Build AI That Works

Mark Smith 0:05

The intelligence age is not coming. We're already living in it. I'm Mark Smith, and each week I speak with the people building, leading, and navigating a world being reshaped by artificial intelligence. We explore AI technology, the changing nature of work, and what it takes to build a meaningful career in the intelligence age. This is the intelligence age.

Growth and Demand in AI Solutions

Mark Smith 0:25

James, welcome back to the show. It's it's been what, maybe a year since I think you were on my last podcast. So not too long. What have you been up to in the last six months?

James Diekman 0:35

Uh it's been a very busy six months for us, actually, Mark. So thanks for having me on again. Good to be back. Um so look, Accelerate Tech, our organization. We've we've grown quite uh quite substantially over the last six months. So we're now at about 32 staff. I think the last time we chatted was certainly less than that, probably not half, but you know, somewhere between that. And uh a lot of that is to meet demand of, you know, we do a lot of government work, so government-based projects. Not all of it is AI, which is interesting. So we can get into I guess how and why, and I guess the technologies that government is is selecting and using in this day and age and what we see moving forward. But yeah, for us, it is, you know, we've certainly doubled down on our Azure practice and foundry and AI come into that in Prodev. We're doing a lot of solutions that have AI built into them. So think of a pure software development application, internal application project, but we're seeing more AI being used within those applications for certain processes and workflows within it. So not just a chatbot being the primary thing. A lot around data platforms, fabric, still your power platform and dynamics 365 going strong as well. So yeah, look, lot a lot happening in that space. We're using AI quite a lot internally, as as probably most organizations are, and learning a lot as we go, and sharing a lot of that with our customers as well on what we're seeing work well, what we're not seeing work well, best practices, etc. So yeah, it's in very interesting time.

AI Usage Trends in Business

James Diekman 2:14

Mark Smith 2:14

I saw some data the other day from Open AI of their, I think they've just passed the one billion active users on platform, and they were taking an aggregate across business. What are people using AI for? You know, in particular in this case, JGPT. There were two areas of the business or of a business that totally separated from all the rest of the organization, and that was engineering. So engineering was far outstripping any other category in people not only using AI, but producing new things with AI that benefited their wider business. The second category was marketing, same deal. Using AI massively for enabling that role function, but then also producing new things that created leverage for the rest of the organization. Then everything else, HR Finance, blah, blah, blah, blah, blah, was basically in a category of they just consumed. Yeah. And then they were users of AI. How can it help me, you know, do whatever I'm doing? But weren't, if you like, counterproducing back, in other words, putting stuff back into the organization outside of their consumption to do their own job role. I just thought it was really interesting because three years ago it was talked about support, customer support, being the biggest AI consuming area. I don't not hearing that as much these days, but it doesn't surprise me about marketing. And it doesn't, and absolutely doesn't surprise me about engineering. And engineering's a broad term, right? So we could talk about it being software engineering or product development or whatever. But I just thought that was interesting. And you mentioned there, you know, you're using it a lot more internally. What are you seeing with the customers that you're working with around where they're getting the value from using this new tech?

Challenges and Successes in AI Adoption

Mark Smith 4:04

James Diekman 4:05

So yeah, I mean look, this is this is something that we we think about day in, day out. And look, I think you know, if you take a step back a bit, look, look at what we've got and what we've had for a while. Everyone, you know, everyone has access to intelligence or even superintelligence. You know, whether you're a government agency or a private organization, you know, one way or another, people can get access to whether it's co-pilot, co-work, Claude, Chat GPT, etc. And you know, to your point before, like we're seeing a lot internally, it's we're using AI to produce things and we're using it to work out or build probably more so deterministic applications and software and workflows, things that we either couldn't do before or was too time consuming to do before. So, you know, deterministic is is certainly what we're seeing a lot of in government. Where we're seeing AI, I guess, come in, or where how how they're using it at the moment is not certainly not broad brush approaches. A lot of agencies, a lot of customers have rolled out co-pilot. And and I think honestly, they're getting they're getting mixed results. Some are adopting it more than others, and you know, you're hearing conversations around, well, you know, we're we've been using it for some time, but then we're seeing the usage drop off. How do we make sure that people people can use it more in their daily workflow? But I think that's kind of the core challenge and core bit of friction. It's you've got this new tool, this new thing, this chatbot that's outside of your daily workflow that never really used to exist, and now you're creating stuff and trying to use that within your workflow. And and look, if you're not giving it the right context, obviously, it's gonna have mixed results. So, where we're seeing the best successes is working with these customers and identifying where in your business processes at the moment or your workflows could AI help the most? Where is it gonna have the best chance of success? Pull that one bit apart and then transform it with either introducing artificial intelligence within that workflow, doing a range of either judgment-based activities, because that's I think what you know, that's what it's very, very good at in reasoning over large amounts of data, or automating workflows within that process or building workflows that never used to exist using tools like GitHub Copilot, Claude Code, etc. And then starting there, obviously doing proof of concepts, doing MVPs, measuring it, getting feedback on okay, well, what were their what were our benefits, what were our KPIs when we started this project? Okay, we thought we were going to achieve a 30% reduction. Have we achieved that? Yes or no? That has been a few times the business case to unlock further funding to proceed and roll things out to production. So we've been um look, I I think like a lot of firms and a lot of partners, everyone's talking about doing AI and they've done um lots of projects for customers. There's a lot that never make it into production. I think you know, you need to need to peel back the surface a little bit and look at the the the projects that partners and and SPs are doing and vendors and see, okay, which ones have actually gone into production, how many? And and even with us, so we've you know, we've probably done over 30, 30 AI projects that have had some form of AI for customers. Maybe look five, five or six of those have actually gone full production. So you know, some never see the light of day, some the numbers just don't stack up, and there's there's a risk tolerance there as well that they want to take a more cautious and measured approach.

Mark Smith 7:48

Interesting. Yeah. Yep, yep, yep.

James Diekman 7:50

So so yeah, where we're seeing the um to sort of tie back to your your original question, it's more pointed solutions within workflows, sub workflows, existing business processes. As opposed to just rolling out Copilot for everyone, which is more of a which is more of a deployment and an adoption thing, in my view. That's changed probably a little bit with with co-work. But again, it's still a user-based tool and you're still you know, it's still something that's not a part of your your daily workflow, at least by default. So so yeah, we tend to do things that benefit more of a back-end process or yeah, it's still human in the loop as well, but uh but yeah, it's operating more behind the scenes.

Government Integration and Local Government Dynamics

Mark Smith 8:34

So my understanding of my time being in Australia and working in Australia is that particularly in local gov councils, things like that, there was a there were kind of two or three main players when it came to the software stack of any council and organization from memory. Civica Tech One is the other one, is it? Yeah, yeah. And what's it is is it the odd one on SAP if they're large enough?

James Diekman 9:01

Some some of the larger ones do, but N4 is the other big player.

Mark Smith 9:05

So it used to be, in my experience, as in councils I always saw as tire kickers. Love the new tech, wanted to do it, never any budget and stuff. And my hat's off to you because you've obviously built a company around that plane. Uh are you more percentage-wise in local or more state government? Where would you see you sit or fed?

James Diekman 9:29

Definitely more state government. Our local government practice, which is a product and it does integration for all of those platforms. So we we sort of, you know, we're we're the plumbers, really. Yes. So we provide the integration into those into those platforms that they use for property and rating, asset management, customer service, etc. They're they're varied in how easily they can be integrated into. And yeah, it it's roughly between 20 to 25% of our total, you know, customer base and and revenue is local government. So once once you and look, there's a lot of there's a lot of new players that have popped up recently with solutions and AI solutions and SaaS solutions for local government.

Mark Smith 10:12

Okay, interesting.

James Diekman 10:13

Um but the the challenge that they face and you know some of the conversations that we've had with other people is oh, why couldn't you just do this with AI? It's so you know, why couldn't you just build this that solves this problem?

Mark Smith 10:23

Yeah.

James Diekman 10:24

It's like any organization, you know, they're not going to just go and rip out these tools and these things that they use and built, you know, invested millions of dollars in and built core processes around and trained their staff on. So if you're a consultant or a business trying to get into that space, uh certainly don't lead with that approach. You the the way we did it is, well, through a particular challenge that New South Wales councils were having with the New South Wales planning portal, we built the integration based off that. But as time has gone on, we've realized actually this integration that we've built can solve many other problems and it can extend and transfer data between other applications. So we we are now talking to some of these and even partnered with a SaaS application in you know, they're in a very particular niche area around developer contributions software, and we do we do their plumbing for them because the last thing a council wants to hear or see is another shadow IT or SaaS solution that business has gone and procured, and it doesn't integrate in, and when something goes wrong, they're the the ones that are going to be hit with the the support or the uh the responsibility to go and fix it. So yeah, there's like yeah, it's not just it's not just AI, uh it's not just automation, it's integration as well. That's core to a lot of these systems working effectively and scaling as well.

Resistance and Regulatory Frameworks in AI

James Diekman 11:45

Mark Smith 11:45

Are you coming across much resistance?

James Diekman 11:48

In t in terms of AI solutions or Yeah, yeah, yeah.

Mark Smith 11:52

As in you you mentioned, and this seems to be more rigorous just in the conversation so far in Australia than we see in New Zealand, the you know, the models needing to be in country, blah, blah, blah, blah, blah, as in when it comes to AI. And, you know, I suppose Australia has has lived some burnt experiences. You know, there was the Queens, Bank of Queensland back when I was there that were running Salesforce, and APRA found out it was all being hosted in San Fran and were not happy about that entire data set being over there, and some people got spanked for that type of thing, and and there's been other cases like that. Do you think that and that's why I wonder if there's a higher degree of resistance as in because the perceived risk profile is higher for these organizations to take a more you know, slowly as we go approach than others. But kind of what is that what's that feel that you're getting from the market as opposed to you know the flip side could be there's just an you know insatiable hunger to get going and do stuff?

James Diekman 13:01

I think I think the hunger has come calmed down a little bit. So yeah, wine back a couple of years, it was yeah, it was it was AI was all the rage. It's you you know, we needed in our organizations, all the major vendors, you know, Microsoft, Google, Amazon are all pushing it really, really hard. The policy and the organizational structures hadn't caught up at that stage. People were, you know, I I guess you know, governing and approving these applications based off how they typically did it, right? Existing processes, architecture review boards, you know, change approval boards, things like that. Now what we're seeing is you know, from contractual terms have been updated, frameworks have been implemented. So New South Wales government has the AI adoption framework. So any AI solution that government develops, whether that's a partner or a vendor coming in to develop or internally themselves, have to run through this framework. And long or short of it is, without butchering it, it's basically there to determine the the level of risk that could an agency could have and how they're going to manage that risk if they go ahead and build that solution. And importantly, what effects more on the adverse side would it have on society, whether AI is making a judgment on something, maybe it's a case or a claim or it's reviewing documents or things like that. Could it have you know direct consequences or indirect consequences if you know a human is is reviewing the outputs of what that might be? Uh Queensland government has a framework as well, the the FARA framework, and uh I think I think most states do, and obviously the federal government do as well. So there is specific instructions, processes, guidelines, guardrails that have been stood up. The next layer down is what we're seeing in government agencies is actual roles have been created, AI specific roles, product owner roles. We we're working with some of them at the moment, AI strategy roles. You know, it's it's they're dedicated, it's baked into the title and the remit of what they're doing. And then when we're talking earlier before before the show, uh there's there was a mandate by the federal government, at least we're all federal agencies, to have a a chief AI officer for each of their departments. And I think it's it's taking a bit of time to embed those into those organizations. Yeah. But yeah, certainly at a state level, we're seeing we're seeing dedicated AI roles being created within technology teams. So there is more process, uh a bit more bureaucracy as well. So it's certainly slowing things down and they're taking a more measured approach around reviewing these solutions and determining if they will potentially cause adverse outcomes or what the risk level is by introducing these. So look, I think it's I think it's a it's a very good thing that they're doing that. That's not just, you know, they're just letting all these things through and uh you know, creating even bigger problems for themselves. Um so yeah, it it's it has to happen. I think I think you'll you know continue to go down that path. But again, like you know, for Australia to be a uh an innovative country and and for the states to be innovative in their own rights, there you obviously need balance on you know policy and and uh and controls and guardrails and and red tape. So so yeah, it's it's it's finding that right balance.

Public Perception of Data Centers and AI

Mark Smith 16:27

Where's the what's the temperature like in the country around things like AI as a whole? As in so we're talking about the general populace, data centers, water used in regards to those, energy consumption, driving power prices, etc. What's the kind of from your perspective, if you could put a temperature probe out there, what's the temperature?

James Diekman 16:50

I think I think data centers have that this they've been getting a very negative connotation and negative perception lately. They've they've always been there, but now this AI is shone a light on them, and you know, what powers AI is these big data centers, and more and more are cropping up. So so there's a lot more public awareness and visibility of those, and I think you know that we there needs to be some very serious conversations and and policy around this because on one hand, you know, it's obviously people are concerned with them and living near them and the potential impacts and obviously the power that these things take to run and the water as well. Get to an interesting conversation that I had recently around on that side with water utilities. But on the other side, everyone is using it, and a lot of people are now using it day to day and relying on it, and businesses are relying on it, even government are relying on it. So, yeah, it's you you know, you can't have your cake and eat it too. We need to figure out a way that it is sustainable to the environment and to, you know, if there's any um potential adverse health implications. I think all those things need to be looked into as well. But yeah, I don't think Australia should well I don't I don't think we can just slow down the progress of it because we're seeing firsthand some of our customers being hamstrung around the intelligence they have access to and that they can use because the model not in country to support the latest models that are coming out. So when uh Opus 5 or even 4.8 or Chat GPT 5.6 comes out, Microsoft say, Yep, we've got it all in Azure Foundry, you know, we've done a great job, it's there. But it is not available in every single country, and certainly not Australia. And it takes a while to become available there and for people to start consuming it in region.

Mark Smith 18:40

So we And that's because it's not just a software thing, right? It's a hardware issue. Is it they don't have the uh the GPU, the the the the AI, whether it's you know Google TPUs or GPUs from NVIDIA, etc., there's a big shopping list of much more wealthier, perhaps, buyers and other geographies around the world that Australia is competing with. And therefore doesn't have the hardware to run the software.

James Diekman 19:06

Exactly. And you know, I think benefits of Australia, we've got a lot we've got a lot of space, but we've got you know potential of of renewables and and things like that to help help assess offset. The you know, the thing on the water side, why I sat on a round table for water utilities back in February, uh Victorian water utilities, and AI was the yeah, and data centers was was a hot topic. And they're they're worried, they're really worried about how they're going to be able to meet that demand because they simply don't have the capacity at the moment for what's being proposed.

Mark Smith 19:41

And so so so I find the water argument a baffler to me, as in when the minute you put evidence around it, it starts to come unstuck. One, a modern data center runs a closed loop system like the radiator in your car. You're not just pumping water in and flushing it out, it is a closed-loop heating-cooling exchange system. There is now three data points that are documented. It takes no more to run a data center than the equivalent of two McDonald's bus or restaurants a year in water consumption. You're gonna have staff there, so the the stuff you're gonna consume is yep, the toilet's gonna flush, they're gonna have a coffee break, they're gonna use water for that. But the actual cooling where the water is generally used is a closed-looped system. And if you take it and compare it to another industry, and I've seen it compared against golf courses, for example, it uses something like a a fifth of what a golf course uses in a year in water consumption. Yeah, we're not running around closing down all the golf courses. If you compare it to, I think it's almond growing, it uses like a fraction compared to what the almond industry uses to produce one single almond. And I think if we looked at it against agriculture and things like, you know, farming cows or sheep or any of these type of things, and when you do these apples to apples, all of a sudden it's fractional, but it's the news headline grabbing things. And and Osachi that recently just came out and said that it's equivalent of two average-sized restaurants and water consumption per annum, an entire data center. And if you look at the all the data centers as a whole, I think it would be fractional compared to other industries that have not had to defend their water usage. That the you know, the the two are energy and water. Water, I think, is a mute point if you really unpack it, and then energy. Mate, Australia's got some incredible sunshine areas. And what solar can do nowadays, and I think you know, my observation, Australia seems to be ahead of the game in solar production, and now the energy infrastructure is cat catching up. But surely the regulators can just say, hey, you need to be net neutral in energy consumption, and here's your consent for X. Amount of water, stick within it.

James Diekman 22:02

Yeah, the water the water ones are interesting. I mean, you know, I I didn't even know that what you mentioned before, but just yeah, being being in this room of water utility. I mean they're all tech tech people anyway, but yeah, it's interesting. The the media and depending on what media you you subscribe to definitely um as we know can spin things in lots of different ways and give people different different views and and opinions. So yeah, you you you're 100% right. You need to you need to dig down a bit and actually re you know look at all how these things actually run. You know, what what yeah Yeah, I I I agree.

The Need for Sovereign AI Infrastructure

James Diekman 22:38

Mark Smith 22:42

The you mentioned there about one of the big challenges Australia has is the the compute for AI is sitting offshore, which therefore breaks a lot of rules and makes it untenable to be used. There was fear spread across the world when Anthropic's Mythos was pulled from the market by the uh US federal government. And did you see anything from a state, fed, local government perspective that went, oh shit, we really need to have our own AI infrastructure on soil and not be cap in hand to another state. It's a decision around you're allowed it, you're not allowed it, you're in the in-group, you're not in the in-group. And therefore, and I'm just going to make a bit broader question so you've got stuff to work with. Kimmy K3 came out, it's Chinese, and it's phenomenal, right? As in the the difference, the delta, I'd say 99% of users wouldn't notice the delta difference from a intelligence served, but at a much lower price point, especially if you can house it on hardware and country. What what are you seeing? What are you hearing? What are you thinking?

James Diekman 23:55

So around that time, we we were having a conversation with a customer about, well, going back to the conversation we had before, they were they were they were limited around not only the models that they could use on shore, but actually the throughput of how many tokens could could go through the could go through their uh TPU. Is it TPU? Yeah, TPU at sorry, PTU, prioritized, no, prioritized throughput unit the Microsoft um vernacular associated with basically reserve capacity for for AI compute. So so yeah, it's not just context windows and it's not just model capability, it's actually the throughput as well. So when you start to ramp up to more users, you you actually uh you gotta take that into consideration as well. So we we just started discussing local models and setting up local models on on their infrastructure or even hosted infrastructure within Azure.

Mark Smith 24:52

Nice, nice, yeah.

James Diekman 24:53

And that is that is a very viable option. And when um I spoke to a few other people when that whole you know Fable got pulled and and that whole fiasco happened, and yeah, I think it highlighted that hey, we are we are very dependent on this. And as you know, I think if if organizations started building more on Fable at that time, like if it happened weeks or even months after, it would have even more of an impact because they they're now becoming dependent on it. But the the yes, we absolutely need more sovereign capability, but I think this is we're we're seeing far more conversations and a trend in local models. And the local models, you know, you can't just go and download some of these and run them on your computer because they just simply won't work or they're just not going to be efficient.

Mark Smith 25:41

But the you need about 600k worth of hardware first.

James Diekman 25:44

Yeah, you need, you know, you need a you need another mortgage for a small data center out the backyard to run them properly. Exactly. And and if you were to go and put it on on rented compute, it's you know, it's just in crazy price. But as the models get better, as that uh quantization layer and and the the hardware optimization gets better, you know, we will be able to run really good models on hardware that is that is not that expensive. And that is very good for government because they're gonna have full control over the infrastructure, the hosting, and the weights of those models as well, because they can they can go and fine-tune them and train them, yeah, train them further, which is I think you know, that's becoming a bit more back in trend at the moment to get better better results because the frontier models they're so tweaked behind the scenes and and sort of yeah, they're they're you know, there's some funny stuff going on behind the scenes there. Um and you know, a lot of conversations we're having at the moment is you don't need an opus level or fable model for all the things that you're trying to do. Like break it down, and you can use some of these lower models or even local models for parts within, let's you know, call it you're transforming a workflow or you're doing any sort of development, use the models where they're best suited. Like, you know, on any software development project, who are the most, I guess, expensive resources? Typically, they're the solution architects. They're the ones that come in the beginning that help figure out how this whole thing needs to stick together. Use if you if you're doing software development with these models, use the fables, use the opus fives, use the chat GPT-516s to do the planning, do the architecture, do the, you know, come up with the UI UX, and then have a lower model do the actual development. Or if you're building a workflow and it's just some basic document OCR scanning that you're doing, you don't need Opus for that. You don't need these higher class models. So yeah, there's there's a bit of education there with customers to help them figure out that you actually don't need this for your for your solution. And I think more so recently with what Microsoft is doing with their pay as you go pricing, Claude Cowork, you've got to pay for that. Copilot credits, you've got to pay for that too. So people are gonna be far more aware of what their spend is and what their token spend is moving forward. And I've spoken to several organizations that had cowork enabled for everyone, but they've scaled it right back to 20 users, for example, um private organizations. That'll be surprising. They're going, well, this is this is risky because you know, like I said before, not everyone prompts these things the same way, they don't have all the context, they can they can go off the rails. I've read some horror stories of I think someone tried to clean up their mailbox and it it was eleven thousand dollars to to do that for some post I read. Because co co-work just, you know, this went absolutely bananas. And yeah, so education on where these models are best used, and that's something that we we we do. Local models absolutely have a role, and that role I think is going to become far, far more important over the next six months to a year. We're gonna see way more, way more development and integration with local models. And you know, I think that will that will certainly lessen our dependency on these frontier firms. They're still gonna be the default for some time, and it's still like we said around before, data centers and local capability, that still needs to happen in parallel. Yeah. Not everyone's just gonna pack up and and move to local and you know break the tether, essentially. So it's a yeah, it's a mix of things.

AI's Role in Project Management

James Diekman 29:27

Mark Smith 29:27

My last question is out of left field for you. And that is I've been thinking about project management in an AI world and wondering if it's one of those things that AI would really excel at running projects with all the constraints, you know, whether it be quality, whether it be cost, whether it be scope, creep, etc. Using any kind of historic project methodology like you know, Scrum Agile, Waterfall, PP, whatever it was. Do you see a world as an and because I know what you do is similar to what I spent years doing with running power platform and dynamics 365 projects? Is that often a project success would come down to how effective your project manager was in their role? Do you see that becoming more and more an area where AI can excel in? Really? So often I I feel like people get confused by going, what task can AI do? And I think of it from what role can AI do? Like a task is one thing, a role is a lot of different tasks and a lot of nuance and a lot of context that must be known to be able to deliver on that role. What are your thoughts?

James Diekman 30:41

So the short answer is absolutely it can, and it can do a lot of the the mundane tasks that any very good project manager or well-funded project simply, you know, things that they just couldn't get to. Project governance, for example. So, you know, to give you an idea of the things that we're doing and experimenting with is uh I think I spoke about this last time, but essentially a project brain, right? So you you know every project gets created has a has a SharePoint side or a team side, and there's there's some knowledge there. But there's also it needs access to your broader organizational knowledge as well. So they're your processes, your policies, your things about your company, your design standards, your software development workflow, etc. Those can all be codified. Those can all sit in a a clawed MD file or an agents MD file, and everyone who is on the project can kind of bring their own agent and then connect into that that knowledge source. And then and then from there you you would have skills, and skills are there to basically do things like write DevOps tickets or update DevOps tickets or update them with estimates or do reporting, for example, a weekly status report. It's there to update the risk register, the assumptions, dependencies, decisions, all of those, you know, when I was doing project management and program management a while back, uh all the registers that you had to maintain was a nightmare. Yeah. And it's risk registers.

Mark Smith 32:06

Yeah.

James Diekman 32:07

Yeah, and they never get the the love and attention that they need. But when you've got this, you know, super intelligence reasoning over all that data and multiple people are connected into it, it can do some pretty cool things. So, you know, it and then if you have your developers using it, it can go and take take an issue. You know, let's say you know projects for example, it can take an issue from in your backlog and then move it to in progress, and then once it's in progress and it's done the the the pull request and you know can move it into done and go through that whole process. So budgets as well, the financial tracking, you can you can hand it off and do a lot of different things. I think the next the next level of what we're what we've actually been experimenting it with uh for for a while is and with mixed results, definitely, but having having an agent join a Microsoft team and actually be useful, not just in there to answer basic questions, but actually be in there and like like a normal staff member. It can you don't have to tag it, it can hear what's going on in the conversation and then actually go and do work. So you know if you've seen Buzz that's come out recently from Block, so Jack Dorsey's new it's like a Slack competitor, but it's an AI agent first tooling. I haven't used it, but I've watched several videos on it, and it's very much AI agents and humans in this one collaboration tool. And yeah, I think I think that's where that's kind of the next level where you can have your solution architect agent, you can have your project manager agent, you can have your developer agent, you can have your QA tester agent, and they're all doing different tasks across different projects and you know, or built and backended in Azure. It can all happen. It's yeah, something that we're we're looking to do next. But absolutely on project management, yeah. Like I love it. You know, it can it can do a lot of stuff, and we've we've used you know use those technologies quite a bit to to help with things like that. So very cool.

Mark Smith 34:08

James, it's been great talking to you as always. If you're listening and uh you want to check out James' bio, it'll be in the show notes as well as any links, etc., that we have or resources we've we've talked about in the show. But thanks once again, James, for coming on. Thanks, Mark. Always a pleasure. You've been listening to the Intelligence Age. I'm Mike Smith. The Intelligence Age is already here. The question is, what we choose to do with it.