing humans. What you want to be doing is enabling humans to use AI to become more efficient. I don't like the directive of get rid of your staff. Why not make them more efficient so they do more service delivery, bringing in more business, so on. So what I would do is I would test one or two very small things that can free up your best staff to do the best part of their job, the most effective part of their job.

Martin Moore (14:05)

Right. That's great, James, because that really coincides with my philosophy and experience around using pilot programs or prototypes to test an idea before you fully implement it and testing it with the people who are most likely to be able to get it done the right way. So that's really consistent with that. And look, technology's been encroaching for years on, on manual labor, whether it's, you know, accounts payable and receivable and getting those done automatically and just handling the exceptions. So these huge rooms of people doing, you know, accounts receivable, they're sort of gone, which is great. And you just, 99% of it is just processed as a very automatic thing using the technology we have. And that's been going on for decades. So AI, I suspect, is going to do the same thing. But what is the magic that's left outside of AI, particularly when agentic AI becomes self fulfilling and you know, in its own learning loop?

James Killick (15:00)

So I think it's going to be, you know, making yourself out to be different. Right. So if everyone's using the same models, they're producing the same content as an example, you know, the human nature, the, the human effect is going to be so important. How can you make yourself see, seem to be different to what the rest of the crowd are doing? How can you zig while everyone else is zagging? And the human in the loop is so critically important? You know, with COVID everyone kind of went inwards. We're all working from home, you know, no one's coming into the office. There's a lot less face to face meetings, I think, with AI automating things in the background, which is inevitable. We're going to need that human interaction, you know, that shake of the hand, the look in the eye. We're going to need the human to come up with that creativity that I can't because I looks back at what's been produced in the past and it takes the best of it. It brings it together, you know, it right now. It cannot create new. That's what humans are for.

Martin Moore (16:00)

How far away are we from that though, from AI being able to create new things based on the past?

James Killick (16:06)

I don't Think anyone truly knows.

Martin Moore (16:08)

Right.

James Killick (16:08)

I think there's been a lot of tests and every now and then, you know, something new is created. Right. We don't say never, but I mean, I think what was it that if you give enough monkeys a typewriter, eventually they'll create Shakespeare. So no one really knows when that's going to be consistent.

Martin Moore (16:25)

Right. So. So just from listening to a whole bunch of podcasts on this, as I've done, there was one really, really good one with a panel of guests that Stephen Bartlett did on Diary of a Sea a little while back that was excellent. But he had a guy called Brett Weinstein, I don't know if you've heard of him, sort of futurist, but complex theorist more than anything else. Complex systems theorist. And his view was this is catastrophic and it's coming soon to your neighbourhood. And I think that panel all agreed that it was closer than anyone thinks. Now they were sort of talking about two year timeframes. I don't know if we're talking about two years or five years, but I suspect that, or even ten. But I suspect sometime it's going to radically change the way our, our environment works in terms of economics and employment and everything else. So do you see the shift when it finally comes as being that dramatic?

James Killick (17:16)

Yes.

Martin Moore (17:17)

Right.

James Killick (17:18)

Short answer.

Martin Moore (17:19)

Yeah. Yeah.

James Killick (17:20)

There's a great document, I think it's called 2026 and that, that shows a version of like the Skynet, you know, we're all done. I know, but you know, it's. I feel as though we've had this rapid expansion with AI and what they're doing is they're projecting that same expansion on that same level. But the truth is that we've hit a bit of a ceiling at the moment with this ability to create. So AI is now amazing at taking information and spitting it out in another way. But it's not got that creation. It's not been able to create its own AI or at least what we're seeing, you know, we don't know what's going on in the background.

Martin Moore (18:04)

Sure.

James Killick (18:04)

When it breaks through that. Yeah. I mean, there's a part of me that's scared, I'll be completely honest.

Martin Moore (18:10)

Right.

James Killick (18:11)

I don't know. Like if the people that are pushing this forward are the corporates that are doing it for profits or they're trying to, you know, the countries that are trying to win the AI arms race, they're going to do it at all costs. So what are they training the AI to be is pretty scary.

Martin Moore (18:28)

Sure.

James Killick (18:29)

There's not enough empathy. There's not enough human nature. There's not enough of the good of humans. If you look back our history, the, the language models are taking all of this information, our wars, all of the horrible things that we've done. And that's its training. We really need to step up as, as people. People to train it better, to give it the best of us, to give it what we want it to be.

Martin Moore (18:54)

Right. And that's, I mean, I love the way you said that, James. The thing that really worries me about that is that someone has to be the arbiter of what is desirable. And with the. Even if you just look at the US Very simply forget, you know, Russia, China, you know, India, you know, African countries, if you, if you look at that, they are vastly separated internally about what the right thing would be to get the AI and so that's, you know, someone's got to have a set of rules and someone's got to be the arbiter of what good is. So, so, yes, I share your fear. There's no doubt about it. It's incredible. So, and I think the governments are always very slow to respond. But even in the course of, of, you know, industrial progress, as we've seen over the last 100, 200 years, they've been very slow to respond to changes in markets there. So I can't imagine how far behind they're going to be on things like regulating AI. Yeah. Are you seeing any sort of attempts at the moment for governments to regulate sensibly, to stop the worst instincts of those who have the sort of keys to the castle on AI to be.

James Killick (20:05)

Honest, I can't speak. Speak much on that topic. My gut feeling is we're already too late on that.

Martin Moore (20:11)

Right.

James Killick (20:11)

You know, we've gone far past it.

Martin Moore (20:13)

Right.

James Killick (20:14)

So anything they do now is going to be reactive. But yeah, I don't go deep into the politics of.

Martin Moore (20:20)

Right, okay, sure. So Sam, Sam Altman is probably one of the most powerful people in the world. There's a few others in Microsoft and, and a few other places that are really hitting this hard.

James Killick (20:29)

Yes.

Martin Moore (20:30)

And they will determine our futures, unfortunately, based. Based on their commercial drivers.

James Killick (20:35)

Exactly.

Martin Moore (20:36)

Okay, well, that's sort of good to know. Right. So before I get thoroughly depressed, let's get back to leadership. So I'm a, I'm a senior or middle manager in a company. I know that AI is coming. I know that it's going to move rapidly, and I'm trying to make strategic decisions that are sort of two, three years out at least. And so are there things that I can bank on and say, this is 80% chance of happening. So to cater for that, I'm going to need to start thinking about shaping my workforce in a different way. I need a different type of person, or I need to put the people I have, both now and in the future, onto different types of work. Because I know that there's going to be a tool that's going to handle most of the grunt work.

James Killick (21:18)

Yep. I've got this concept that I've been teaching, which is AI orchestration. Essentially, you need to be the orchestrator of not just the AI, but the people using the AI. So what does that mean? I think more than anything, it's basic training. What can you do with AI? Because there's no point in putting an initiative together and then giving it to your employees and then having no understanding of what that means.

Martin Moore (21:44)

Right. So although we've been doing that for decades, mate, it's not new. No, I'm just kidding. But in the context of AI, it's going to be different.

James Killick (21:53)

Yeah. Well, the people that win are going to do this, right? They're going to keep their humans in the loop, they're going to get them excited about the initiatives. And I think it was Stephen Bartlett that actually did it within his business, where he did a competition and he said, you know, come up with these great ideas. He had so many ideas that were, you know, groundbreaking for his business, and some of them got chosen. And those people that, that had the ideas, they felt as though they were part of the journey. And so that's massively important. And then, you know, if you're a leader, what do you want to do? Okay. You need to orchestrate your team so that each one knows how to use a little bit of AI within their role.

Martin Moore (22:33)

Right.

James Killick (22:33)

That's going to help them.

Martin Moore (22:34)

Right.

James Killick (22:35)

I think as a leader, the worst thing you can do is leave it up to them, because you'll go for Claude, somebody else will go for Gemini, somebody will be using free chat, GPT and giving away company secrets.

Martin Moore (22:46)

Sure.

James Killick (22:46)

You've got to be the facilitator.

Martin Moore (22:48)

Right.

James Killick (22:48)

But this is leadership in general.

Martin Moore (22:50)

Right, sure, sure. But. But the, the exposure now is so much greater because, as you said, James, you know, you'll have situations where someone doesn't understand the security implications of having data out there in the public domain.

James Killick (23:02)

Yeah.

Martin Moore (23:03)

And much, much easier to breach that. Yeah. And I remember, you know, sort of 30 years ago, working in a company where they had diskless workstations because they didn't want anyone being able to put in in those days, three and a half inch floppy disk or a flash drive and taking data outside of the organisation because they were very security conscious and these days it's so easy to let data leak.

James Killick (23:27)

Yep.

Martin Moore (23:28)

So is there any organisation wide remedy for that in terms of saying as a company we can find a way to lock this down?

James Killick (23:36)

I think the easiest way as a business you're either using Google Workspace or Office 365. So they have their own inbuilt AI, they have the same ISO standards as their email systems and their drive folder systems, cloud storage. So the easiest thing to do if you're unsure is just to use those systems that come with your business licenses.

Martin Moore (24:03)

Right, okay. Which is sort of interesting because if I'm one of those employees and I want to check something that I don't want the company to know about that I'm checking, the last thing I'm going to do is use the company's Gemini or Copilot. The last thing I'm going to do. So I'm going to go on my own chatgpt and say blah, blah, blah, answer the question. So, but that's just a human behavior thing which we're, which we're not going to solve here today, James.

James Killick (24:25)

But they can't track you.

Martin Moore (24:27)

No, that's what I mean. Well, sorry, as an even, even if.

James Killick (24:29)

You use your company's Gemini, the company people could not track that.

Martin Moore (24:34)

Sure. But that's, but my first instinct is that's the company environment. If I want to, if I want to do something that's not under the scrutiny of the corporate environment, I'm going to do it here. And when they say we can't track you, everyone goes sure you can't. Sure you can't. All right, so this is, this is super interesting. Right. So, so getting back to how we use this in companies and how we get people trained in it, I feel as though I'm good at prompting the GPT. Like I feel as though I'm good at asking a very specific question that I thought through and articulated. When it comes back, I feel as. I'm pretty good at looking through it and saying that doesn't sort of sound right. Give me a reference from that, tell me where you found that, tell me exactly where that's come from and you know, what are the assumptions around that, do you have any conflicting reports or studies, etc. Etc. So I feel as though I'm pretty good at prompting it, but I still Feel a little bit exposed in terms of not being able to see a hallucination?

James Killick (25:36)

Yeah.

Martin Moore (25:36)

Do you think that's a reasonable approach to take to have that level of confidence that I'm getting good data back but that I still have a little bit of skepticism around the information I'm fed?

James Killick (25:47)

Yeah. 100. You don't know what you don't know. So if you're asking ChatGPT or another model to teach you something, you know, you, you don't know if it's teaching you the right thing. There are some tricks so you can pin one LLM against the other. So you can say Chat GBT gave me this. Claude, is it right? Brutally. Yeah. I always use word words like that. In fact, unfortunately, I think this is again the way that's been trained. The more negative you are, the more accurate it is.

Martin Moore (26:15)

Oh good.

James Killick (26:16)

If you say I'll give you a million dollars, Chat GPT if you give me the right answer, or you say I'll document a million dollars if you give me the right answer, it's more likely to do on the negative one to give you a better answer.

Martin Moore (26:27)

Right. That's interesting. I didn't know cared that much about the million dollars. That's good. There is, there is life in these things. I gotta be really worried that that is super interesting. And so a healthy level of skepticism, which you should have anyway, is a leaf.

James Killick (26:41)

Yeah.

Martin Moore (26:41)

And you know, not always believing what you read, not just subscribing to conventional wisdom. That's a healthy approach to take to AI.

James Killick (26:49)

Yeah. In fact this is my whole methodology. So I really feel as though people with domain experience will be the ones that win. So go really niche with AI and use it to amplify what you already know and what you already do.

Martin Moore (27:03)

Right.

James Killick (27:04)

Rather than. So if I know marketing, I should be doing marketing with Chat GPT.

Martin Moore (27:09)

Right.

James Killick (27:09)

I shouldn't be doing law and the lawyers shouldn't be doing marketing. Why? Because ChatGPT will hallucinate. It will also give you average to bad strategies because it's pulling every single strategy that's out there. You as a marketer know what's good. You as a mark, you as a lawyer know that which cases actually exist and which ones don't.

Martin Moore (27:31)

Right.

James Killick (27:31)

So it's your domain experience experience that makes you powerful.

Martin Moore (27:34)

Yep.

James Killick (27:34)

And so you take that and you use AI to specialize, not be the generalist. That's what everyone's doing at the moment.

Martin Moore (27:40)

Yes.

James Killick (27:41)

There it drives me crazy. It really does. Like business owners are not talking to specialists when it comes to AI, they're thinking, oh, AI knows this, so I can do it all myself.

Martin Moore (27:53)

Right.

James Killick (27:53)

If you want to put a successful AI program into your business, the first thing you should do is, is talk to an AI specialist that has done something specific for your industry or your domain knowledge. Not even a generalist AI.

Martin Moore (28:06)

Right, Right. Yeah, there's a. I think there's a Deloitte partner in Canberra wishes that they'd done that. So that's, that's. That's a good idea. Good, good advice, good advice.

James Killick (28:16)

So just coming back to what we were discussing earlier. I think the point about AI not replacing humans, but instead humans using AI replacing those that don't is such an incredibly important one because there is a window of opportunity and it's now. So if you think about expectations of delivery. Right, let's say that a random task might take seven days, and that's without AI. So that's the expectation. If you use AI, you might be able to cut that in half, but the person you're giving it to doesn't necessarily know that that's the case. So you can impress. This is the way that you can move up the ranks, or if you're a business owner, the way that you can impress your clients to get more referrals. But expectation is going to shift eventually.

Martin Moore (29:02)

Sure, sure. So when it comes to this equalization concept, yeah. I feel as though AI has been a great leveler. Two years ago, someone like me, who I believe I write pretty well, I would stand out with my writing from a whole bunch of other people. Now anyone can write, okay, because they can put it through the GPT and get it rewritten or get it commented on or whatever they want to do. So it's sort of a leveler. But when I read something, I can tell it's written by AI. I don't know why, it just feels like that.

James Killick (29:33)

Is it the dashes?

Martin Moore (29:35)

Well, see, people talk about the EM dashes, and I wrote my book, no bullshit leadership in 2020, and I've got em dashes all through it because that's the way I write. Instead of colons, I'll often use EM dashes because they're just a little bit less formal. But now that I've heard that AI uses them all the time, I'm trying to change my writing style, for goodness sake, so that people don't think I'm a fraud. It's hilarious. But anyway, so it's leveled the playing field a lot. But I hate the way it writes. I just do. And the reason is I try and write in the style of the Economist. Very direct, punchy, little bit of wry humor throughout the thing. And when AI writes, it's just, it feels like it's blancmange. It just feels gray and beige and all the same. And it has this particular style it uses. Now, is that going to change? Am I just being like prissy about it? Or is that genuinely an issue that everyone's writing is going to look the same?

James Killick (30:33)

Oh, yeah. I mean, that's going to happen. And AI does things logically in patterns. So if you ask it to do a list with bullet points, slightly to give you a heading, three bullet points, a heading, three bullet points. But that's not how humans write. Some might have two, some might have four. So there are ways around it. You can actually tell it to be inconsistent, to do different numbering structures. You can also give it a full tone of voice document. So that's what I do, right? I've had AI analyze countless posts, papers that I've written, things like that, and I've said, you know, extract all of the key points of my tone of voice and I've given it the things that it should do. I can't even remember what that is now, but it's all of the scientific parts of tone of voice and it's created a multi page document. And so every single project that I create in ChatGPT or Claude has this tone of voice document as the. As part of the knowledge base, right? And then the system instructions, and I probably should just tell you how all of that works in a second. So yeah, but the system instructions say reference this and then I also say use Australian English, no Americanized spellings, no em dashes, no oxford commas. And then even when I do the prompt, I over emphasize it. Remember I'm English or Australian. Sure, sometimes English.

Martin Moore (31:55)

I love the Oxford Comet, but hey, that's just me.

James Killick (31:57)

There's so many elements that I used to use just like you before, and now I stop, right? I used to love the word game changer, right? But now everyone's a game changer.

Martin Moore (32:07)

Totally.

Martin Moore (32:07)

Totally.

Martin Moore (32:08)

Yeah. It's funny, isn't it?

James Killick (32:09)

Do you want me to explain the whole project side?

Martin Moore (32:11)

Please. That'd be excellent. That'd be fantastic.

James Killick (32:13)

So this is coming back to the point about people using AI like Google. What they need to do is give it context. Now the best way to do that is, for instance, ChatGPT has this thing called projects. So you can create a project for a client, or you could create a project for a department or specific tasks that you do regularly, like LinkedIn. Then you give it system instructions and you say, act as if it's LinkedIn, the best copywriter in the world. You know, you've helped millions of people get millions of followers, blah, blah, blah. And then you add knowledge base to the. That. Okay, this is me. And you add stories, because that's what starts to humanize the AI. Right. This is my story about this. This is my story about this. Here's a couple of jokes that I like right now. You were saying about it not being funny, about it being bay.

Martin Moore (33:00)

Sure.

James Killick (33:00)

So you give it all of this context plus your tone of voice, the system instructions. You reference each of those documents. Why are they there? When should the AI use those documents? And then the hard work of the prompting is done, and then you go create a new chat within that project to do a specific thing.

Martin Moore (33:18)

So these are just standard. Like In a standard ChatGPT frame, you set up a project that has a number of different chats in it.

James Killick (33:26)

Yeah.

Martin Moore (33:26)

And you. You. One of those can be instructional before you get to the actual content of what you want to ask it to do.

James Killick (33:32)

Yes. Every project has a system instruction. A system instruction, which is like a. Which is a big prompt.

Martin Moore (33:38)

Right. Which is part of the overall project. And then you do the prompts in the. Okay, great.

James Killick (33:41)

And then you add the knowledge base just as files within that as well.

Martin Moore (33:44)

Right. Okay, good, good. So I think I'll look over to our team here. I think there is a tone of voice, system instruction for me in our chat. You know, I still don't like the way it writes. I don't know, it's just me. Call me crazy, but I always want to change it when it comes through. Okay, so I guess this is probably going to improve over time. Right?

James Killick (34:04)

I would say challenge accepted. I will do your tone of voice document until you like it.

Martin Moore (34:09)

Thanks, James. I love it.

Martin Moore (34:11)

I go.

Martin Moore (34:11)

I would never use that word, but like I said, I'm probably just being a little bit too finicky. Excellence over perfection. Right. But once again, you know, I have said I will. I will never use AI for a podcast.

James Killick (34:22)

Yeah.

Martin Moore (34:23)

Like, I just want to generate it from my own head. In a world of, you know, everyone else doing everything through AI and taking the productivity gain. I want this to remain real. And so that's. That's just basically a choice I've made. You'll suck up my time like no tomorrow, But I think it's worth it. All right, James, just to finish off, I'VE got some rapid fire questions that EM has given me. Ok. She said I have to ask them, so let's see what happens. First, one thing every leader should automate this year.

James Killick (34:49)

One thing. Meeting transcripts.

Martin Moore (34:52)

Meeting transcripts.

James Killick (34:53)

Yes.

Martin Moore (34:53)

I like that. Are they easy to automate?

James Killick (34:55)

Yes. As long as you have somebody that's. Or somebody. Some AI that is recording the meetings.

Martin Moore (35:01)

Okay, good. So you could do that with Fireflies or Fathom. Yeah, Fathom. Okay, great. One leadership skill that becomes more valuable because of AI management.

James Killick (35:10)

I would say because now you have to not only manage humans, but the AI agents as well.

Martin Moore (35:17)

Ah yes, good answer. I like that. The biggest myth that leaders believe about AI and technology in general. General.

James Killick (35:24)

That you have to be technical to be able to use AI effectively. I'll tell you, I've never written a line of code and I also got my 84 year old father to use AI.

Martin Moore (35:35)

Nice.

James Killick (35:35)

He is now much more productive.

Martin Moore (35:37)

Right, okay, fantastic. I love that. And last one. The mindset shift. Leaders need to thrive in the AI era.

James Killick (35:45)

I think they're going to need to be really flexible. We are moving so fast and that rate is only going to increase. So you've got to be ready to adapt to a lot of different changing scenarios.

Martin Moore (35:58)

Right. Okay. Fantastic. Great answers. James. Thank you so much for joining me today. It's been enlightening for me. I've really enjoyed it. Where can people find out more about you?

James Killick (36:07)

So I do have a school community. It's called Make Money with AI. Kind of self explanatory.

Martin Moore (36:16)

Good.

James Killick (36:17)

I also run a business that specializes in sales. In particular AI powered sales. So we help people turn leads into revenue quickly. That's Enjin co n J I n co.

Martin Moore (36:31)

Fantastic. And I believe you have a giveaway freebie for our audience which would be awesome.

James Killick (36:36)

Yes, I've got a bundle actually. So I'm going to give away my prompt library.

Martin Moore (36:41)

Right. Fantastic.

James Killick (36:41)

GBT that's designed to create other GBTs.

Martin Moore (36:45)

Right.

James Killick (36:45)

This has been super effective for me because I love gbts and I use it probably once a week. And I'm also going to give away an AI readiness blueprint.

Martin Moore (36:55)

Oh, nice. Love it, love it. That's fantastic. Great. So and, and they'll be able to get those from links in our show notes, which is fantastic. Thank you very much James for joining me today. Really appreciate it. All right, so that brings us to.

Martin Moore (37:09)

The end of episode 377.

Martin Moore (37:11)

I really hope you enjoyed it.

Martin Moore (37:12)

But as I'm sure you know, listening is easy, leading is hard.

Martin Moore (37:17)

That's why we created Leadership beyond the.

Martin Moore (37:19)

Theory, our flagship program that turns insight into action and action into results. This is where we unlock the secrets of elite leadership performance and give you the tools that you need to retain your value irrespective of fast moving technology. Shifts. I'm looking forward to next week's episode.

Martin Moore (37:37)

Until then, I know you'll take every.

Martin Moore (37:38)

Opportunity you can to be a no leader.