The Cheapest Way To Learn AI Is To Use It

Bren: Hello and welcome back to This Is Not A Data podcast. This week, I'm flying solo as Leno is enjoying some well earned time away, but fear not, as I'm joined by a good friend, Jamie Fry. Jamie and I used to work together years ago and he now runs a consortium of senior cloud and data specialists. Jamie's business, the HawkeFry Group, is up 40 % this year with the same people and he's taken more holidays than in the previous three years combined. That is impressive to say the least, but how has he done it? We get into the real return he's seeing from AI. and where the human side still slows everything down. If you've been wondering whether AI really pays back on the investment, this one's for you. Let's get into it. Hello and welcome back to another episode of this is not a data podcast. Now I'm sure you everybody listening will know how it works by now, but today is actually slightly different because Leno, my good friend and co-host is off on his holidays. He is enjoying some well earned time off, but that doesn't mean that we are not joined by an expert in the data space because today I have with me. A long-standing friend actually and somebody that we used to work together a while ago but today I have Jamie Fry who's building out the Hawke Fry group but I'm listen Jamie I'm gonna do your massive disservice but do you want to just give a bit of an intro so you can you can say exactly how you want it to sound

jamie: Thanks, Brendan. Yeah, I mean, we do we do go way back. Brendan was once my line manager back in the day.

Bren: That's a long time ago,

jamie: yes, I run the Hawk Fry group. We are a consortium, so it's a little bit of a different business model to your traditional agency. there's eight of us. We all have worked together for around eight years. And a lot of us are senior kind of well, principal staff level in in the cloud space in data. predominantly so we're made up of cloud clo cloud architects, data engineers, analytics engineers. To be honest, a lot of us have kind of the full kind of skill set over the years. And the reason we are a little bit different is each of us runs our own individual business and we roll up into the Hawk Pro group as a consortium. So we work under a framework agreement. So it allows us to go to to market together as one entity, allows us to team up on projects together, and it's just A lot more fun working together and supporting each other than it is if we were all to go solo. So it's a bit of a lifestyle business, but one that also is quite successful, so it does make you know good commercial success. And it allows us really to just pick things that we enjoy doing. And as a result, we do a lot of applied AI recently, which is probably what we're gonna talk about today. And I would probably say we've got a lot of real use cases we can talk about. to kind of show where it is making an impact and where it is changing the industry quite a bit. and also I think i in a year's time it's all gonna be different again. So we'll probably be changing our delivery again within a year.

Bren: It's great to hear that it's been a success having followed you for a wee while. is great to see. Although you say that you've got the full end-to-end stack, I am immensely envious because some of the technical code and some of the bits that you guys can bring to life is very impressive. But I think that the topic of conversation around AI and building a business and people is so pertinent at the moment. And you mentioned that in 12 months time, it could look very different. How have you used... AI to or have used AI to help scale and grow your business? I guess it's the first question.

jamie: Yeah. So we've been we've been using it for three years really and obviously the the first year it was very not super capable, but we put a lot of effort in into trying to get it to work anyway, because we kind of felt that well this is only gonna get better, so the more effort you put in to understand it now, the easier it will be later. Which has paid off.

Bren: And was that in front of clients as well?

jamie: not in front of them. I think the first year we were using it behind the scenes.

Bren: Okay.

jamie: Second year we did start to use it in front of clients. And in fact, the first first AI thing I put in production for a client was when the Gemini model first came moved out of preview to general availability. And it was the only AI model that could do like ingest like PDFs and images and stuff at the time. Right. So Chat GPT hadn't couldn't do that yet. And this was because Google already have their own Vision AI, so it was easy for them to go and put it in. And that that system was they would throw in emails and documents, whatever they want, into this web app we built and it would just extract all the data they needed into a structured database and then from there it just funneled into automated reports they had and things like that. So kind of like a a a small use case but one which for them saved them a lot of manual time and kind of proved to them that AI actually had a use case and was useful. move on a year with the same client We're now using AI with them to build all their internal operating systems. So we're not embedding AI into those systems, we're just using it to move incredibly fast and build their in-house software. Because they're in an industry that's very underserved in software. You pay a lot of money for something which does like 60% of what you need. Right.

Bren: Yeah.

jamie: And so we've basically just been sprinting with them to f to to build all their in-house operations software, which is which has been amazing because you can put real numbers on that from like time saved. the the amount of data they've now got which they didn't have before to help them with decisions. and I think that's been great because their view was always we can try and embed AI into things but it's like well you can just use AI to get to the end goal quicker, not necessarily embed it into everything. So we've gone

Bren: And he's not.

jamie: from like building data platforms and things with them and other clients to now also building in-house software systems and so the whole kind of we look at we look at a business now as Well I can look at a business as not just from like a data perspective, like one corner area. I can look at it from the entire business, where are we going? What can we do? And because we've got the right skill set to work with the AI, we've got systems thinking. I don't necessarily need to know all the coding framework languages in detail, but I know enough to read up and go, This is what we should use, and then tell the AI this is what we're using, right?

Bren: And when you talk with clients, kind of in that example, and you're them move forward, is your first thought AI can do this? Or is it actually kind of understanding what the problem is? how do you approach?

jamie: It's the same

Bren: such a...

jamie: as it's always been. You approach the problem. So if y I mean, you know this as of working in strategy before and sitting down, you know, i if you're if you're someone who works within your company and there's issues, what would you do? Well you sit down and you talk about it, you figure it out. So you go, Where where's our blockers? Where's our opportunities? Okay, our margin is you know, let's say your margin's 10%, but it needs to be 20%. Like you go and look at the numbers to figure out where that is and how can you adjust that. So it's exactly the same approach you would take before AI came in to to try and solve what your problems are. The only difference here is you've got more tools at your disposal to solve them. You haven't got to go and buy you haven't gonna go to like a big four or a big expensive agency who's then gonna turn around and say, Yeah, it's gonna cost six s well, seven figures to go and build that system in-house, at which point you go, well, the cost of build and the time to get there is probably the biggest issue. It's not worth it, therefore we'll go buy something off market that can kind of do the job and help us out. So I think the tooling at your disposal to solve problems is so much better. You

Bren: Mm.

jamie: need it's and you can get there so much quicker. So the time to market to solve that problem is quicker, right? Which means you can fail quite faster too.

Bren: And obviously the question that I've had bit of a ding-dong with a few people on recently is the ROI on using AI. I've said it's minimal at times. What's your view?

jamie: so I've got a real number on this. So I've I've just had my year end and so I was working out on the projects where we're using it quite heavily, what's the return? And the return that we've got is a 5x at the moment. So that's

Bren: Wow, okay.

jamie: 5x on both the margin but also on on the speed, right? So the reason it's not 10 or 100x is because there's still when you look at a project end to end, not everything You're not using AI for the entire thing, right? So I'm not using AI to talk to people and like on a meeting, right? so there's a lot of things that AI is not able to speed up right now, which still slow the process down. But it has significantly sped up the technical delivery side of the process. So if I'm building like a green-field data platform for a client, a lot of the effort is now up front and on the end. So up front trying to figure out what their requirements are, what their business is, how you're modeling, you've got to model their business in data, right? So you've got to really understand their business. You've got to put it out of people's heads. People aren't natural at trying to communicate these things to you. So you've got to spend the con you know, do your consulting piece of pulling it out. Then when you know what you're doing and you're going, then we s you know, we we have our own harness in house around building data platforms and modeling data, which is very, very effective. So that gets us to where build actually getting the platform up and running extremely quick. And then the remaining part is then user acceptance testing really is making sure we've got everything right. And again, that takes time because it's people want to go and check the data themselves. They want to go and make sure actually what you find is some of the things that you've been told by them is not actually true and they've misunderstood parts in their own business, right? So it's back and forth with stakeholders and other people in the business. And you know that human to human communication is not efficient. It takes time. So

Bren: Yeah.

jamie: so the so overall it's five X, which is obviously phenomenal. But if But it's never gonna be a hundred X because the human element to it as well and w which goes with it.

Bren: And obviously the genesis of this podcast was the people side and everything outside of just the technical elements of the data world. And it's interesting you say, they're almost a touch point between each human interaction is the part that can be truncated, the person is still bottleneck is probably a bit unfair. Was it, is that.

jamie: Yeah, it's probably the wrong word. Like it's it needs to happen that way. You can't like you

Bren: Okay.

jamie: need to have peer person communication. but it it's a part of the process which it can only be sped up really with like I suppose a pr approaches to how you communicate and how you work through these things. So do I think we can speed that up more? Yes, but it's not gonna come from AI, it's gonna come from me putting in like ways of working up front with the client to get things out of their head quicker and also spoon feeding them the UAT process better. So like that's something I've done with one the other week where we made it so easy for them it's almost hard for them not to to do it. So it's things like that which can kind of get me to that bigger piece. But but the whole 5x piece overall is is been amazing because it means one is obviously my margin is is way better. But I'm also able to do it quicker, which means I can fit more in over the same time span, which means more revenue, which is also what's occurred. And I can also bring the price down for the clients too. So for some clients where it might not have made sense pre AI for the cost, now it does make sense. Right? Which is great.

Bren: And you've seen is that an active conversation because the general assumption is if a client or a company uses AI ergo, it should be cheaper for the customer. Is that a conversation that's been had from your side?

jamie: they don't really bring it up like in that way. the way I usually message this is we will use AI and we will actually drop our wrappers or harnesses into your environment so you have it as well. So

Bren: Right.

jamie: they're not just getting things quicker, they're actually getting a new capability because they can take that harness and they can basically employ an average person who can then run with it, right? They haven't got to go employ the really highly skilled person so it costs them a lot of money. They can employ someone who's not maybe not quite is dangerous enough and not quite there yet, or maybe still on their learning journey, who can kind of kind of get to grips with it. And they can continue to build themselves as well and and main and maintain what we've built with the assistance of AI. And that's because of the harnesses that we've dropped in with it. So actually I don't so then the conversation of this should be cheaper never comes up because we've already given them extra value that you're not getting anywhere else.

Bren: And that's almost sort of old school business approach, isn't it? Under promise and over deliver. And even in the world of AI, that still sounds like it rings really true.

jamie: And and we fix prices as well, so we take the risk away most of the time.

Bren: Okay.

jamie: So obviously when there's very unclear requirements, it's almost impossible to fix price. You have to take a different engagement approach. But initially, if you're if you you're setting up a data platform from scratch and you need they've already got some very key business critical use cases that go along with it, you can spend a qu a bit of time up front understanding all that and then give them a fixed price and say it's gonna be delivered you know, it's gonna be delivered by this price no matter what. And that's really easy to anchor it against the open market where people try and do time materials because you're gonna be like, well, you know, your minimum you're spending if you do T and and things is gonna be X, and you might not even get it, so it could cost you Y. And it's like you are gonna get it, and you're gonna get it critical. So it's almost hard to say no. And it's it's

Bren: Yeah.

jamie: just it's been a very easy playbook this year to do that. And obviously we're able to do it because, you know, we put a lot of time into learning to use these tools and we have a lot of expertise in in our which we put into our harness. And so we know that that reward has now come come in and we are debating whether we even maybe open source this or give it away or put it behind a paywall or something to let other people use it. I'm ingreing about that at the moment. you know, es essentially give people the shovels. so they can do it themselves.

Bren: And how, what's the take on sort of multimodal and not being sort of hamstrung to a particular model? mean, you've seen the press today, was it this week? Obviously, I say obviously OpenAI has decided not to release its latest version for whatever reason.

jamie: Mm-hmm.

Bren: In your, kind of in your work, in your world, do

jamie: Yeah.

Bren: you have the ability to switch? Is it a sort of, in a vertical, is it a future-proof way of doing it? Because clients, do they worry about that sort of thing?

jamie: Yeah. Yeah, we're pretty we're pretty modo agnostic ourselves. So I mean we use three models predominantly.

Bren: Okay.

jamie: I mean two two is our daily drivers, so I mean we use Claude and Open and so yeah, Anthropic and Open AI as our daily drivers. And you you get to kinda learn each model has its own little like personality. It's quite funny to work

Bren: Yeah.

jamie: with so you kinda get to know what they're good w what which models are good with what, so you can flip between them. and the reason those two are daily drivers is because they they have good harnesses themselves which work with what we've built and we have them play off each other. We also use Grok a little bit, but that's more for a bit of fun. so we have our

Bren: Okay.

jamie: own kind of AI employees that are always on and live in the cloud. And so you can communicate with them via your phone, via Slack, whatever. And they have access to our code bases and things. So they can just you can even give them tasks in our project management and then they're set to wake up in at midnight and look at those for six hours and just work on them. So we actually have had

Bren: Okay.

jamie: There's one client which we're happy to do this where we've had an AI employee pass part of the delivery. And so the

Bren: Wow.

jamie: client can even tag them when they can work, you know, which has been cool. so you know, those ones run on a mixture of Codex and Grok. And the reason

Bren: Okay.

jamie: we have that is they have their own little strengths. Grok is also a little bit fun. So we have a an AI employee who's basically a bar, a bar pub owner in our Slack channel,

Bren: Right.

jamie: is like a British bar. And you know, you could just call his name and he keeps a running tab of everyone's like drink orders and stuff and he like calls it in every two

Bren: Amazing.

jamie: weeks. You know, and you can only do that with Grok because of the way the model is. Like, the other models are little bit too tight with that. so so yeah, we Yeah.

Bren: Yeah. And when it comes to, I mean, you talk about agents and I mean, I've been playing around with a little bit of a chief of staff to try and build my business. you, how does it work within that, inside the HawkeFry group? How have you decided to kind of, I guess, scale and grow and use AI? Cause if you don't, you're probably missing a trick.

jamie: Yeah, so we've got we actually built some we two two of us actually went out to Mexico back in May and we spent a week building this because we were like, let's just do it. And

Bren: Okay.

jamie: we we went and took the Hermes harness, which is a little similar to OpenClaw, but in my opinion, like it's more opinionated and it would just we works a lot better. We took

Bren: Okay.

jamie: that and then we kind of built some of our own security and develop and DevOps practices around it. So we improved that. And then we brought in our own some of our own harnesses, our own kind of coding harness and our own tools and systems that we use as well. So for example, we just use linear for our project management, which is actually a place where our c agents communicate, which I'll go into in a second. we have our own pipeline and CRM tool. So we actually just built them as a data warehouse with a CLI on top, so it's headless, so there's no UI, and our agents interact with it.

Bren: Just explain that for anybody listening. When you say headless, what do mean?

jamie: It means there's no there's no web application, right? There's no

Bren: So you just boot it up effectively.

jamie: it's just a hosted warehouse in the cloud with a what we call command line interface on top, which is what

Bren: Yeah.

jamie: the agent communicates with so he knows how to pull and write data. And so

Bren: And that's what Salesforce have decided to go because they've talked about going headless. Is that right? Do you know?

jamie: I d I don't know, but it's to me it's like if your agent's doing everything, you don't need a UI. And if you do need a UI, you can spin up a temporary one to look at it. Right? So if I want to know what my numbers

Bren: Yeah.

jamie: are, I just go on my phone and I type out to my chief of staff what's my pipeline looking like and she'll come back with the numbers straight away and then I can go, actually I just closed that one yesterday and she'll be like, Okay, and she goes and writes she updates it. So all my

Bren: Okay.

jamie: pipelines fresh. Likewise if I was on if you were if this was a sales call after this call, I type a command in on my computer and what it would do is it would take the Granola script, because I Granola, which is record the transcript. It's great. Yeah. I

Bren: Mm-hmm. Yep. How good is granola by the way? mean...

jamie: mean it's it can there's definitely some things that can be improved, but it's good. and it processes that script. What it does is it look if it works out as a sales call, it then goes and updates my sales pipeline with what we talked about in the call. So updates that,

Bren: Nice.

jamie: updates the CRM as well. So if there's a new person on the call, add them to the business, blah, blah, blah. And then it also goes and backs up into my company knowledge vault as well. So we create like a network graph of all the links between individuals and businesses and projects. And

Bren: like a second brain.

jamie: yeah, yeah, basically. And that that's all that's all done straight away. And so of course, my chief of staff can access that too. So my all our agents can access that company brain, they can access the pipe drive, the pipeline. and you know, we've bundled that all into our AI employee kind of operating system. So if I deploy a new AI employee. it's all it it's all all those tools are built in with it. So we we spent a lot of time doing that because we were thinking about potentially doing that as a certain new service for companies. The reason we decided against it is self-hosting and managing that on someone else's behalf is huge risk because if the AI does something it shouldn't do, you're liable.

Bren: Yeah.

jamie: So the liability would be huge. I don't think I'd even be able to get insurance for it. So so my thinking was well

Bren: Okay.

jamie: we could in theory then deploy it on their behalf. So they own all the infrastructure. But then there's lot of people who just they're never going to be able to deal with that really, even with agents at their fingertips. And so we may end up open sourcing that as well at some point.

Bren: Interesting.

jamie: but yes, in in-house we have these always-on agents which we work with, and then we also work with agents on our computer just using like claw code and codecs. That's our daily drivers that do a lot of the workhorse stuff. But then the always-on agents you can tag to to they live in on a virtual machine, you can tag them to be like, Help me on my project or or whatever

Bren: Yeah.

jamie: you need. Yeah.

Bren: And does that mean that you do you intend to grow HawkFry people? mean, I know it's a consortium, but in terms of employees, does the agent approach mean that that's not a route you want to go down or is it just giving you capacity and bandwidth to be able to kind of stay on top of what's happening? Because the people side is obviously hugely important, but if you can grow and scale and keep margins good using AI, that would make sense, presumably.

jamie: Yeah, so I mean we've I think we're think we're forty percent up this year. and that's with the same inputs, right? So there

Bren: Holy. Wow.

jamie: isn't any additional headcount or anything. it's pretty much all been just been able to do more in in the year, really. And

Bren: No. Well, if you've been in Mexico on the beach sunning yourself, presumably things are going well.

jamie: Well, this is the other thing. I've I've taken more holidays this year than I have in the whole three years of the business as well. And I'm

Bren: Yeah.

jamie: not even I don't work weekends anymore like I used to have to back, you know, when I first started the business. And so I've actually had a better life balance this year as well. And we've been able to do even more. So

Bren: Nice.

jamie: this and I think people need to forget forget about this because you you have a lot of people who say they're tired because they they feel like they've got to always be on using all their AI credits and stuff like that. That's just To me that's crazy. It's like there's nothing wrong with having a day or two or you haven't used AI or you're not needing it's not doing something. Like

Bren: Yeah, yeah.

jamie: you you've got to remember when to stop and slow down. and I think for us, AI has allowed us to do more, but we've we're we're quite self aware in like we say no to work as well when we don't wanna do work. So it's allowed us to also be able to say no more to things which don't either don't suit us or we just don't you know, we're just like, well, I don't really want to be doing that at that point in time 'cause we already have enough on and so I don't want to be working evenings. So, which is great 'cause it means the work you do do, you can like they know they're getting a lot of our focus and they know they're getting a really good quality output. 'cause you don't have that that pressure to to to always be bringing bringing in more.

Bren: I mean, apart from, I've got so many questions, but you know, from a, just quickly from a founder's perspective, and you talk about not having to work weekends anymore. Did you, was burnout a real thing? Were you aware in terms of the pressure to be always on and did AI kind of help or hinder in any of that?

jamie: Yeah, It's I mean burnout has always been an issue for me, even before as a founder. It's kind

Bren: Okay.

jamie: of in my nature to get a bit obsessive with what I'm working on. and so that's always been an issue, burnout. Like I'm it's in my nature to go and spend a few weeks day and night working on something and then not do much for two weeks. That's always been a nature of of mine and also some of the guys in the consortium as well. they're like that too. But the AI has been a allowed me to I suppose it's it's allowed me to get to that end goal quicker so I'm not spending three weeks working on that problem to get to that end goal and burning out. Instead I've spent a week of really hard work with AI, maybe twelve hour days, working on it and then I've done it and then I've kind of got it out of my system then, right? So I don't know, maybe it's an ADHD thing or something, I don't know. But I I would say burnout's always going to be there just 'cause of the nature of who we are. Like

Bren: Hmm.

jamie: so and I think with some people who say they have burnout of AI it's because they don't have the ability to stop. They have that they keep going. So I think it's more of a human thing of just knowing your limits and knowing when to stop and put your tools down and go outside, touch grass rather than the AI being the problem.

Bren: Yeah.

jamie: so Yeah, that'd be my experience of it. Obviously with a f as being a founder, it's a bit different 'cause you're trying to get a business up and running at the same time as delivering, doing the sales, expanding to literally everything. But the cool thing about being a founder is those things compound. And so this this year, for example, all of our work has been from customer referrals or our partners. Like I haven't had to go out and do a lot of business development this year. So the previous two years have compounded and allowed me to have that breathing space.

Bren: Yeah.

jamie: Which has been nice.

Bren: And presumably with people in the consortium being able to just pick up the phone or have a chat with somebody and just say, today's a really S day, know, and just being able to share that burden has been helpful.

jamie: Yeah, I mean, sometimes life happens, you know, some of the guys you know, one of the guys lost some family members this year, which was obviously sad. So it meant we could

Bren: Gosh.

jamie: step in and help them out, which if they were by themselves, they wouldn't be able to, you know, even be able to do that. So, yeah, the it's it's a com comradeship approach as well. So, it's a high trust approach. Like we don't just let anyone in the consortium, it's people we've already worked with and we've been in the trenches with. So, you know, some people we've met like clients and then they've decided to come in, right? So we've already worked everyone in the consortium has already worked with each other in some professional capacity, be it a business together or be it like one, you know, one of you was a a consultant, contractor, engineer, whatever, work agency on the other person a client. So so yeah, and I think it gives people the freedom as well, 'cause, you know, at some point some of those guys might decide they want to do something else and there's no there's no nothing stopping them, which is great. And everyone has their own things happening at the same time. So I think one of the guys, he he's working three contracts with me at the moment, but he also has another two contracts himself that he does for some other things. So and w we don't necessarily do the whole time materials thing. We do we try and do a well we'll we'll we'll say what we'll deliver and we'll we'll give you a price for it. So it allows us to then actually figure out our own diaries to work with the clients and ourselves and also it means we can speed things up with AI. Because if you think about it, if someone said to me, Well, this used to take three months but now it's only going to take two weeks because you can use AI, so I only want to pay you two weeks worth of work, I'm gonna tell them to jog on obviously, because it's like, well,

Bren: Yeah.

jamie: you go do it yourself then. And obviously they won't because they haven't got the expertise to drive it. So

Bren: Thank

jamie: it's kind of like, you know, the whole paying for time thing, I know I'm going on a tangent here, but the whole Paying for time thing has never made sense to me. And I think the only reason people do it is because it's easy to transact. Alright? So you go to market, you go, I want a contract for three months. It's like, great, you can go have one for three months, but you might not get what you need. So you might have open and if you're trying to compare someone's day rate against someone else's day rate, it's is stupid because someone else the person with a high day rate could probably do it in a quarter of the time. So overall it costs you less. So the whole the whole piece around you know, the whole piece around selling time doesn't make any sense to me, which is why we we don't, so most of the time. There's very f there's some exceptions, but most of the time we don't.

Bren: And I think to me that's where the world is going because certainly the certainty that clients and customers want is really what they're after, isn't it? They want an outcome and a certainty on what it's going to cost them to get that outcome. But for everybody listening, mean, thank you for explaining what you've been up to. Some of it was relatively technical and in the detail. For anybody that's listening, how would you suggest that somebody gets started with AI in inverted commas? Like if somebody came to you and said, I don't know, what's the standard question that comes to you that say, you know, God, can't, I can't get access to this report because the data is pretty crap or whatever it is. I need a data platform. All of those sort of broad broad statements.

jamie: Yeah.

Bren: What, what's the first step somebody should take rather than just getting overwhelmed and being like, gosh, I've been left behind.

jamie: Yeah, most of the time people come to you with an idea, so they either come to you and say, we need AI or we need data platform or something really high level. And

Bren: Yeah.

jamie: then you go, Well, okay, what's the actual problem? What's

Bren: Yeah.

jamie: the actual like what is it you actually need? So it just comes back to that kind of getting to know their business. And they go, Well, actually the problem is the real example, we can't deploy capital fast enough. Okay. So you've so their their assumption is AI can help with that as like a potential silver bullet.

Bren: Yeah.

jamie: And in some cases, yeah, it can. In other cases, actually no, it's just a different approach you need to take. so my fur so if if you do want to try and understand AI and its capabilities to know whether it can solve the problems you've got, honestly the best thing to do is just just try it and use it yourself. So that's the first basic step. You'd be amazed at how many people come to us and they haven't even used AI. Like already they're going out shopping for it. And maybe they've used it a bit in their personal life on their phone as a glorified Google engine, right?

Bren: Google search.

jamie: And I see this all the time on trains to London. I always look around what people are doing on their laptops because I'm interested to see if or using AI.

Bren: You

jamie: And where I do see people using like ChatGPT and things, what they're doing is they're putting things into it to write an email for them and then they're pasting that out into the email to send. And then they go back to their Excel spreadsheet and start doing things in manual Excel. And I'm like, you're using it for completely the wrong thing. Like you should be writing the email and then you should be getting AI to do the Excel stuff. So so I think

Bren: And is that because people just are not aware of it stuck in the process or you haven't really thought about it? Because you're right that the whole point of AI is to help kind of alleviate some of that pain, the copy and pasting is a prime example.

jamie: Yeah. It's it's it's it's thinking. So the reason I say just get to you try try and use it first. So like obviously those people were already you trying to use it, which is great. And that's like using it for writing emails and stuff, that's like baby steps first stage. But you st you need to do that. You need to do that to understand it. And eventually you get to understand how to prompt. Yeah,

Bren: Yeah.

jamie: you understand how they work. Then you start then you come into the world of hallucination, right? You're like,

Bren: Yep.

jamie: it's hallucinating like that's you you kind of it's a whole you need to go through this learning journey yourself.

Bren: Hmm.

jamie: And At some point it clicks and you go, I can use it for XYZ like everything, XYZ. And at that point it clicks, you really start to understand. Then you start to get dangerous with it. All right. And the people who get to that clicking first are generally the people in your business who are tinkerers.

Bren: Yeah.

jamie: You know, there's kind of like people you usually find writing XL macros and stuff like that. so like they haven't got a tech background, but they've kind of just got they've been given. they're in a businesses or Microsoft, they haven't got very good tools. And then they they just kind of work away around with the tooling they've got to to do something. those tend to be the people who click first, I've seen, and are the most dangerous with it at the moment, than actually even like people who have a tech background. interestingly enough. So yeah, if you wanna get to grips of AI and you're understanding it and you're going out to people and you're paying, you know, hundreds of thousands for like an AI strategy and stuff like that, like Before you even get to step, just just use it. Just use it to get an idea. And don't go all gone ho by buying licenses for everyone and just expect it to return something because everyone then will be going on that baby steps journey. And guess what? You get more chaos because Jane over there has decided she's going to use AI to write horror emails without understanding that it can make stuff up. And then, you know, Bob over there is decid who's a tinkerer who's decided to build his own in-house CRM system.

Bren: Yeah.

jamie: which undermines what the IT have done, putting in place one already. And suddenly you've got two sources of truth for CRM. So you you the thing is you can actually cause a lot of chaos as well if you kind of throw it out there at everyone before really understanding what it can do in educating people. So what I'd say is use it and but use it in a sandbox environment first and just let everyone go crazy in a sandbox way to understand it before you start then looking at how can I now change how my processes and my business works a bit.

Bren: It's funny because I put a post out on LinkedIn in the last few days, but I got weak. We made the mistake in the previous place where we gave 50 people a co-pilot license and said, here you go. You bang on about it. Fill your boots. Three months later, Dave rewrote an email. Karen created a GIF and somebody else sort of sent a funny note. And it's like, cool.

jamie: Yeah. Yeah.

Bren: But that really didn't, and we, you looking back, and this is what I wang on about now, is it's about being problem first. What is the problem you're trying to solve? And then tackle that. But your point is absolutely spot on. Just try it. If it all goes wrong, close the chat, close the application, start again. Like,

jamie: Mm-hmm.

Bren: you can't go... touch wood too far wrong but if you don't try and understand what it's capable of and I think you're right you have that matrix moment where you just see everything and you go holy shit this could this could just like do everything and then

jamie: Yeah.

Bren: the hard part is not trying to go after everything be like I should just keep building stuff like this because you end up

jamie: I mean that's that's where I am at the moment. I've got so many things I could go after and just have to like slow down, you know, and Yeah.

Bren: Well, mean, Jamie, listen, I mean, the last half an hour is absolutely flown by and I would dearly love to carry on the conversation. But for everybody listening and they want to kind of get started with AI or they think that they could do with some conversation with you, what's the, how can people get in touch?

jamie: yeah, so on if you go to hawkfry.com there is a a contact form you can fill in there. alternatively you can connect with me on LinkedIn at Jamie Fry. or if you're just interested in following what we're doing, we do have a Hawk Fry Substack. And we're looking to post a lot more on there soon. So I mean this year's been so busy just heads down doing stuff. We've got all this gold mine. And I'm we're trying to figure out how do we get this out into the public to educate people. So at some point we'll have some YouTube videos out and things like that, but a lot of it will probably come from the Substack first. So yeah, I think those are the best ways.

Bren: Wicked, well, I will definitely keep an eye out for the sub stack and I won't tell everybody that you and I did start a podcast a few years ago that we chuckled about before we started recording, which we don't know whether we'd still survive now. I

jamie: Yeah.

Bren: mean, Jamie, thank you so much for the conversation. To everybody listening, this is the part that I need your help on because really we want to. make AI work for us and play with the algorithm. So if you did like the conversation, please like and subscribe. I mean, that's what the cool kids say anyway, but ultimately we're just after a review and only five star ones. So please

jamie: Yeah.

Bren: do leave us a review, get in touch, ask any questions. But Jamie, really appreciate your time. Next week, hopefully Leonard will be back from his holidays and we'll carry on the conversation. But until next time, thanks very much.