AI Without Strategy Is Theatre. Process First with Kate Fabian

[00:00] Kate Fabian - Adoptech: Oh, I guess [00:05] Matt: I'm glad you said row. That [00:07] Dave Pengelley: What? [00:07] Matt: was that could [00:08] Richard Webbe: So [00:08] Matt: have been [00:08] Richard Webbe: Dave, [00:08] Matt: weird. [00:08] Richard Webbe: where where did you bury the body? And who knows? [00:11] Dave Pengelley: No, no, I was saying I was just checking Kate's our guest. It's a first time long long long time listener, first time caller or watcher or joiner or whatever it is. Uh and I was just making sure she's comfortable being on the top row. I mean I can easily just throw anyone up there. She can she can anchor the thing for like if she wants. [00:28] Kate Fabian - Adoptech: I'm I'm not sure what [00:28] Dave Pengelley: Like [00:28] Kate Fabian - Adoptech: saying. [00:29] Dave Pengelley: that's fine. Oh [00:31] Richard Webbe: She's a lot better looking, a lot more professional than us, so anchor her wherever you want. [00:35] Dave Pengelley: well you know we want to lift lift women up in the world so we'll we'll put her put her up on the on the top [00:41] Kate Fabian - Adoptech: Thanks, [00:42] Dave Pengelley: row [00:42] Kate Fabian - Adoptech: thanks, Dave. [00:44] Matt: I don't know what I'm [00:45] Dave Pengelley: um [00:45] Matt: gonna do today without Kate in the comments. It's gonna be weird. [00:49] Dave Pengelley: well tell your friends drop it in your chat groups everyone get out there let people know we are live live on the youtube's [01:01] Dave Pengelley: Matt's Matt's commenting for us. Good on you, Matt. [01:03] Matt: That's right, gotta get that algorithm pumping. [01:07] Dave Pengelley: The trick, the trick is we've all actually got to like tune in five times on different devices so that way we just sort of simulate that there's there's other more people out there. And then the other room goes, wow, there's like 15 people watching. It's just all of us five times each. But then it goes, Oh, now I'll send more people there. I'll show people. Look, there's a show with 15 live people on it. Like that's amazing. Uh at midday on a Wednesday in Sydney. [01:30] Matt: Would be be to get Sam and Elon to argue about us. [01:34] Dave Pengelley: Well, yeah, I don't know. Is that then what did uh Twain say? The only thing worse than bad publicity or people talking negatively about you is no one talking about you at all. [01:45] Kate Fabian - Adoptech: Yeah. [01:46] Dave Pengelley: Yes, yes. We will we will get on to uh all that sort of stuff. But uh Kate, welcome to the show. [01:52] Kate Fabian - Adoptech: Thank [01:53] Dave Pengelley: How how [01:53] Kate Fabian - Adoptech: you. [01:53] Dave Pengelley: how are you going? [01:54] Kate Fabian - Adoptech: Yeah, pretty good, pretty good. It's nice and warm up here, sorry to say, guys. Um are in the midst of beautiful North Queensland winter. [02:03] Richard Webbe: So where are you, [02:03] Kate Fabian - Adoptech: Yeah. [02:04] Richard Webbe: Kate? For everyone who listens tunes in, where are you? [02:06] Kate Fabian - Adoptech: Final North Queensland. [02:08] Richard Webbe: That's [02:08] Kate Fabian - Adoptech: Just [02:08] Richard Webbe: all yeah. [02:08] Kate Fabian - Adoptech: uh Cairnsway. [02:10] Richard Webbe: There, Kim's. [02:11] Kate Fabian - Adoptech: Yeah, [02:11] Richard Webbe: Okay, there [02:12] Kate Fabian - Adoptech: helping [02:12] Richard Webbe: you go. [02:12] Kate Fabian - Adoptech: regional businesses with all this crazy [02:14] Richard Webbe: And what [02:14] Kate Fabian - Adoptech: IAI stuff [02:16] Richard Webbe: I [02:16] Kate Fabian - Adoptech: going [02:16] Richard Webbe: was gonna [02:16] Kate Fabian - Adoptech: on. [02:16] Richard Webbe: say, what does Adopt Tech do? [02:19] Kate Fabian - Adoptech: We basically help businesses do what they do only faster, better and cheaper with the power of AI. [02:26] Richard Webbe: Well, good phrase. I like [02:28] Kate Fabian - Adoptech: Yeah. [02:28] Richard Webbe: that. You've said that a [02:29] Kate Fabian - Adoptech: Yeah. [02:29] Richard Webbe: few times. [02:30] Matt: Well done. [02:30] Dave Pengelley: You you worked on you worked on that tagline, haven't [02:33] Kate Fabian - Adoptech: That's [02:33] Dave Pengelley: you? [02:33] Kate Fabian - Adoptech: yeah, that's my that's my one-liner pitch. So yeah, nothing confusing. Just basically my job is to help businesses um get their head around this this whole this whole space and basically figure out where to start. So what's the problem they're trying to solve, how to uh map out that process of of where they are now and where they want to be. [02:53] Matt: Everybody [02:53] Kate Fabian - Adoptech: Every business has [02:53] Matt: everything [02:54] Kate Fabian - Adoptech: a has their um their [02:55] Matt: it [02:55] Kate Fabian - Adoptech: their bus [02:56] Matt: it's [02:56] Kate Fabian - Adoptech: their strategy [02:56] Matt: charity, [02:57] Kate Fabian - Adoptech: and so [02:57] Matt: so [02:57] Kate Fabian - Adoptech: my goal is [02:58] Matt: I [02:58] Kate Fabian - Adoptech: to just help them leverage AI to reach that strategy quicker. [03:03] Richard Webbe: Excellent, excellent [03:04] Kate Fabian - Adoptech: Yeah. [03:04] Richard Webbe: assessment. [03:06] Dave Pengelley: Nice. Is there a particular industry you work with more than others, or you're you're across the board? [03:11] Kate Fabian - Adoptech: I haven't niched down totally but my June my most of my experience is around in the construction space. Yeah so trades and services and yeah working more with larger businesses now. [03:23] Matt: So like that's a big spread, like a trade as like a single sole operator up to like a like a development company. Company, like construction wise. [03:32] Kate Fabian - Adoptech: You actually nailed it, yes. [03:33] Matt: Wow. [03:34] Kate Fabian - Adoptech: So yeah, so I do implementation and training for job management software for businesses. So that's could be a solo trader, could be you know 10 or 15, 20 person um organization up to yes, I just recently had a client in Papua New Guinea who was actually a land developer there. [03:51] Matt: Wow, [03:51] Kate Fabian - Adoptech: So [03:51] Matt: that's cool. [03:52] Kate Fabian - Adoptech: yeah, so wide and [03:54] Matt: So [03:54] Kate Fabian - Adoptech: varied. [03:55] Matt: how like and and what you you do mainly like sort of audits and implementations. implementations or do you sort of um don't want the you don't want to use the term handball, but like uh employ subbies to to get the work done or how's it work? Is it [04:10] Kate Fabian - Adoptech: Yeah, so I I like to hold hands all the way through. I don't like to hand over a report and say good luck with that. So because I've come from like a training implementation background, I can do the the planning beforehand, actually execute the implementation and then run the workshops and teach people how to use it so that whatever we decide to move forward with actually sticks with. the business once I'm born and um I'm not sort of just setting them up to fall off a cliff as they say. So [04:35] Matt: Yeah. [04:35] Kate Fabian - Adoptech: really around that you that user adoption piece is is my [04:38] Matt: That's [04:38] Kate Fabian - Adoptech: favourite. [04:38] Matt: cool. How do [04:39] Kate Fabian - Adoptech: Yeah. [04:40] Matt: like do people find you and say, Hey, can you just teach me this AI thing? Or like how how does it usually come about? [04:48] Kate Fabian - Adoptech: It depends. It depends. Just networking is I guess the the main thing. With the with the job management software the clients come through through the software platform. So at the moment with with where I am regionally it's just about marketing and getting people out there but typically businesses will say I just want AI like what what what tools do I use and I always peel back and I'm like stop at the tools already the tools come later [05:13] Matt: Yeah. [05:13] Kate Fabian - Adoptech: so I'm around [05:14] Matt: When [05:14] Kate Fabian - Adoptech: the the people people in process first [05:16] Matt: that's what we speak about all the time, and you'd know this too, like listening, how you know they say we want AI, but like what actually is their problem? Like have you ever been able to give back to these people and be like, uh, you don't need AI, you just need to change this. [05:30] Kate Fabian - Adoptech: Yeah, actually for some of the recommendations I make are just around getting to tools to talk to each other or where they've got an overlap. They can actually retire one of the subscriptions they're paying for and just put everything into the other system that they're paying for. And yes, there's often cases where the people don't actually need AI at all. They just need to leverage the subscriptions or the tools that they've already got and use them better rather than introducing new things. [05:55] Richard Webbe: So it's a part of that whole digital transformation model that everyone's going through. True, and AI is just one of those components. [06:02] Kate Fabian - Adoptech: Exactly. [06:03] Dave Pengelley: Yeah. [06:03] Kate Fabian - Adoptech: Yeah. [06:03] Dave Pengelley: Well well welcome to the show, Kate. Uh you guys [06:05] Kate Fabian - Adoptech: Thank [06:05] Dave Pengelley: have said the word [06:06] Kate Fabian - Adoptech: you. [06:06] Dave Pengelley: AI many, many times, so let's do this. [06:22] Dave Pengelley: They always end so suddenly, the uh AI generated bumpers, don't they? It's just like you're just getting into it and then do just stops every time. [06:30] Matt: Yeah, I I would get real into it. I'm like, ah, this is really nice. And then yeah, it [06:33] Dave Pengelley: uh [06:34] Matt: cuts [06:34] Dave Pengelley: i [06:34] Matt: off. [06:34] Dave Pengelley: know i like i like my cinema advertising bumper that one's that one's a good one um so it's been a big week in ai all the things ai there's there's lots of news lots of different developments happening across the board for everyone and all the things uh so i thought we'd kick off um with uh segment a bit of a bit of a fan favorite and a one we all like that uh we can discuss a few things that's going on we'll get the the news is gonna blow I know we're going to talk for ages about the news. So let's kick off with this one. [07:15] Dave Pengelley: Wins of the week. The week. [07:20] Dave Pengelley: So uh how has it has anyone had a win, anything that's surprised them with AI? They've tried a new tool, they've tried a new technique, they found a skill in a skill library or Um something happened with a client where the light came on, like any any wins of the week this week? [07:35] Richard Webbe: I [07:36] Dave Pengelley: Week, [07:36] Richard Webbe: I I [07:36] Dave Pengelley: week, [07:36] Richard Webbe: enjoyed [07:36] Dave Pengelley: week. [07:37] Richard Webbe: cross-referencing my LLMs on a subject I was working on recently. And they were giving some surprisingly new answers. And I'm playing, is it Fable 5? Is it what's the one? Yeah, I was playing with that. That's a lot faster and a lot more powerful. So I was getting different responses to what I'd normally get. But sorry, my stuff's boring. Kate, you looked like you had something you're about to say. [07:58] Kate Fabian - Adoptech: Mine could be boring too, but It's actually on the Fable V topic. I try very hard not to get shiny shiny tool syndrome, which is why I tune into you guys every week to get the lowdown so I don't have to have to do all the playing myself. But with the Fable V free credits, I did have a bit of a play, and I was watching the usage, you know, clock up very quickly and all the tokens get chewed. So I actually said to it, you know, let me know if I could be using something else for this. And it very kindly was like, You don't need me for this, you need to just jump back on Sonnet, and it really helped. me get a grasp of like what to use fable four and what not to use it for so for the average person they don't need that that compute and that capacity of fable five so that was that was a cool little win for [08:42] Richard Webbe: That's [08:42] Kate Fabian - Adoptech: me. [08:42] Dave Pengelley: Nice. [08:42] Richard Webbe: very cool. [08:42] Dave Pengelley: Nice. [08:43] Richard Webbe: That is cool. [08:44] Dave Pengelley: When I was using clawed code back in the day with Opus pre-fable, I'd often say to it, I'd give it a bunch of like we'd back and forth and work out the plan. And then I'd say, All right, now go push that out to the lowest cost sub agent that's capable of doing the job. Use Sonnet, use Haiku, uh just Get it done, and it would then go, okay. Well, this is pretty straightforward, it's clearly defined. Haiku could do that because it's very descriptive exactly what we're doing. You don't need much reasoning. And so I would let the smarter model of Opus decide which of its baby models it would pass tasks out to again to save that token usage. [09:16] Kate Fabian - Adoptech: Yeah, good tactic. [09:18] Dave Pengelley: But Matt, you were saying, were you saying, Matt, to me the other day that Sol doesn't know about that? And if you're using GPT 5.6 Sol, it just subagents everything out as Sol agents. [09:28] Matt: Oh yeah, I would have to do Direct everyone to Theo's recently uh posted video about that because the whole subagent definition tooling that exists inside Codex CLI and Codex app itself, uh which I don't know if we're gonna be calling it ChatGPT app. It it it's confusing and weird, [09:47] Dave Pengelley: I know. [09:47] Matt: and I hate that whole rebranding, [09:49] Dave Pengelley: Yeah, just [09:50] Matt: but [09:50] Dave Pengelley: just just when we thought that were let like leveraging everything around codex, I've gone back to Chat GPT, and it's like, why open AI? Why? [09:56] Matt: yeah. Um [09:56] Dave Pengelley: Yeah. [09:57] Matt: but no, so like the the real Simplified answer is when you're speaking to an agent through the codex ecosystem, regardless of model, they have tools that allow them to spin up sub-agents. There was, let's say it's V1 and V2. This is exactly how Theo said it. V1 was very simple. It was very hub and spoke. It's very like, okay, I'm going to send it off to that one and it will give me a response. V2 is possible to be recursive and they can speak to each other within that recursion. And it's a mess. So there's also the inability with V2 to specify model and reasoning parameters. So yeah, if you are on Sol Ultra High or Max extra high, whatever, and it goes and spins up agents, it's going to be other like clones of itself, which is exactly what Fable used to do as well. When I first tried the first release of Fable and I did a bit of a quick test with workflows, this wasn't using ultra code or anything. This was just use workflows, spin up Agents do the thing. Next minute I had 90 uh agent sessions of of Fable, each doing granular tasks. But [11:08] Richard Webbe: Matt, [11:08] Matt: um [11:09] Richard Webbe: are you worried that that's in the J space using AI to speak? [11:13] Matt: no, no, no, no. No, I it's it's relevant. The JSpace is yeah, [11:19] Richard Webbe: Subconscious. [11:19] Matt: that's that's another strange thing again. Yeah, we touched on that last week and I haven't revisited that. I I kind of see JSpace kind of like daydreaming in a sense. Um [11:30] Dave Pengelley: So yeah, so okay, so yeah, and I'm finding this with Hermes as well. The way the sub-agent stuff works, I don't think I can define within an agent if I've chosen the model, it's always just going to use that model or whatever thing. I don't know that like Claude Code knew it could create haiku agents or it could create sonnet agents. I'm not sure some of these other platforms are that sophisticated when it comes to their sub-agent sort of specifications, which means you've got to be careful with what you're doing and what you're doing with things. Um [11:58] Matt: They are. They they are capable. And it it just can't it's kind of like an edge case. It's kind of like any system that we build and we're like, oh, we didn't think about that. The y the user pressed that button and then that button in that order and that caused [12:08] Dave Pengelley: But but [12:09] Matt: something to [12:09] Dave Pengelley: but [12:09] Matt: break. [12:09] Dave Pengelley: it's way the way the harness is configured, right? Like like if the home like yes, with Hermes you can create a skill that knows to go and call a specific CLI with a certain thing, but if you're trying to just run it within the Hermes harness and create sub agents, I don't know that it's necessarily I don't think you can find different models. You might be able to, but like I think [12:26] Richard Webbe: Yeah. [12:27] Dave Pengelley: it's default to do using whatever model you. you're running as the main model in that particular profile. I'm not talking about sub profiles and using other profiles. I just mean within one profile. Um I don't know because it will create sub process agents asynchronously and stuff now. [12:43] Matt: I'm not certain either, but like this is where you could just be custom and you can you can get it to build its own school around del uh skill around delegation using external agent systems and that delegation skill could use its own CLI, you know, like spawning a Hermes agent with blah blah blah, you know, all the flags. Um, so I do that at the moment where I have my adversarial reviews use other agent CLIs with different models because it they pick each other up on different things. Um, and why would you orchestrate that manually? You know, like let the agent do its own orchestration. That that's probably my biggest win for the week is realizing leverage is necessary. And at this point, if you are currently running your own loop yourself within your own your your own loop, you're kind of you're losing a lot of potential output here. So like for me personally, I run about 10 different projects at once all the time, every day. And I've got to make sure that I understand each of those things. What's next? What are you currently working on? What are we about to ship? Do I have to review anything before we put it to production? And all of these various bits and pieces. But I'm finding more and more that every time I do one of those steps, it's the same thought process. That I could just turn into another skill loop, you know, and [14:03] Dave Pengelley: Yeah. [14:04] Matt: if there is any ambiguity escalated to me. So that's what I do now. A lot of my stuff is becoming a lot more autonomous and utilizing some fantastic control layer mentalities and tools that Blake's been working on. Yeah, things are getting very, very interesting and very exciting. [14:24] Dave Pengelley: I [14:24] Richard Webbe: That's [14:24] Dave Pengelley: think [14:24] Richard Webbe: kind of kind of like getting out of the way of your own agents. [14:28] Matt: Yeah, exactly. So, like I could give you a real like 20-second summary of what this kind of system looks like. Imagine you have a job queue and everything that gets added into this job queue is unconfirmed. So the very first step is adding something to the queue. The second step is checking that that queue is possible to get worked. You know, that that item is it confirmed? Cool. All right. Now the that the dispatcher can spin up another loop to do that task. And then once that task is listed. Is complete, you have your review loop and all the validation and back pressure steps with that, which [15:04] Richard Webbe: Yeah. [15:04] Matt: obviously can then spin up fixes until that is now green for staging or preview, at which point it could then start the waterfall of shipping. And then in this process, there could be more uh observations that get added to the queue. And then again, the cycle just keeps going. [15:23] Richard Webbe: Wow. Good [15:24] Kate Fabian - Adoptech: I [15:24] Richard Webbe: explanation. [15:25] Kate Fabian - Adoptech: love your analogies, [15:26] Richard Webbe: Yeah, [15:26] Kate Fabian - Adoptech: Matt. [15:26] Richard Webbe: yeah, very well done. [15:28] Kate Fabian - Adoptech: I like to think I'm not too tired. Technical, but I fully understood all of that. [15:32] Richard Webbe: Exnt [15:32] Kate Fabian - Adoptech: Thank you. [15:32] Richard Webbe: it weird. I [15:34] Kate Fabian - Adoptech: He's [15:34] Richard Webbe: can't [15:34] Kate Fabian - Adoptech: got a [15:34] Richard Webbe: know [15:34] Kate Fabian - Adoptech: magical [15:34] Richard Webbe: what he [15:35] Kate Fabian - Adoptech: way [15:35] Richard Webbe: said. [15:35] Kate Fabian - Adoptech: of doing it. [15:37] Matt: Well, I mean, if you can't this is my this is my um ethos that I've started to take on now because I want to try and get more into speaking and presenting and um just being able to educate as well. So if you can't explain something in the least amount of words possible and have the most instant comprehension, then you don't know it well enough. [16:00] Dave Pengelley: Yeah. Yeah. [16:01] Richard Webbe: Smart people explain complicated things simply. [16:05] Dave Pengelley: I think my my my biggest issue with all this sort of loop stuff at the moment has just been token limits because I'm not paying for the expensive plans. And because I've been trying to budget down my um my LLM plans, it's definitely restricting my experimentation. Because I'm super conscious that I'm going to hit limits that I run out of things much quick more quickly. Uh I experimented with some of the other cheap models like the M3 and stuff, and it wasn't up to scratch. Scratch. Even now with GPT 5.6, I've been trying to use Luna a lot, and Luna clearly has limits compared to some of the other models. [16:38] Matt: All right. [16:38] Dave Pengelley: Uh, as far as it's much cheaper, and you can use it a lot longer with compared to you know, Sol or Terra or Grok 4.5 or whatever. But there's a reason why it's cheaper because it's not as good, even though they said it's as good as 5.5, it's great, it's this, it's that. Part of me goes, Well, I'm not sure. I think 5.3 codex might. Might have been better at some of these coding tasks because it was built specifically as a coding agent versus these generalist lightweight cheaper agents that are meant to be as good but are not. So [17:10] Matt: I would look at um I would look at Luna like haiku. That's the probably the best, simplest analogy [17:16] Dave Pengelley: Yeah. [17:16] Matt: for for most people to understand is it's capable, but yeah, it's not as capable. [17:21] Dave Pengelley: But on [17:22] Matt: And [17:22] Dave Pengelley: all the rating shit sheets and all the benchmark tests, they said how amazing it was and how it was close to this other model that was almost as good as five point five. Quarter the cost and everything. So you go, okay, I'll take that on face value and I'll try and use it as a straight up 5.5 replacement. And it's not bad, but you know, I've been trying to do some complex website stuff with um cinematic scrolls and um 3D layers and all kinds of weird stuff. And it just doesn't quite get the brief sometimes. Uh and so I was having this issue with this particular thing. I was trying to get a render and running around in circles. Eventually, I was like, okay, I got my Groc 4.5 refreshed, I can go. Go back to that. And for Grok 4.5 actually nabbed this specific brief really quickly. Um, same brief that I gave Luna, Luna did a whole cycle on it and just gave me nothing. And I'm like, what have you even done? Whereas Grok 4.5, which is not perfect either, but it it worked it out much more quickly. So yeah, finding the right models and you do get what you pay for, I think. That that's ultimately part of this as well, right? Um, and if you're trying to, you know, yes, you can get things done with less tools. But sometimes you're better off just going out of Bunnings and buying buying the like talk about the chainsaw. Sometimes you've just got to go and spend the money on the chainsaw in order to get the job done efficiently. [18:39] Matt: I've got another um I've said this in the past, but like this is another workplace organizational structure type analogy where you know you have your floor worker, then you have your your uh department manager, and then you have your executive person. There's a reason why they're all paid different money most of the time. Let's not get into like the weird imbalance of fairness. Um where you you would You you would uh delegate a query or some sort of task out to the executive, and you would expect to not need to say much because they know what they're doing, they they have the experience, they have the the power, the you know, the various aspects to get the thing done. Whereas if you go to the floor task, uh the floor member and you give them the same task, you would be expected to pretty much list every single detail in order, including where to go when you get stuck. You know, I think that's the exact difference here between Sol Terra and Luna is that you can just screw it. With Sol and it will figure it out for you. Luna, you've got to tell it everything. And you're absolutely right. Like that difference between benchmarks and reality as well. Like if you actually dig into what the benchmarks are, you'll instantly realize that it's not that benchmarks are fake, it's that the tasks that they're doing on the benchmarks are nowhere near what we're actually doing in real life. [19:53] Richard Webbe: No, they're designed to look good for certain tasks. [19:56] Matt: Yeah. And then you get the problem where like the the benchmarks. Benchmark uh training data is like bleeding into the models as well. So when they go to do it, they're like, hang on, I know what this is. Like, and then it it kind of biases the answer in that regard. [20:11] Richard Webbe: Bit like Volkswagen with their uh emissions detection. [20:16] Matt: Yeah. [20:16] Kate Fabian - Adoptech: Going [20:16] Richard Webbe: Sorry, [20:17] Kate Fabian - Adoptech: back [20:17] Richard Webbe: Volkswagen. [20:17] Kate Fabian - Adoptech: to the uh [20:18] Richard Webbe: Sorry, did I say that out loud? Sorry, I was just thinking that. [20:21] Matt: No. [20:21] Kate Fabian - Adoptech: going back to the uh executive analogy outside of the uh outside of the models, if you're giving your your tool of choice context around Your business, then you again don't have to explain that um you know ground floor worker information when you're interacting with it. So a lot of people pick up the tools and they're giving these you know short sharp um you know prompts where you could actually build out the context prior, give it all that understanding. So when you do give it that short sharp prompt, the answers you're getting back are much more customized and accurate as opposed to not having that context built in. [21:00] Richard Webbe: Review and once you know what that review is, now let's apply the tools. I think you and I spoke about that on the phone a while ago, Kate, and that's exactly what it is. [21:08] Dave Pengelley: Yeah, the the org the org chart analogy is interesting. I mean, I'm talking to businesses now, I'm pivoting some of my uh go-to-market direction for for the business I'm operating around this concept of fractional leadership because there are businesses that are looking to you know bring on like an AI doer, someone who can hands-on, use some AI, run claude code to build a website, do other bits and pieces. And these are business owners who don't necessarily [21:31] Dave Pengelley: Know the ins and outs and they're just gonna hand like chaos to someone and hope that person thinks about it sophisticatedly enough. And there's a lot of great people that are technical and they're great at just jumping on and doing, but they're not necessarily gonna be able to translate the business drivers and and ask some of those why questions. And so uh I'm talking to sort of a few um potential clients and stuff at the moment around that sort of fractional leadership piece where bring me in as a part-timer to become that sort of that translation bridge. So that way you don't have sort of business leaders and operators trying to talk in technical terms to very technical people. They can talk about the business outcomes and the drives and what they're trying to do. And then my job is then to try and translate that stuff through to the operations team. And then I'm accountable for the deliverable of what's getting done by the technical people. Um, but I can sort of translate between the business and the technical and back and forth. That way the business operators can continue running and leading their business and they're not sort of get stuck in the weeds of like why Did you code that thing, or what that what's that thing, or why why is that agent costing me so much money? And I've so I can sort of blend that strategic and technical lens together in that sort of fractional leadership capacity. So I think that's something that is interesting, uh especially with these agents, and and you know, it translates all this stuff's fractal. [22:48] Matt: Hmm. [22:49] Dave Pengelley: Um [22:49] Matt: That's [22:50] Dave Pengelley: you [22:50] Matt: super [22:50] Dave Pengelley: zoom [22:50] Matt: cool. [22:50] Dave Pengelley: in. [22:51] Matt: That kind of like it's so it's similar to what we've said in the past. It's cool that you get to sort of start like exploring that and embodying that a bit more. It's similar to what Kate said before as well. Like You know, most of the time these companies don't have an AI problem or a lack of AI problem. They've got an unclear work problem. [23:07] Richard Webbe: Yeah, they don't know what the problem they don't know the problem they're trying to solve. [23:11] Dave Pengelley: Yeah. [23:12] Richard Webbe: They [23:12] Dave Pengelley: And [23:12] Richard Webbe: don't know it yet. [23:13] Dave Pengelley: I I find myself guilt guilt, and I've discussed this and been very candidly transparent on this show as well. Around sometimes I get so excited about trying to do something, I don't necessarily do all the planning, and I become a little bit spoiled by these magic black boxes. And I just expect the AI to be able to do it, or like you're in my business folder, you've got all my context just you should know me by now and you know and then I give it a Luna and I'm like why why couldn't Luna do this? Um and so it forces me to go back to the drawing board and go okay hang on hang on it's not that magic it doesn't have that ambient context that I always talk around it doesn't know all the things in my head and just because I have an understanding and I can visualize in my head if I haven't properly translated that through or I'm just you know that that that AI slopness where I just don't take enough control over the copy and the narrative of what's being written I just expect To kind of work it out because it kind of knows things about me, I'm not going to get the best outcome. So I've been forced to slow down, reset, go, okay, that was great. Prototype that stash that. Don't use that. We'll go back to where we started and let's break it down. Give me the story. Let me read it through. Verify that. Okay, now give me the next steps. And really, you know, the stuff that you and Matt talk about, that Blake talks about around that whole planning should be 80% of the process. You know, I'm was always the guy that never wanted to write pseudocode. I just want to jump straight in. And I think in this era, you've really got to spend the time on the planning if you want the best result, if it's something that's sort of a bit more meaningful. And if you don't have a clear view of what it is you actually want, how's the AI? You might get lucky and the AI can inspire you and get you off the blank page. But it's important to know what you're actually aiming at or because the AI is not magic. [24:54] Richard Webbe: I did notice that with uh Fable the other day when I was asking to do some inappropriate tasks, like me asking my ten year old. To stack the dishwasher and expecting it to work. And um, and it did actually come back with some context for me, which I enjoyed. And it said to me, Richard, looks like you're trying to do this. I think you're starting at the wrong point. Can you go over here? I really enjoyed that. I think we all expect that, but of course, our expectations are always a long way further down the track than the reality because Hollywood introduced us usually to all this stuff first, or some writer or some media person, and then we expect it to be a lot more interesting. Hey, um, just to change. Tack for a second. Um, Facebook and Meta and a whole lot of people are now offering sovereign AI data centers. [25:38] Dave Pengelley: That that sounds like that sounds like news, Richard. We're still talking about wins of the week and how [25:42] Richard Webbe: Oh [25:42] Dave Pengelley: we use this [25:42] Richard Webbe: sorry. [25:42] Dave Pengelley: stuff. You're killing me. You're killing me. [25:44] Richard Webbe: Sorry, [25:44] Dave Pengelley: Jumping [25:44] Richard Webbe: I'll shut [25:45] Dave Pengelley: out of [25:45] Richard Webbe: up now. [25:45] Dave Pengelley: order. What is that? [25:47] Richard Webbe: I do [25:47] Dave Pengelley: Um [25:48] Richard Webbe: that every week. [25:48] Dave Pengelley: yeah, [25:49] Richard Webbe: Sorry. [25:49] Dave Pengelley: I mean, I I uh I'll I'll finish on on on you know humble brag win. It's not fully win yet. I'm not there yet. Uh I've been rebuilding a website for myself and trying to make it look not like all other AI Slop websites and this is where I've been fighting with Luna and Grok and everyone else. So this is this is incomplete. This is still running on my on my local host dev server. I've put a draft versions online, but I I just want to share because I think it looks really cool. Um what I'm doing on this site where I've got the whole scroll thing going on with these [26:19] Richard Webbe: Oh, [26:19] Dave Pengelley: little floating [26:19] Richard Webbe: I like [26:20] Dave Pengelley: shapes [26:20] Richard Webbe: that. [26:20] Dave Pengelley: coming [26:20] Richard Webbe: Oh, [26:21] Dave Pengelley: in. [26:21] Richard Webbe: that is cool. [26:22] Kate Fabian - Adoptech: Very flesh. [26:23] Dave Pengelley: Yeah something something a little bit different than your normal website. So I'm still working out on what the what these shapes should be to sort of match the story and talk And line up with the card. Um but yeah, I think you know, having having the logo there in the floating form, pretty cool what you can do with this sort of stuff now. Like if you have a vision for what you want, you can do so much, but you've got to have some idea of what the vision should be. Um, if you didn't know you could want to even do this, then you're just gonna get the same website as everyone else, right? [26:53] Richard Webbe: Well I [26:54] Dave Pengelley: So [26:54] Richard Webbe: loved your phrase back there, David, what we were just talking about. Without uh strategy to say without any AI becomes theater. I like [27:01] Dave Pengelley: Uh [27:02] Richard Webbe: that. [27:02] Dave Pengelley: that one. [27:04] Richard Webbe: That one, that's a [27:04] Dave Pengelley: Yeah, [27:05] Richard Webbe: good phrase. [27:05] Dave Pengelley: yeah. [27:05] Richard Webbe: And I like how the 3D perspective, like you're walking through a story or you're walking through a program and [27:11] Dave Pengelley: Yeah. [27:11] Richard Webbe: the different issues appear. It's very well done. [27:13] Dave Pengelley: And [27:13] Richard Webbe: I [27:14] Dave Pengelley: and [27:14] Richard Webbe: like [27:14] Dave Pengelley: and [27:14] Richard Webbe: that. [27:14] Dave Pengelley: you know, I'm I'm I'm an eighties kid, so that whole tron grids kind of I [27:20] Richard Webbe: I [27:20] Dave Pengelley: I [27:20] Richard Webbe: was [27:20] Dave Pengelley: avoided [27:20] Richard Webbe: expecting [27:20] Dave Pengelley: going [27:21] Richard Webbe: Jeff [27:21] Dave Pengelley: full [27:21] Richard Webbe: Bridges [27:21] Dave Pengelley: neon [27:21] Richard Webbe: to [27:21] Dave Pengelley: with [27:22] Richard Webbe: go [27:22] Dave Pengelley: it. [27:22] Richard Webbe: past on his motorbike. [27:23] Dave Pengelley: Yeah. [27:25] Dave Pengelley: Um but [27:26] Kate Fabian - Adoptech: Very [27:26] Dave Pengelley: yeah, all [27:27] Kate Fabian - Adoptech: cool. [27:27] Dave Pengelley: this all this sort of stuff's out there, and but you know that That's you know, uh a hundred iterations to get there, which probably should have been 10 iterations if I'd slow down. But I, you know, [27:38] Kate Fabian - Adoptech: Yeah, it [27:38] Dave Pengelley: could [27:38] Kate Fabian - Adoptech: comes back [27:39] Dave Pengelley: get [27:39] Kate Fabian - Adoptech: to [27:39] Dave Pengelley: ex [27:39] Kate Fabian - Adoptech: capturing that [27:40] Dave Pengelley: get [27:40] Kate Fabian - Adoptech: that's [27:40] Dave Pengelley: excited and want the magic black box to do all the work. Um but I again that's one of those things. If I'd given that to like Soul and had the credits to burn on Seul or Fable, I might have got there in a lot less iterations too, by going through things like Luna and looping and looping and looping, and it doesn't get it. And [27:58] Kate Fabian - Adoptech: a good idea. [27:58] Dave Pengelley: it, you know, it takes. 20 minutes to try and work out how to do the geometry and I'd and different versions of and some some of the the what I've settled there for the look and feel is just because I'm hitting limitations where I can't get it to do what I think I want it to do. Like I had one point where I was saying, What if I had like little drones that are flying across the surface? And then as you get to each card, they form up to make the constellation and join the lines together, and then you know, kick that sort of sticky pins for a little while, and then as you keep scrolling, then they break back down into a drone fleet and go fly to the next one. I couldn't Luna. At best, I got like three little dots that kind of rose up and then went down a little again. Like I couldn't could describe it to the point where I could make it do it. [28:37] Richard Webbe: That's a [28:38] Dave Pengelley: Um [28:38] Richard Webbe: key point, isn't it? Description. [28:40] Dave Pengelley: yeah, [28:40] Richard Webbe: Like we're saying [28:40] Dave Pengelley: and [28:41] Richard Webbe: in that context. [28:42] Dave Pengelley: and just burning tokens, burning tokens to the point where I'm like, I need to get a result or I'm gonna be locked out of my entire AI systems for five days waiting for resets or spend more money. So I do miss my Claude 5X plan. I do. Hermes is fun and does some great things, and being able to swap in different models in an experiment is Good, but I also do miss just having one really good set of models that just worked all the time for me. Um, even if I was paying for it. [29:07] Richard Webbe: Has [29:07] Dave Pengelley: So [29:07] Richard Webbe: anyone come up with a token exchange app? [29:11] Matt: What [29:11] Dave Pengelley: meaning [29:12] Matt: do you mean? [29:13] Richard Webbe: Just trading token credits between organizations and balancing out their needs rather than just with the vendor. Just an idea. [29:21] Kate Fabian - Adoptech: Sounds like a cool idea. [29:22] Richard Webbe: Someone's vibe coding right now, and I've given up the idea. [29:25] Matt: I don't understand. What do you what do you [29:27] Richard Webbe: So [29:28] Matt: what do [29:28] Dave Pengelley: I'm saying. that's what [29:28] Matt: you mean? [29:28] Richard Webbe: Dave's in Sydney, I'm in Melbourne, you're in, you know, Orby Radonga, Kate's up there, we're all working on different projects, and someone says, gee, I need an extra hundred tokens quickly. Right? [29:40] Dave Pengelley: Oh, you mean like like like she like saying like you've got an extra seventy percent of your five hour usage window that you haven't used or whatever, and and I need something, so I could sort of sublet [29:52] Kate Fabian - Adoptech: Sort [29:52] Richard Webbe: An [29:52] Kate Fabian - Adoptech: of [29:52] Dave Pengelley: your credits. [29:52] Richard Webbe: Uber app [29:53] Kate Fabian - Adoptech: yeah, [29:53] Richard Webbe: for tokens for the Uber [29:54] Kate Fabian - Adoptech: sort of [29:55] Richard Webbe: generation. [29:55] Kate Fabian - Adoptech: like an enterprise plan, but not. [29:57] Richard Webbe: Yeah. Yeah. Because someone's always got You know, there's always room I find in our world to even out efficiency and and uh um usage, fail false usage, right? I always think of those guys, I forget what the name, is it uh late table or something like that, and they get the leftover food from restaurants that's before it goes off and they repurpose it for [30:19] Dave Pengelley: Hm. [30:19] Richard Webbe: charities, other events, other things. I think there could be an exchange for tokens that might be worthwhile [30:24] Dave Pengelley: I think I think [30:24] Richard Webbe: using. [30:24] Dave Pengelley: it breaches a bunch of terms of service from a lot of these companies and 'cause again they don't want you maxing out your tokens on On all your subs because they lose money on that. Um [30:32] Kate Fabian - Adoptech: Yeah, [30:33] Dave Pengelley: it's [30:33] Richard Webbe: I [30:33] Kate Fabian - Adoptech: they [30:33] Richard Webbe: wonder [30:33] Kate Fabian - Adoptech: want [30:33] Dave Pengelley: it's [30:33] Kate Fabian - Adoptech: you [30:33] Richard Webbe: how much. [30:34] Kate Fabian - Adoptech: to I was just gonna say they want you to run out and then pay you extra API usage [30:39] Dave Pengelley: well [30:39] Kate Fabian - Adoptech: on [30:39] Dave Pengelley: yeah, [30:39] Kate Fabian - Adoptech: top. [30:39] Dave Pengelley: I mean on Sunday when uh the the the draft of that side that doesn't have all the little geometric shapes and stuff and and different copy on the cards, but it's got the little the scrolling plane thing. Um I wanted to get something out by the end of the weekend and I hit Sunday night and my GPT five-hour window was was was slammed and I couldn't actually use any more GPT and my Grok window was full. slammed and I was like, they're my main two ones that I'm paying for. And I'm like, oh sorry. I um [31:07] Dave Pengelley: I'm like, I don't, I don't this again, this comes down to the the uh the the technical enough but I'm like I don't know all the different git commands to make sure I'm gonna capture the right differences to just commit the right changes to push them up. Like I can do some basic git stuff but I just rely on the AI to do all that for me, right? Because I'm just like okay git push git push and and commit and so on and sync. And then I'm like ah I really want to push these Changes up and not because I've got sub-repos and stuff anyway. So I was like, ah, I think Noose Research, they said they're using they've got a free access to Step Fund 3.7 or something at the moment. So I was like, I'm just gonna use that free step 3.7. I Googled it and went, Oh yeah, it knows about coding stuff. Brilliant. All right, step fund commit my repo. And it did. And I was like, ah, phew. I got it. But I ran out of AI, and I and I and I think as businesses, if they're moving into having operating Operational critical systems running on AI tokens and they're running on sub plans, or they're going to hit usage limits, or they're going to have have cost caps. Where even an open router, you're right, you sort of go, Yeah, we don't want to spend more than you know $500 a month or a week or whatever on these AI tools. What happens to your business critical system? We discussed this last week when all of a sudden all your AI stops. Like, how long can you afford to be offline for [32:23] Richard Webbe: I think there's bill shock, to be honest, happening all over the world with organizations and how many tokens you're using and stuff. like that. You know, I used to think about uh I was consulting on VMware to one of the uh big four banks and uh we ran an audit on their virtual machines that they because you know you could set it up with a credit card and um kind of like Amazon got in with their setting up their cloud components with a credit card. And I think one of the banks had two and a half thousand virtual machines they were paying for that they didn't know about. And so I imagine those subscription apps that are checking our credit cards and that telling us oh you know you went for subscription on this service us for the last two years and you've never used it, there's got to be room for efficiency, AI efficiency uh tools over the top of what we're using to help us manage those resources. [33:10] Dave Pengelley: Well and and and as you know Richard, we we went and uh spoke to a uh a potential client for some of our syllogism workshops together and they're deploying various AI tools through their business, but they don't know if they're getting any value from it. They they ambiently think, yeah, sure, like people are happy. I think we're seeing some productivity from it, but they got no benchmarks, they've got no measures, they they They have no idea whether it's worth the money they're spending on the AI tools they're providing for their staff. [33:35] Richard Webbe: That's the space I love, I love to live in. I'm I was doing some work in London funny out for Cisco and British Telecom, and I did a financial analysis on client server. So, what's the story? Should I have remote servers doing the processing for me, or should I give more power to everyone on their laptop? Because it's less less cost doing it there. And it was really just the cost of the processing power additive to the cost. of the networking traffic and then amortized across the users to see if you get an efficient run. I love ROI and and mathematical calculations. And I think there's a big industry for us as we've got this new resource to manage it and understand where the costs go. [34:14] Kate Fabian - Adoptech: Yeah, [34:14] Matt: One of [34:14] Kate Fabian - Adoptech: I [34:15] Matt: the [34:15] Kate Fabian - Adoptech: think [34:15] Matt: big [34:15] Kate Fabian - Adoptech: it comes sorry, don't you go? [34:17] Matt: You go okay, yeah, you go right away here. [34:20] Kate Fabian - Adoptech: Sorry, I was just gonna say one of the um one of the uh it comes back to what we're talking about before is um you know making sure that you are tracking before You start measuring that that benchmark, having the ROI, being able to measure that. If it's not make you know saving you time, saving you money, um, or making you money, then you know it's not worth doing. So having a register of what you've got implemented, what's it for, what's the benchmark, and being able to track all that so that over time um you can you can see where the value is, and obviously, you know, that comes back to Dave what you were saying about this fractional piece is it's not something you just implement, you turn on, you teach people how to. and then you walk away. It's like a constant revision that you're gonna have to to keep up with to keep checking on this and and as things change, you know, are we adding more, are we changing models or do we just stick to you I don't think you can just turn something on and walk away and just leave it. It's just gonna be this ongoing thing that businesses have to account for um [35:20] Richard Webbe: And [35:20] Kate Fabian - Adoptech: moving [35:20] Richard Webbe: and I [35:20] Kate Fabian - Adoptech: forward. [35:21] Richard Webbe: I you I know you and I spoke about your tradies are quite smart, aren't they, Kate? They know their business really well. I think you and I had a chat about that. And that helps you go, oh good. You know your business, here's what we're going to add value. One of the uh funniest things, I'm going to slag another big industry here. Being a sales leader and a uh sales manager and an entrepreneur, the thing I find most frustrating is the introduction of CRMs into the world. Apologies to Dave with his long-term experience with Salesforce and other CRMs. But what I noticed was a CRM was a great idea to keep the intellectual property and the knowledge of the relationship between the client and the company. Everyone goes, that it's like it's a no-brainer. I don't need to do an ROI on that. We keep our clients going. And then, of course, when the ERP, when the CRM system comes along, the accountants get a hold of it and they turn it into an ERP system. And then all of a sudden, you've got your salespeople entering in thousands of points of data on behalf of the accounting system, not the sales business, and they're not getting much selling done. And everyone wonders why revenue goes down. [36:24] Dave Pengelley: No, I'm [36:25] Richard Webbe: And [36:25] Dave Pengelley: I'm I I'm gonna fire you on this one, [36:26] Richard Webbe: I've [36:26] Dave Pengelley: Richard. [36:27] Richard Webbe: seen that countless times. [36:30] Richard Webbe: But I have I've seen it so many times and [36:32] Dave Pengelley: So [36:32] Richard Webbe: I've experienced it myself. I go into [36:33] Dave Pengelley: that's [36:33] Richard Webbe: a company and I go, I've got to see this [36:35] Dave Pengelley: that's [36:35] Richard Webbe: client. Yeah, [36:35] Dave Pengelley: that's [36:35] Richard Webbe: Richard, [36:35] Dave Pengelley: just [36:36] Richard Webbe: go and sit there for two weeks and fill out the forms [36:38] Dave Pengelley: that's [36:38] Richard Webbe: for us, [36:38] Dave Pengelley: that's [36:38] Richard Webbe: would [36:38] Dave Pengelley: just [36:39] Richard Webbe: you? [36:39] Dave Pengelley: bad bad systems, um and bad processes. You can't blame the CRM for that. [36:43] Richard Webbe: But [36:43] Dave Pengelley: I mean [36:43] Richard Webbe: that's [36:44] Dave Pengelley: a good [36:44] Richard Webbe: what I'm [36:44] Dave Pengelley: c [36:44] Richard Webbe: saying. [36:44] Dave Pengelley: a good [36:44] Richard Webbe: I'm not [36:45] Dave Pengelley: CRM [36:45] Richard Webbe: blaming the [36:45] Dave Pengelley: impl [36:45] Richard Webbe: CRM, [36:46] Dave Pengelley: implementation. [36:46] Richard Webbe: I'm blaming the accountants who get a hold of [36:49] Dave Pengelley: Well, [36:49] Richard Webbe: the [36:49] Dave Pengelley: but [36:49] Richard Webbe: process. [36:49] Dave Pengelley: it's not just accountants, it g it can be sales managers, it can be anyone. Like salespeople uh can be their own worst enemy, especially once they get into management. Uh sorry, any sales leaders that are watching this. But uh, like, [37:03] Dave Pengelley: and the thing is, like, top operators are always gonna be top operators, and the reason like why your top performing people are always gonna be top performing is because they know what they're doing, they've either got really great memories and they're just naturally gifted, or they've got really good systems in place and they they know how to keep on top of their work list, they know how to get the next thing done, they follow up with customers, they do all those sorts of things. That's just what they do, they're wired that way. Your bottom performers are probably always gonna be bottom performers for. Whatever reason, they're just not wired for performance. They don't care. They don't have the right intrinsic motivation because intrinsic, extrinsic sort of um things can only get you so far. They don't have the systems and so on. So they're always going to struggle. But a good system like a CRM that gives you follow-up reporting and reminders and visibility, the idea is it lifts that, especially that middle percentage up closer to the higher performers because all of a sudden they're going to get that visibility and reminders. It builds in some of those systems and processes. gives them those leading indicators on their dashboard when they look at their pipeline report they go, oh I do need to follow up that customer because it's been three days since I heard from or oh I did send five quotes. I forgot I'd sent that five fifth quote and I didn't follow it [38:13] Richard Webbe: Dave, [38:13] Dave Pengelley: up. [38:13] Richard Webbe: I feel like we're having our first fight, but we'll get past it. But I think you're 100% right, 100% correct. But that's if the system's implemented properly and it represents the reaction, the relationship, the sales cycle. The process, everyone doing things. In large companies, did did [38:35] Richard Webbe: in large companies, I see so many of the wrong people taking control of the sales process to make their life easier and not benefit the customer or the company. And it's a disaster. And I saw, I just got sent an email the other day, and they're implementing AI on top of their sales process, right? And then the sales manager sends a note around: this is taking too long to get your quotes out. And of course, The person who worked there pointed out, yes, it's all taking longer because they've introduced more steps so that it works with the age. It's just insanity. [39:06] Dave Pengelley: Yeah, look, I I don't argue there's there's very lots of very bad implementations. Uh and it's either because you know internally they don't actually use it. We used to say if it's not in the system, it doesn't exist. So if you've got your sales manager who's coming from another industry and they're downloading spreadsheets once a week for the pipeline review instead of just loading up the live dashboard, you that's never gonna work, right? If if they're not holding anyway. Um, but the other thing is. Is the other thing is, and this is true across the board, and we've got to be really careful around this in the AI implementation space. And there's [39:35] Richard Webbe: Absolutely. [39:35] Dave Pengelley: going to be a lot of operating and I mentioned the fractional thing, and that comes back into play accidentally in this conversation from the point of view that you I would see a lot of businesses who would say our implementation isn't working very well, we're not happy with it, and so on. And you look at what their implementation partner, the third-party service provider, that was coming in to do it, they were just doing exactly what the customer told. We talked about having opinionated AIs last week. You need, as an implementation person, to have an opinion on how something should be implemented. And so the worst implementations are when they would, they just did as they were told. And they didn't go, hang on, we're actually the experts on implementing CRMs and systems. We actually understand like data modeling and we think we think about all these sorts of things. So when you say that you want 500 tick boxes on that form, we don't think you should do that. We have a different opinion. Maybe you should slimline that down, maybe you should have different Different versions of the form for different different sales processes and for different teams. But if you don't have that opinionated expert opinion, opinion that will guide the person on getting a better result, they won't get a good result, no matter [40:43] Richard Webbe: Right, [40:43] Dave Pengelley: what the [40:44] Richard Webbe: nice, [40:44] Dave Pengelley: system is. [40:44] Richard Webbe: nice, [40:44] Dave Pengelley: And [40:44] Richard Webbe: right? [40:44] Dave Pengelley: in the world of AI, that's where people like us come into play, where if we just do what the customer says and just slop it out, then why even bother hiring us or or anyone else? We need to go in there with opinions on better practice. So that way they get better results and challenge them on what is the value of that? Where is the ROI? How is that, like you said, Kate, gonna make money, save money or save time? [41:10] Richard Webbe: Yep, look, Dave, you're 100% right. And I'll drop it now, but just [41:13] Kate Fabian - Adoptech: I [41:13] Richard Webbe: quietly. [41:13] Kate Fabian - Adoptech: feel like you're [41:14] Richard Webbe: When [41:14] Kate Fabian - Adoptech: agreeing. [41:14] Richard Webbe: you're dealing with the accountants who are paying for the CRM software, they have a big say and they want to make their life easier. And they all hate salespeople. So they make the salespeople do all the administrative tasks. Getting a salesperson to do admin is the dumbest. thing in history. Getting a salesperson to spend time with their customers is the best thing. But like you said, that's where they lose their objectivity. I'll shut up now. I was defending the salespeople then. [41:40] Dave Pengelley: I know, but I mean admin's part of life. You gotta put something somewhere. [41:44] Richard Webbe: Really? Damn it. Okay. [41:46] Dave Pengelley: And [41:47] Richard Webbe: Sorry, [41:47] Dave Pengelley: look [41:47] Richard Webbe: Kate, [41:47] Dave Pengelley: and [41:47] Richard Webbe: you [41:48] Dave Pengelley: Richard, [41:48] Richard Webbe: had something [41:48] Dave Pengelley: no [41:48] Richard Webbe: to [41:48] Dave Pengelley: no doubt [41:48] Richard Webbe: say. [41:48] Dave Pengelley: you were in that top tier with the excellent memory and you just remembered everything and you didn't need to write anything down. Unfortunately, many of your sales peers are not in that caliber and they do need to write things down. And if you put it in a structured system, Then they can get reminders and workflows and automations. Isn't that what we talk about? Workflows and automations? Uh [42:05] Richard Webbe: Yes. [42:05] Dave Pengelley: in order to help them have better outcomes and be more productive than otherwise. So [42:10] Richard Webbe: Yes. [42:10] Dave Pengelley: that's [42:11] Richard Webbe: Kate, you had an opinion for a second there. [42:14] Kate Fabian - Adoptech: Uh I was just saying it sounds like you were going through agreeing, not arguing. [42:19] Richard Webbe: We are a bit. It's a bit of street theater. Goes well with the podcast. [42:24] Kate Fabian - Adoptech: Yeah. [42:25] Dave Pengelley: same the thing. [42:25] Kate Fabian - Adoptech: I think it just comes back to that whole that whole planning stage which is what we talked about thing. There's no point adding AI to a broken process, it's just going to be a faster broken process. So figure out, you know, the like you say, the problem that that you need solved. Um, you know, quite often the businesses don't know the problem that they need to be solved, which is why that analysis up front is so important. Um, as an example, there's a a client that's entering in like an onboarding of a of a uh employee into three different systems in an Excel sheet. So you think, oh, that's not much, but if you've got you know X number of projects a year times you know how many people you're employing. The the figure [43:05] Matt: I [43:05] Kate Fabian - Adoptech: that [43:05] Matt: think [43:05] Kate Fabian - Adoptech: we came to was [43:06] Matt: it was [43:06] Kate Fabian - Adoptech: 40 hours a year. So it doesn't [43:08] Matt: too [43:08] Kate Fabian - Adoptech: sound like much [43:09] Matt: large [43:09] Kate Fabian - Adoptech: but that's a [43:10] Matt: and [43:10] Kate Fabian - Adoptech: week [43:10] Matt: if you [43:10] Kate Fabian - Adoptech: of [43:10] Matt: want to [43:11] Kate Fabian - Adoptech: time of copying and [43:12] Richard Webbe: Yeah, [43:12] Kate Fabian - Adoptech: pasting. [43:12] Richard Webbe: yeah. [43:12] Kate Fabian - Adoptech: You know so until [43:14] Matt: put this [43:14] Kate Fabian - Adoptech: you do that analysis and actually figure out where the the time waste and the the manual work is happening there's [43:20] Matt: in the world, [43:21] Kate Fabian - Adoptech: you're not going to get any decent [43:22] Matt: those [43:22] Kate Fabian - Adoptech: leverage [43:23] Matt: are [43:23] Kate Fabian - Adoptech: from from AI just by slapping it on a process. You've got to actually map it out and figure out where those ones are. [43:30] Richard Webbe: Time cost of money. [43:33] Richard Webbe: Matt, you've got something on your mind. Oh no, he's kicked us off. And your cost of inaction. [43:39] Dave Pengelley: Yeah, [43:39] Richard Webbe: I like [43:40] Dave Pengelley: just [43:40] Richard Webbe: the [43:40] Dave Pengelley: just sort of saying like like this sort of math, people don't do it. But if you're doing something twenty times a week, it takes you ten ten minutes and five times, that's like half a person. Like at fifty [43:48] Richard Webbe: money. [43:48] Dave Pengelley: bucks an hour, that's like forty-three thousand dollars a year that that thing's costing you. And if you sort of scale your team up, the numbers get really crazy. Maybe it's a fifteen minute task or twenty minute task, right? Like People don't do these simple back of the knock and maths, but the the numbers get really scary really quickly, and they're pretty simple numbers to run. [44:11] Matt: Agreed. I have so much. And I'm not gonna elaborate too much because you guys, um, you've all said it very well. And I agree with all of you. Like you, you all have very, very important takes that people need to take on. But the really, really important word, I think, across all of that is the word that Dave used a lot, which is opinion. You know, an opinion is never true. It is that's just the state of an opinion. It is never true. It is an opinion. So it is what we bring as the humans. And sure, some AI things, especially as they get smarter, will start to have opinions. But what are those opinions actually, you know, structured from? You know, it's far less parameters than what we can do as humans. It's it's why we have that kind of innate ability. Those calculators though, there's one thing that they never Ever, ever actually, John. And I've been trying to say this. Well, I'm trying to find a place to say this the whole episode so far. When when Kate was mentioning time saving, cost saving, you know, all these various things. What do we do with the extra time? You know, nobody knows this. And you ask any company, [45:20] Dave Pengelley: We burn [45:20] Matt: they'll [45:21] Dave Pengelley: more tokens, man. We burn more tokens. [45:23] Matt: Yeah. So these these companies, these accountants, these salespeople using their CRM, let's say we automate them for them. Let's say they bring in I've been in me, you know, this is what I've done for many customers and many clients so far. Come in, rip out their systems, rebuild it with the correct data and context layer, correct control layer for everything. All of a sudden the sales team's now happy, the accountants team are happy. Now they've got all this extra time. What do they do with it? [45:48] Richard Webbe: We go to the power, you know that. [45:50] Kate Fabian - Adoptech: I [45:50] Matt: Yeah, [45:50] Kate Fabian - Adoptech: think [45:51] Matt: but [45:51] Kate Fabian - Adoptech: it gives you I think it gives you a choice. You can either sell more or you know, go watch your kids' footy or you know, whatever it is. that it's important to the business or you that that the the efficiency of giving you the choice. [46:07] Matt: I [46:07] Richard Webbe: Yeah. [46:07] Matt: agree with that. And a [46:08] Richard Webbe: I [46:08] Matt: lot of [46:08] Richard Webbe: did [46:08] Matt: the people [46:09] Richard Webbe: a [46:09] Matt: are are choosing to do nothing [46:11] Richard Webbe: I [46:11] Matt: or [46:11] Richard Webbe: actually [46:11] Matt: choosing. [46:12] Richard Webbe: did a I [46:12] Kate Fabian - Adoptech: Cool. [46:13] Richard Webbe: actually did a study with uh Cisco systems many years ago and uh one of the value propositions that came out of intranets and a whole lot of things like that, and particularly in their argument for using, you know, their system or another's in their environment was they were gonna say gonna say Like a 150 headcount or 250 headcount. Um, and everyone said, Well, we can't go to companies and say this is going to help you sack 250 people. So we actually did a deep dive, particularly in the IT and networking and voice department. And we came up with a lot of other work where you could tag it with how much more money will we get back on their time, as in time cost of money, if they're doing this, rather than doing this. And I I think the the litmus test is on the That is how much intellectual property is applied to what they're doing. If it's just a process, move it on. And then give people the way to be people and be creative and solve problems rather than just pro because most of what we're doing is automation, right? At the end of the day, it's automation with intelligence. So it doesn't, those the AI gives better guardrails [47:20] Dave Pengelley: I [47:20] Richard Webbe: on [47:20] Dave Pengelley: mean [47:20] Richard Webbe: those automations. [47:20] Dave Pengelley: to to [47:21] Richard Webbe: So [47:21] Dave Pengelley: to your point and round out the CRM thing, Richard, one of the features that I don't know if it ever made a market, but my uh former employer was pitching in the early days of AI on some of the voice transcription stuff. stuff was that concept of the salesperson gets in the cab after a meeting and they they talk to their phone and say update the opportunity with acme services to close next Friday, progress it in the pipeline and change the value to 50k. Bang. And then get the AI to transcribe, translate, and then update the record for you on your behalf so you weren't doing the admin. So [47:54] Richard Webbe: So [47:54] Dave Pengelley: that comes [47:54] Richard Webbe: how we grow [47:55] Dave Pengelley: that's that's [47:56] Richard Webbe: up. [47:56] Dave Pengelley: what they were. I don't know. [47:57] Kate Fabian - Adoptech: And [47:57] Dave Pengelley: I don't know where [47:57] Kate Fabian - Adoptech: salespeople [47:58] Dave Pengelley: they landed, [47:58] Kate Fabian - Adoptech: don't [47:58] Dave Pengelley: but [47:58] Kate Fabian - Adoptech: don't have to do that. Talking about [48:00] Dave Pengelley: Right. [48:01] Richard Webbe: Well [48:01] Dave Pengelley: They've got [48:01] Richard Webbe: you've [48:02] Dave Pengelley: their [48:02] Richard Webbe: got [48:02] Dave Pengelley: virtual [48:02] Richard Webbe: those little [48:02] Dave Pengelley: EI. [48:03] Richard Webbe: you've got those little plaud things now that automatically you know transcribe everything you're saying and doing. I'd I I I just use my phone to do it all the time now to be honest. But I am [48:11] Kate Fabian - Adoptech: the [48:11] Richard Webbe: often I'm [48:12] Kate Fabian - Adoptech: one. other [48:12] Richard Webbe: a big one for time and place and context. Once you've had a meeting, your endorphins are running, you're hot, an understanding of what's going on, you're motivated. It's really good to capture everything then. So yeah, I was big on that. I would I would dictate um emails and stuff in the car on the way back to the airport or something and get it sent. Before I even hit the hit the gate. [48:32] Matt: I [48:32] Dave Pengelley: Nice, [48:32] Matt: have so [48:32] Dave Pengelley: I'm [48:32] Matt: many more things to say, and I feel [48:34] Dave Pengelley: I'm [48:34] Matt: like I never am gonna get them out, [48:36] Richard Webbe: No, [48:36] Dave Pengelley: I'm [48:36] Matt: but [48:36] Richard Webbe: Matt, get them out. We will shut up. You are much smarter than the rest of us. Go. [48:40] Matt: it's got nothing to do with smartness. This is just from my own experience in the field. You know, I'm not speaking about you know hypothetical scenarios [48:50] Dave Pengelley: not sure. [48:50] Matt: here. A lot of the sales teams that I work with that I've built them systems for, these guys are hammered. They have meetings back to back, and you know, they get off. Zoom call that went over time because the person was screwing them around with, you know, that were they gonna pay, weren't they? Or that card didn't work. Now they need to get onto their next call. They have no time to do that quick capture. So all the systems that I build, they it's done from the actual meeting [49:16] Richard Webbe: Oh [49:16] Matt: itself. [49:16] Richard Webbe: yeah, yeah. [49:17] Matt: You know, that and a lot of people don't realize this. And with that whole time-saving things and people not knowing what to do with it, I this was one particular article that I found. I think I'm not sure if Dave's got that. Andy, but the um it's it doesn't really matter so much. The it was this huge survey done with all these like improvements to industries and companies and um the yeah, it was actually the BGC one, yeah. Um, where 66% of people that were surveyed received limited to no guidance about what to do with the save time. So imagine you're uh you're a generic office worker and you're usually just, you know, you clock in, you clock out, you're hammered with your work every day. And now all of a sudden there's these improvements happening. And you're not, you're not um given this scenario that if you don't perform, you're gonna get uh kicked out. You know, AI is not replacing you per se. But we've got all these performance gains and all this time saving that you can now spend towards other things. Like there there's no strategy for when you are getting these magnitudes of performance gains, what you should be doing next. And I guarantee that. Ninety percent of people out there are not as driven by initiative as a lot of these people think that they [50:36] Richard Webbe: I'll [50:36] Matt: are. [50:37] Richard Webbe: give you three perspectives on that from experience. Number one, the more idle time we give our children to play, the more creative and intelligent they become. [50:46] Matt: That's [50:46] Richard Webbe: I think the [50:46] Matt: true. [50:46] Richard Webbe: same applies to employees. If you take back the pressure of time and give them back idle thoughts, they will be much more efficient, much happier, and much quicker performance from that point of view. And then when they do. that they'll be much more responsive for customers and for the business to get the data it needs to do what it does as it does. But I I think that's the big one is don't fill up the available space with other stuff. [51:12] Dave Pengelley: I think I think play with purpose though, not just [51:14] Richard Webbe: Yeah [51:16] Dave Pengelley: like have some direction, like use the free time to innovate RDA, just do the things you would never do otherwise. Like the old uh Google was famous for having that sort of the Friday afternoon innovation time and things like that. Like don't just go play at the the ping pong ping. In the kitchen. [51:31] Richard Webbe: I don't know. That works sometimes. I mean, I had a team, I had quite a large team in Australia, as largest telco, in a role I was running a few years ago. And um, I spoke to a few of them and they were saying, we're gonna quit. This is dreadful, we've you know, we're flogged. And so I said, That's it. Like almost like you said, I said, Friday afternoon's pizza afternoon, and we're gonna do that every two weeks. And uh at first I didn't get many turn up, all of a sudden I got all 20 or 30 turned up, and from there, our efficiency in our customer relationship just Went through the roof. I just gave them time, as you said, Dave, for guided play. And it really worked well. Everyone bonded well and stuff. So look, there's a lot of room in there, Matt, without us holding hands. If you let kids go or adults go or anyone go and say play, they'll play with the things that make their life more beneficial. [52:20] Matt: I disagree. [52:24] Richard Webbe: Well, that's good. That's good. You think they'll just idly waste the time, Matt. [52:30] Matt: No, no, I don't. I I think that they will find what they're interested in and where they get dopamine from because that's all we're driven by. [52:37] Richard Webbe: Yes. [52:37] Matt: Um I was a a kid, uh a little bit in a shorter distance away than what you were. And I can tell you that when I was a kid and I had my own time, all I did was play games because it was just where my dopamine came from. [52:53] Richard Webbe: Probably shapes who you are today. [52:55] Matt: Yeah, that Rugrats game really did uh do a lot for me. [53:00] Richard Webbe: I I was a big fan of Rugridge, just saying that. [53:03] Dave Pengelley: I think there's a but there's obviously a lot of things conflating here, and some of that social bonding and just team building time is different than necessarily innovative time and it drives different kinds of productivity. Um we are we are the winds of the week went way longer than I was expecting. I thought I thought news would would blow out, but we'll let's let's at least try and cover a few news things. [53:30] Dave Pengelley: All right, so what in the last week since we joined last time, we've talked about it a bunch already, GPT 5.6 dropped and Groc 4.5 dropped. So two major model releases came out. Um has anyone played with either of them beyond me? I played with both. [53:47] Richard Webbe: A little, not enough to make an opinion on. Except it was faster and it gave me some different answers, which I think I thought had better perspective or better context. [53:57] Matt: I probably spent about a billion tokens of The last few days. Um [54:02] Dave Pengelley: Because you're in a serious codex plan, aren't you? [54:06] Matt: yeah, I'm in a I'm in a serious relationship with a lot of my [54:09] Dave Pengelley: Yeah. [54:09] Matt: tools. [54:12] Dave Pengelley: See, I'm still on just like Chat GPD business plan. And so I get like the business user amount. I'm not even on specific codex plan. And it gets weird once you're on a business plan that it's like that or enterprise. If I want to have a specialized codex plan, I think I've got to get an extra one. And then I'm double handling different. Oh, why do they make it so hard? Anyway. Um, yeah, that thing is good. I mean, it's interesting looking at the usage and and how much I hit Lunar versus Soul, and then Sol uses so much more usage so much more quickly than Lunar. It's it's it's wild, the the usage. But again, and we saw this, I think, Matt, um, which model was it where it was showing that if you use the cheap model, um, it can use as many tokens sometimes or take longer because it's got a reason and loop that many more times than the smart. Model model that just sort of one shots [55:01] Matt: Probably [55:01] Dave Pengelley: it. [55:01] Matt: GLM 5.2. [55:04] Dave Pengelley: Yeah. [55:04] Matt: It um, you know, put it benchmarks high. And yeah, it will just be as expensive because even though it's a cheaper model because it reasons longer. Um, but yeah, look, my advice to everyone as far like unless you're paying for the higher tiers for these tools or you have the budget to have API costs for the big tools, um, in the GPT land, just stick to Terra. I like Luna's not for us. Like if you think about that statement, Lunar is not for us. Lunar is meant to be sub agents. It's meant to be for the other agents to orchestrate with. And um, but yeah, the I think honestly, like Seoul is ridiculous. Like the whole 5.6 release has been absolutely insane, but I think everyone was kind of expecting that. Um, the the real, I reckon the real story is um is grok. Like 4.5 is actually insane. Um, it's no, it's not as good as Seoul, but for the difference between 4.5, Point two to four point five, it's like ridiculous. And then if you look at that as a uh extrapolation, you know, the trend graph. Um since SpaceX AI have done their partnership with Cursor, you know, 4.5 gets a little bit of a hint of that, but the real release, I don't know if it's gonna be another incremental or if it's gonna be Groc 5, the real release is gonna be nuts, and yeah, uh if hopefully anthropic slash um open AI have got an Another another increment out before then because otherwise there's a very high chance that that Grok and Cursor will will take the lead. [56:38] Richard Webbe: We're seeing a lot of that uh lipfrog technology capability. I know there's a bit of um a lawsuit kicking up between uh Apple and OpenAI was in the news in the last week or so about stealing libraries and definitions and things like that. Um it's gonna be a a race uh on and on. What's happening on the pricing of all these things? Is is Anyone discounting to compete? I was interested about that, and I wanted to ask you guys. [57:04] Matt: It's very complex. And it's all based on demand, like every economy. So while they don't have enough compute for everything on slower old hardware, yeah, we're gonna be we're gonna be paying. And then as they increase their actual compute capacity, faster hardware, better technology, less power usage, that's gonna obviously give us savings. And you can see that in the improvements that the labs are making in their efficiency. Of their models as well, because [57:31] Richard Webbe: Yeah. [57:31] Matt: there's a lot in there to do with how the power is actually used. You know, [57:36] Richard Webbe: Yeah. [57:37] Matt: you could look at it as like a mechanical engineering situation, like the duty cycle of the whole system. [57:42] Richard Webbe: Yeah. [57:42] Matt: Um, it's getting more efficient in general. So, yeah, stuff's getting more powerful. So, yeah, we're going to start paying more tokens and use more power for the individual compute elements, but you'll see more efficient gains proportionate to that as well. [57:57] Richard Webbe: And it's interesting in that same I've noticed [58:01] Dave Pengelley: Oh, [58:01] Richard Webbe: a lot of now socially orientated groups starting to [58:06] Dave Pengelley: that's [58:06] Richard Webbe: complain [58:07] Dave Pengelley: got [58:07] Richard Webbe: about [58:07] Dave Pengelley: the record. [58:07] Richard Webbe: data centers like it's illegal chemical dumping. Do you know what I'm saying? They're anti data center movement around the world, which, in my opinion, if you did a um a holistic view of data centers, thus saving resources everywhere because of how efficiently they're treating things and doing things. But the people who don't understand Stand up. No more data. There's people with placards in parts of the US, no data centers, no [58:35] Matt: Yeah. [58:35] Richard Webbe: contents. Like, what are they talking about? It's like no more hospitals. What is wrong with you? [58:40] Matt: Yeah. I I can only I can only attempt to speak respectfully about those people, but uh I I believe it's a context issue. Again, you know, it's like we always talk about it, they don't actually truly know and it's just an opinion. [58:53] Dave Pengelley: Well [58:53] Richard Webbe: Education. [58:53] Dave Pengelley: and and I think [58:54] Richard Webbe: Education. [58:55] Dave Pengelley: I think that there's uh there are some short term issues with the power costs and things where these Guys are going in and they haven't created their own power plants yet and they are pulling a lot of power. And so the grid is getting more expensive in those areas. And so there's like concerns around this thing's tripling my power price. And Australians know a little bit about rising power prices and the stress that can cause. So we can sympathize with some of those things and zoning issues and noise potential things. Like I have any development issue, right? Where where I've I've got this beautiful landscape behind me, and now it's going to be filled up with these giant heat sinks and fans and noise and other bits and pieces. That people concern about. So again, some of it is inf in information, and there's a lot of misinformation and disinformation out there and scaremongering around the negative things of all this sort of stuff versus the positives. And there isn't enough necessarily necessarily uh editorial balance out there. But I also think you know, sometimes there are some legitimate concerns around um industrializing some of these remote communities that they weren't ready for, and they just yeah. [59:55] Richard Webbe: Yes, there are. And on that basis, since in Opinion, AI is driven by what you just said, Dave, power. You want to, if we want to really automate the world and make great leaps in health and medicine and support and people and resources, the more AI, I think, is better. So we need more power. And there was a young man, I've forgotten his name, you can look it up, but he can build a nuclear reactor in 296 days. And he's going to IPO in a little while. And he's it's his own startup, and I think he's in his 20s, and he's building micro-nuclear. Reactors that are very safe, and where people say it takes 10 years to build one, he said he can do one. I think it was in 296 days. Isn't that astoundingly great? Because as you were saying, Matt, need drives change. Demand drives change. [01:00:48] Richard Webbe: And on that same vein in the news was, of course, the launch of this sovereign AI as a service is coming out from Meta and a number of other organizations. So you'll have your own private RAG LLM running on your own private cloud version of that, which is specific to that rather than your general data processing angle. And I found that really interesting as well because it's like we were saying, the cost model is shifting to where the most efficient way to do things is. [01:01:19] Dave Pengelley: It's uh it's interesting if it's hosted in a cloud somewhere, how sovereign is it? Like the whole point of sovereignty is it's like within your borders and you have a lot more control over it. [01:01:26] Richard Webbe: Well, they [01:01:26] Dave Pengelley: So [01:01:27] Richard Webbe: have to demonstrate that. So I remember back in the day going. To the Department of Defense, selling telecommunication. And in the old days, they would only ever buy dark fiber, right? That is, we'll just buy the pipe and we're going to put our, you know, defense level security on it. Nowadays they do buy VPNs. They do buy shared infrastructure resources. And we know the biggest cost saving comes for any of anything, whether it's power, water, AI data, it comes from shared infrastructure. So the industry is on itself to prove that that sovereign Chinese walls. So to speak, that we're talking about, um, can be demonstrated to be safe and secure. And I think that's the challenge of the industry. [01:02:07] Matt: That's the um that you just use the word secure. That that's I believe what the actual term's gonna be. It's not off site sovereign um hosting. It's it's caught it's gonna be called secure compute, you [01:02:19] Richard Webbe: Yeah. [01:02:19] Matt: know, as like a as a as a weird marketing, you know, product term. [01:02:23] Richard Webbe: Yeah. [01:02:23] Matt: Um and the the point is extremely relevant to what Dave says, like how can you um What it like how? You know, what does that mean? So I would liken it similar to what you can do right now with some of the AI inference endpoints where you can actually have zero data retention, where that particular infrastructure, you know, you send your request off. It could be overseas, it could be wherever, and there's literally no logging. Like it's it's completely zero. They do the processing for you, and then it comes out. Some [01:02:54] Richard Webbe: Yeah. [01:02:54] Matt: of the particular providers have to withhold logs for like 30 days for their own audit. Reasons or something, [01:03:00] Richard Webbe: Yep. [01:03:01] Matt: but I believe that's what it's going to look like. It's going to be basically like a VPN [01:03:05] Richard Webbe: Mm-hmm. [01:03:05] Matt: where it exists externally. We get the cost saving because of that shared infrastructure, like you said. And there's no point for the outlay of the huge hardware on your own site plus the power costs and all of the other [01:03:17] Richard Webbe: Yeah, [01:03:17] Matt: rest [01:03:18] Richard Webbe: yeah. [01:03:18] Matt: of [01:03:18] Richard Webbe: No [01:03:18] Matt: it. [01:03:18] Richard Webbe: one no one has their own power. So interesting on that point, and that's it's a ri really good point. Um uh and and I'm sure Dave will appreciate this, um, that the Office of AI has just been launched by a federal government. And I think that's a really smart thing because compliance and regulatory compliance and watchdogs will be needed to keep an eye on all of these things so that the industry and as individuals we have trust. [01:03:46] Matt: I hope they get the right people in there. [01:03:48] Richard Webbe: Well this is it. I mean I remember maybe years ago Alan Stockdale, I don't know Alan personally at all, but he became the office of multimedia in Victoria. He was the first minister to be a multimedia minister When multimedia was exploding. And uh I don't think it went anywhere because there was nothing to do really. The industry took care of it. But yeah, we need some real talent inside that department at the federal level to understand and look down the track. And as we know, governments don't always get the front-leading people, but I'm sure the industry as a whole will form a uh lobbying body to uh to keep that going. [01:04:24] Kate Fabian - Adoptech: Interesting times. Very interesting times. [01:04:31] Richard Webbe: Yeah, I'm not sure we can say the Prime Minister sees control. But it's I think it's a good initiative and and um and I think it's worthwhile. [01:04:41] Matt: As long as he's not um [01:04:45] Matt: conversating with Dario, then I think we're then we're okay. [01:04:50] Richard Webbe: Well, it's it's funny, like we're saying earlier, people want everyone to drive electric cars, but they don't want data centers. Okay, the amount of power electric cars chew up compared. data centers is unbelievably different. And so you can't demand one thing without having a balanced understanding on the other side. So having [01:05:08] Dave Pengelley: And it [01:05:08] Richard Webbe: this Sorry, [01:05:09] Dave Pengelley: Yeah. [01:05:09] Richard Webbe: go on. [01:05:10] Dave Pengelley: No, no, I was just gonna say like like it's amazing. Like a lot of this uh AI thing is uh we want data centers in Australia. Um but again we have such huge power infrastructure issues facing us over the next decade. Uh it's uh like yeah. What's going on? Uh we are we are we're hitting the hour, we are over time. Um I Remember, like as far as the two different models, Grok 4.5 and Chat GPD 5.6. So I did do a one-shot test comparing the two of them, getting them to write a game. I I gave them the exact same specification and prompt, and they both wrote a usable, playable game. Um, both were stronger in some aspects than others, but um, just from uh visuals, I thought this was the Grok one. Don't worry about what the game is, you can just see you know, there's circles, there's some little bars, and so on. Like that, that's that's a game. Uh, the chat. 5.6 sol version, a lot more polished as far as the overall UI, the look and feel, everything going on there. Um, so but different playability issues. I actually liked the look and feel of the chat one, but some of the actual mechanics of the Grok one were better. And I told chat on the second prompt, go steal some of the code from the Grok one because I liked the way it implemented some of the swarming of enemies and stuff better. So [01:06:26] Dave Pengelley: well, brown. [01:06:30] Richard Webbe: Just [01:06:30] Matt: Body. [01:06:32] Richard Webbe: if you ask [01:06:32] Dave Pengelley: Brown [01:06:32] Richard Webbe: ten pictures [01:06:33] Dave Pengelley: brown [01:06:33] Richard Webbe: on [01:06:33] Dave Pengelley: and [01:06:34] Richard Webbe: a [01:06:34] Dave Pengelley: scattered. [01:06:34] Richard Webbe: guy, [01:06:34] Dave Pengelley: That's what that's what uh Richard thinks my brain looks like. All right, well on that one [01:06:38] Matt: I've [01:06:38] Dave Pengelley: on [01:06:38] Matt: got a [01:06:38] Dave Pengelley: that [01:06:39] Matt: I've [01:06:39] Dave Pengelley: note. [01:06:39] Matt: got a good like final question that Kate could probably round us out with. [01:06:42] Kate Fabian - Adoptech: Ooh, [01:06:43] Matt: Um [01:06:43] Kate Fabian - Adoptech: putting me on the spot. [01:06:44] Matt: well this is just from your own experience. Like we we all have different [01:06:48] Richard Webbe: a bit can't get little more. you [01:06:48] Matt: opinions. There's that word again. Um for the business owner that's listening tomorrow, you know, watching the replay, um, what is the first question they should ask? Before looking into another AI tool. [01:07:04] Kate Fabian - Adoptech: Richard's favorite question, what is the problem that you're trying to solve? And so it's practically how that works is like figure out the thing that you're doing often that's wasting time, e.g. the example that I had before with copying and pasting or entering data into three different places in a spreadsheet. So there's four different places. Look at your yeah, I would say map your processes out and find if you see if you can identify those bottlenecks and if you can't and you need help with that there's obviously you know the calculator resource that Dave shared the um the National AI Center website has a brilliant bunch of resources on there to help you with mapping and and other um you know policy development that kind of thing and thirdly if you're stuck reach out to to one of us guys [01:07:52] Richard Webbe: I'm gonna reach out to Kate or D or Matt or Dave. I'm not gonna reach out to me. [01:07:57] Dave Pengelley: Um that that that's the copy. Paste is one of the simplest, easiest examples for anyone to understand. I was talking to uh to someone once, and when I mapped out the volume of copy and paste they were doing between three different systems for every order they got, it was about half a day a week that they were spending just doing that. And they're like, That's not that bad. I'm like, It's not that bad. You couldn't do something better with half a day a week on growing your business. But it's just hard to make decisions sometimes. The uh the fear of the unknown. Like uh I've said before, our biggest concern. Competitor at a place like Salesforce even was do nothing. [01:08:33] Richard Webbe: Yeah. Well, [01:08:35] Dave Pengelley: Because [01:08:35] Richard Webbe: IBM [01:08:35] Dave Pengelley: inertia's [01:08:36] Richard Webbe: share [01:08:36] Dave Pengelley: hard. [01:08:36] Richard Webbe: IBM share price collapsed uh substantially this week after they've pointed out with one of their reports the need for monolithic large vendor software is going to disappear very fast because of custom AI changes inside the organization. And [01:08:52] Matt: Or [01:08:52] Richard Webbe: I had [01:08:52] Matt: even [01:08:52] Richard Webbe: someone someone [01:08:53] Matt: just their [01:08:54] Richard Webbe: say [01:08:54] Matt: own [01:08:54] Richard Webbe: to me, [01:08:54] Matt: custom [01:08:55] Richard Webbe: sorry? [01:08:55] Matt: bespoke software. [01:08:57] Richard Webbe: Sorry, what did you say? [01:08:58] Matt: Even just their own custom Spoke software within organizations, which is [01:09:01] Richard Webbe: Yep. [01:09:02] Matt: what I spent my day doing. [01:09:04] Richard Webbe: Yep. Yep. So, you know, that whole process and how they do things and why they do it will be captured much more efficiently with the support of people like us with AI and some of these monolithic companies. I won't say their names, it would be unfair, but you'd know them. They're the probably 10 biggest software companies in the world are at severe risk. [01:09:23] Dave Pengelley: They should all spend more money on their Formula One advertising. That will fix everything. [01:09:30] Richard Webbe: Very pointed comment I pulled up. [01:09:33] Dave Pengelley: Everyone, all all the tech firms, all the modern tech firms. [01:09:35] Richard Webbe: They're [01:09:35] Dave Pengelley: You see [01:09:36] Richard Webbe: all [01:09:36] Dave Pengelley: the [01:09:36] Richard Webbe: on a [01:09:36] Dave Pengelley: logos [01:09:36] Richard Webbe: Formula [01:09:36] Dave Pengelley: for [01:09:37] Richard Webbe: One [01:09:37] Dave Pengelley: it [01:09:37] Richard Webbe: car. [01:09:37] Dave Pengelley: was it was all the crypto firms sort of five years ago, and now it's all the other tech firms sort of all getting on the uh they're plastering themselves over F1 cars. Um, [01:09:45] Richard Webbe: That [01:09:45] Dave Pengelley: not [01:09:45] Richard Webbe: is [01:09:45] Dave Pengelley: to mention [01:09:45] Richard Webbe: very [01:09:46] Dave Pengelley: the field [01:09:46] Richard Webbe: true. [01:09:46] Dave Pengelley: itself. But anyway, [01:09:47] Richard Webbe: That [01:09:47] Dave Pengelley: uh [01:09:47] Richard Webbe: is very [01:09:47] Dave Pengelley: we'll [01:09:48] Richard Webbe: true. [01:09:48] Dave Pengelley: wrap it up there. Thank you, Kate, thank you, Matt, thank you, Richard. Uh pleasure chatting with you all. Uh [01:09:53] Kate Fabian - Adoptech: Thanks [01:09:54] Dave Pengelley: always [01:09:54] Kate Fabian - Adoptech: for having [01:09:54] Dave Pengelley: educational. [01:09:54] Kate Fabian - Adoptech: me. [01:09:55] Dave Pengelley: Uh hang on the line, Kate, while we wrap up, uh, so we can say thank you and goodbye. But uh Thank you, Australia. Thank you, everyone around the world for watching the AR operators. Make sure you like and subscribe. Tell the algorithm if you want to know more about upcoming live events that we're doing, scan the QR code, that one up there, uh, to sign up to our mailing list. Uh, we won't spam you. It's just for primarily event updates at the moment, not even just weekly show stuff, just when we're doing our IRLs, which we're planning more of. So if you want to join the mailing list, it's that one there. And uh [01:10:28] Richard Webbe: What do [01:10:28] Dave Pengelley: we'll [01:10:28] Richard Webbe: you think there is? [01:10:29] Dave Pengelley: see. [01:10:30] Richard Webbe: Dave, well done. [01:10:31] Dave Pengelley: We're getting there. One one one thing at a time. [01:10:33] Richard Webbe: Sophisticated. [01:10:34] Dave Pengelley: But uh we'll see you next week, guys. [01:10:37] Matt: Yep. [01:10:37] Kate Fabian - Adoptech: See ya.