How companies get real value from AI - Orla Glynn

Speaker A: Running a small business means wearing a lot of hats. Between managing employees, handling payroll and keeping up with paperwork, there's a lot to juggle. Luckily, Gusto makes it easier. Gusto is online payroll and benefits software built for small businesses. It's all in one remote, friendly and incredibly easy to use so you can pay, hire onboard and support your team from anywhere from automatic payroll tax filing to simple direct deposits and benefits options for nearly every budget. Gusto helps simplify the process. Plus enjoy unlimited payroll runs for one monthly price with no hidden fees or surprises. And with automated tools for onboarding and offer letters, you can focus on what matters most. Try gusto today@gusto.com tech and get three months free when you run your first payroll. That's three months of free payroll@gusto.com tech one more time gusto.com tech.

Speaker B: Hello and welcome to Don't Stop Us Now AI Edition. I'm Clare Hatton.

Speaker C: And I'm Greta Thomas. AI has been described as the most transformative technology since the harnessing of electricity, and we are here to keep you in the loop for what you need to know and do in order to stay relevant in this fast changing world.

Speaker B: Each episode we bring you leading experts and together we explore how you can stay ahead of the curve and in the know for what skills will be valued in the future, how jobs may change, and how industries will evolve.

Speaker C: So why not subscribe to Stay in the Loop and without further ado, enjoy this week's episode. Hello and welcome to this week's episode of Don't Stop Us Now AI Edition.

Speaker B: Here's something our guest today has been pondering and solving for Every organization can buy the same AI, but not every organization gets the same advantage from it. So what are the most successful ones doing differently?

Speaker C: Orla Glyn helps leaders Answer this question Orla, uh, our guest this week, has spent more than 20 years leading transformation inside banking, energy and telecommunications companies. And for the past 10 years, she's led major enterprise data and AI transformation for big and complex organizations, including Australia's largest telco telstra.

Speaker B: Absolutely. And Orla's experience is pretty rare, so we're really excited to share this conversation with you. Ola is now founder and managing director of her own AI consultancy, the Overstory Group, as well as a board director and advisor. She's also been named one of the top 10 women in AI in Asia Pacific.

Speaker C: That is pretty cool.

Speaker B: It is.

Speaker C: All Orla's work is about helping leaders figure out what AI is actually for in their organization and how to Create genuine value and advantage from AI. In our conversation, you'll hear Orla explain the eight capabilities organizations need to be AI ready, why data is less like oil and more like soil, how to think about introducing agents into your organization, why it's so hard to be a leader right now, and, um, which size of company is currently seeing the most growth from AI.

Speaker B: Orla also shares how she's building her own AI native business, why she avoids relying on just one AI provider, and how to keep watch over AI agents once they're working in your business.

Speaker C: And she makes a compelling case that how AI shapes our future is a design decision not just for the heads of AI frontier companies, but for so many, many more of us.

Speaker B: You can say that again. So, without further ado, enjoy our conversation with the insightful and forward thinking Orla Glynn.

Speaker C: Orla Glynn, welcome to Don't Stop Us Now AI edition. Thank you, Claire and Greta. It's fantastic to be here. We're really excited to have you on the show and to get started. You may be aware that a question we ask all of our guests is, imagine you're at a dinner party and you're sitting next to someone you hadn't met before, and they turn to you and they say, so, Orla, what. How do you typically answer that these days?

Speaker D: That actually happened me recently. I was at a dinner in Sydney during Vivid Festival. I was, um, invited to a roundtable. There was about nine CEOs and CIOs around the table. And it's amazing when they ask you that question, it's the simplest question and it actually, um, can be the hardest question to answer when you're under pressure. So I just told them my story, which was I've, um, worked inside large corporates for about 20 years, driving major transformations across financial services, energy, and telecommunications. And what I sort of started to see was that every organization can buy the same AI, but not everybody gets the same advantage for it. And about 12 months ago, I decided to step out of corporate and I realized that there was a gap in the market. So I founded my own organization. And now I help leaders work out what AI is actually for in their organization.

Speaker C: Fascinating. And I bet you have some pretty good conversations around, uh, said dinner tables when you say that too.

Speaker D: Yes. The bit that probably resonated at the end was I actually said, you know, my most important job is I'm actually a mom. You know, I really think about my daughter. She will be 18 in 10 years time. And I really think that, you know, the decisions that we make as leaders now really have will shape how she grows up and how the world will be. And I think that, you know, one of the reasons why I stepped out to really think about my role now in this AI world is actually what do I want it to be like and how do I actually want to help leaders make the right decisions around AI?

Speaker C: Yeah, I think that's cropping up increasingly frequently now as people think of, ah, the consequences as they realize the magnitude of potential change that lies ahead of us. Now, I detect a bit of an Irish accent there. And, um, you're talking to us. I think you're based in Melbourne. How did you come to be in Australia and move through finance and accounting to data and AI?

Speaker D: Yeah, so, well, I guess like most Irish people, we always tend to travel all over the world. So, um, I had worked, um, in Pricewater as Coopers in Dublin for five years and took, you know, the six months off like most of my peers back then. And um, I came to Australia and, you know, fell in love with the country, much to my poor mother's, um, great disappointment. And you know, I landed a job in Westpac, uh, in Sydney in 2008. And then I moved to Melbourne. I was about three months in the role when Lehman Brothers went under and I was managing the securitization, uh, portfolio for the bank. So I got to see the GFC right happening at that moment in time and effectively the next five years of my life, whether it was spent in that world and watching the technology and the models and everything that was happening. And then, um, I was fascinated by how technology really drives a lot of business outcomes and business decisions. And then, um, in 2012 I got the opportunity to head up NAB's data and analytics center of Excellence. And then from that, you know, that just, uh, opened up a whole new career opportunity for me. And I've spent the last 12 years then driving data and AI across those organizations.

Speaker B: Fantastic. And you know, I know you've had some great experience because I think you've also headed up data and AI in Australia's biggest telco Telstra. So, you know, today we'd really love to explore what leaders of different sized businesses can do to really capture the real value of AI in a safe way, of course. So starting at the big end of town with an organization that's sort of got considerable size, multiple divisions, multiple locations, not to mention reams of data, which is probably all over the place, how do you think about sort of prioritizing what you need to do to harness AI successfully in an organization like that.

Speaker D: Yeah, so I guess, you know, first of all, you know, every organization can buy the same AI, but not every organization turns it into advantage. So it really comes back to, you know, the conditions that exist inside the organization which are very much set by leadership. So, you know, first of all, you know, where I've seen this be most successful is when, you know, it is very much seen as a whole of business transformation. Also that it is led by leaders right from the top down. So the board understands the implications and, um, opportunities and risks of AI and then that the executive team are very much accountable for how AI gets embedded within the business. So, you know, the first thing very much is it is a business strategy. You know, it is, you do not necessarily need to have an AI strategy. You need to have an AI roadmap and an AI plan. But it ultimately it is how the business is evolving as a result of AI. That's the first layer is very much your strategy and your leadership alignment. Um, and the second then is very much, you know, how do you understand where value is generated in your organization? So one of the biggest challenges at the moment is most organizations are just putting tools into their companies and handing them to their workforce and kind of expecting value to emerge from using the tools. But actually, you know, you have to really, and uh, this is probably where that accounting skill set came from, was and has helped me is that you really have to understand the P and L and the value in the organization and the KPIs and what are you measuring and how do you actually know that if you've put in a new tool or a new system or uh, you created a model, that it's actually moving the needle and delivering the benefit? And it could be something like, is it the customer experience you want to change? And therefore is it like how do call centers operate and how do you reduce the volume of calls duration to improve the customer experience? And therefore what's that value to the organization versus is it actually, you know, we want to create new business models and actually have, you know, a new product that's going to market. But you really have to look at it around those major value pools and where you're going to put your bets and then you have to design the organization around that. So

Speaker B: yeah,

Speaker D: how you design for risk, how you design for, you know, the complexity of the processes that exist in the organization and then how you design the actual data and technology landscape, the kind of spine of the organization that supports it. So if it's at the top end of town, it is a big multi year, multi investment, uh, to drive the value.

Speaker B: Yeah, absolutely. And, you know, I think this sort of discussion around return on investment is interesting because it is such a big change for these organizations. And actually it's not just large top end, it's, you know, medium sized, even small organizations. It's a very big change in the way that they're having to strategize lead structure, look at value pools. The return on investment question, I guess it might be too early for the return on investment to be coming back. One of the really interesting parts and one of the critical parts that we see is data. And with, you know, companies having so many different pools of data, so many different types in multiple locations, you know, not tagged, multiple different databases, no clear sort of golden record. What do you advocate in terms of getting data ready?

Speaker D: Yeah, so I guess first of all, it really depends on the organization and where they're at from their maturity perspective. So I guess first of all, I will start by assessing, you know, where the organization is across 18 capabilities, of which data is one of those. One of the biggest challenges for large organizations is the fact that they have multiple legacy data platforms. You know, they've acquired new companies, they have huge complexity, and therefore they have to rationalize, modernize, decommission their platforms. And you know, what you're seeing now in annual reports that are coming out, the organizations that made those investments, you know, two, three, five years ago, they're starting to get the return now on AI. But that is a significant investment when you have to replatform and modernize your data. So the foundation of getting your data right is critical. Ten years ago, organizations were looking at, okay, well, we just have to get all our data into a data warehouse, a data lake on the cloud, and then once it's all there, we then can do all this great work. And that had its major limitations. What I look at, and when I look at this with Climbs, is first of all, I just want to understand what data do you have? And is the data set up in a way that the data can be protected and used carefully? But then it's also looking at the. So if you can imagine that data is your horizontal capability in your organization, but then we look at what is that value pool? And then we say, well, where are we going to get the greatest value? And therefore, let's use that data set to support that value outcome, um, as the first data that you're making certain has the right data quality, data lineage, et cetera. That ultimately is where you'll start to get the greatest outcomes. But you also have to manage the fact that technology teams have to think about what systems need to be decommissioned in line with their timeline. So it's a, uh, need to think about both the product roadmap and the technology roadmap at the same time.

Speaker B: Yeah, for sure. And also particularly in financial services. Well, actually across the board, sort of all the privacy legislation that's coming in, whether you're in Australia or in your. If you're in other parts of the world, that's uh, obviously another layer that you have to think about.

Speaker C: Yeah, there's huge complexities in codependencies, aren't there? Yes.

Speaker B: Product

Speaker C: roadmap to the technology roadmap. And then there's data permissions and who is allowed to access different parts of the data that all get uploaded. So, yeah,

Speaker D: completely. And then if you're starting to think of agentic capability, you really have to understand your data and um, your permissions and your controls and what data can be accessed. So years ago we'd say data is the new oil. Recently I heard someone say data is actually the soil and it's ultimately what you cultivate. So therefore how we actually invest in that data so that you can grow from that data is really, really important.

Speaker C: Yeah, I think that's a really great way of thinking about it. I want to come back to something you earlier. You know that it's not about organizations, boards and leaders having an AI strategy, but rather they think about a business strategy and an AI roadmap. You know, how does AI enable, uh, a potential reinvention? I think this is something that leaders haven't had to encounter in a. In the same way previously with other tech evolutions, in that there is that sense of. There's almost a blank sheet of paper. Where do we put our focus to create new value? Is it completely new product or service line? Because we've got this capability of scale now that AI gives us, or is it, you know, as you said earlier, customer service or something? Are, uh, you sort of sensing or seeing different leaders struggling with this new kind of strategic question that before strategy has been, has tended to be more reactive, but there is a real opportunity now to be completely creative.

Speaker D: When I talk to C Suite about this, I talk about it around three horizons. The first is how is your business strategy evolving for AI and Horizon 1, which is very much around driving efficiency growth? It's very kind of like here and now. And most organizations, that's where they're focused Today. The second is the next horizon, which is, you know, what new business models can you actually think through? And then the third is actually what is the disruptor to your entire industry? Right now the boards are probably thinking a little bit more around horizon three and thinking what's happening to the organization. The people inside the organization are generally stuck in the horizon one, and I think that's the challenge. The second issue though is that, uh, what I'm noticing is there's this challenge of wanting to move fast, but not actually having the capacity as an executive team to sit and have the conversation around what does AI mean for in our organization? And I think this is probably one of the biggest challenges that I'm seeing is companies putting in the tools. Tools, but then not even spending a day or two together really mapping out and thinking about the implications of AI, uh, for their organization across those three horizons. That in itself is a challenge. So, you know, when I think about the strategy and the work that we're trying to do with organizations now, it's not about, you know, producing a PowerPoint presentation like you can, anybody can produce a PowerPoint presentation. Claude can do beautiful presentations. But it's actually the facilitation of the conversation to get them to think differently and um, to get them to think about what does this mean for their organization, their industry and um, their team and the people that they're leading. And actually the leadership that's required now in AI in this era is very different.

Speaker C: Yeah. And that ability to fathom the potential changes that could be made on the strategies, choices that they have, as you say, taking that time to sit around the table.

Speaker B: And I think also my observation is a lot of leaders aren't even using the tools themselves. And so therefore, at least not deeply. Not deeply, yeah. They're using it for a PowerPoint like, uh, writing emails or, you know, the sort of the small productivity pieces. But they're not actually really getting into the tools to understand what they can do. So therefore, it's really hard to be creative if you can't even understand what they can do.

Speaker D: I think it's really hard to be a leader right now in the sense of, like, there's just so much complexity. And, um, they're trying to run businesses today and deliver the outcomes that they need to deliver for this financial year and the pressure that they're under for immediate decisions, while also thinking about this new skill and this new capability that they need to have and, um, their organization needs to have. And I feel a little bit like, you know, We've spent a lot of time training up the workforce and not as much time actually investing in the executive teams and the leadership teams to give them not just the tools but actually the capacity that they need to lead differently.

Speaker C: I'd love us to sort of focus on that area where most organizations uh, are right now, that sort of horizon one. So horizon one and let's say mid size organizations. You hear so much about the pilots not getting to be scaled or AI projects just not seeing the benefit come through to the bottom line. You know, what's going wrong with a lot of the AI initiatives out there today, do you think?

Speaker D: I think it's important when we talk about AI that we're clear on um, what we're talking about. So the first is machine learning models. Okay. So I would say pre2022, that's what we were focusing on was building large, complicated machine learning models. So anything from customer, uh, lifetime value personalization, looking at the network, how the network is operating, or in energy building virtual power plants. So very complex machine learning models. They are complex to build and they are still, I would say, where a lot of the value in AI still sits. But we've kind of forgotten to talk about that now because we've got purpose in generative AI.

Speaker B: Exactly.

Speaker D: And that still needs to be, you know, what is the data? How do you have the right skills? How do you have the people who understand the problem and the people who understand the data science and the technical component working together? And in particular, you know, companies just um, understanding when you build those models that you have to invest in them after you build it, like build them because you have, you know, drift, et cetera. So that problem is complex and hasn't gone away. Okay, then you've kind of moved up to generative AI, which is sort of what we all see and look at now, which is we have ChatGPT and Claude and you can generate anything which is sort of more, what I would say is your assistant AI. So that is how do I use those tools to be more, more efficient or effective in my job? You know, organizations at the mom know, as I said, they've rolled out the tools and they've trained up people. Some people are using the tools for is some is things like connecting it to all of the slack and communication tools in their organization and just synthesizing what their day should look like and you know, making them a little bit more efficient. And you know, CEOs are asking me why are all these tokens being used? And I was like, well, you've got 25 different comms platforms and that's what your team is actually using it for. They haven't gone and said, okay, let's look at what are the really complex workflows or processes in my organization. Let's design for those. So that's a big part of the generative AI component. And then you move up into agentic AI or agents and you've got to think about, you know, what are those autonomous decisions that need to be made and where is AI going to act. And I think the challenge there is that, you know, this concept of value is, you know, they haven't, we haven't actually defined what we mean yet by value. So some value might be, well, my team is far more efficient and effective than they were before and they can deliver more in less time. And therefore where does that dividend sit? Is it just that they now get more work and absorb more work, or is it that I can grow my organization without needing more employees? So uh, I generally see companies less than 1,000 employees doing really well at the moment because they actually don't have this challenge of driving productivity the same as other organizations and they're actually doubling or increasing their growth. So I find at the moment those mid sized companies is where I'm seeing significant growth. So I'm not sure if I answered your question properly there. But effectively it comes down to understanding the different types of AI and where you're applying it in your organization.

Speaker C: Yeah, no, and I couldn't agree more that with the hype around generative AI that the smarts of machine learning AI and value creation, potential competitive advantage potential is kind of the sort of the lonely forgotten sister at the moment it seems a lot of the time and that people shouldn't be forgetting about that.

Speaker B: Yeah, no, for sure. And you know, talking about agentic, I think certainly in Australia, you know, obviously agentic is being used in some companies, but the risk conversation is the one sort of dominating at the moment, particularly with that hugging face incident and the Medicare hack and et cetera. What's your approach at the moment in terms of whether you build an agentic capability? Would you build your own? Would you go out to platforms that have got agentic capability and bring those into the organization. How would you be thinking about this if you wanted to start putting agentic into your business?

Speaker D: Yes, I guess that just comes back to what is ultimately your build partner buy strategy. Um, regardless though, around whether I would recommend to build partner and buy. What I do recommend is that control plane that you have to have in place so that overarching architecture that actually monitors and oversees agentic capability, whether it is the agents that you're building yourself, the agents that your third party vendors are building, or you know, where you're externally buying that in. And you know, what we're seeing now, especially with APRA and CPS230, is actually how are you going to monitor and oversee those? And you know, that is why, you know, that control plane is so important.

Speaker B: Yeah, this kind of always on risk monitoring. And for listeners who haven't heard about the term, sort of the control panel or control tower, can you explain a bit about what it actually is and how you would think about building one

Speaker C: or buying one

Speaker B: or buying one? Buying one, probably more likely,

Speaker D: yeah. So I often like to use the analogy of an airport. So if you think about the control tower in an airport where you have planes flying in and planes taking off, you need to have that same concept when you're thinking about agents. Airports don't have a policy, procedure and standard and then hope all the planes will land and take off safely. And that is the same when you have agents in your organization that are going to act, decide and make, um, autonomous decisions. So imagine that the control plane or the control tower in an organization is that technology intelligence layer that's overseeing all of your agents. When you have that in place, what you can actually see is how your agents are behaving. And um, a good example is, you know, you could have an agent with all of the controls and guardrails and responsible AI designed into that agent. But if the agent has a goal to answer a problem for a customer and it realizes, oh, I actually need to get the customer's name and the customer's date of birth. Even though you set a control that says don't hold PII data, the agent in a chatbot might actually end up getting that information. And if you don't monitor that and oversee that, you won't know whether you need to retrain that agent or fire the agent if it continues to move outside its parameters. And therefore, when you have lots of agents in an organization, you need to have that control in place and you need to be having oversight and governance of it.

Speaker B: Yeah, absolutely. And it's not just the agents internally within the organization, is it as well, you know, you're going to have third parties who have got agents, you're going to have customers who've got their own agents coming in to talk to your agents. It's getting quite complicated

Speaker C: and it's not that simple. I think the control tower tech layer because agents are showing just how sophisticated and goal focused they can be to um, pursue things that they shouldn't be pursuing in order to come up with an answer. They know their kind of human wants.

Speaker B: Can I ask from your perspective who are the main tech players who are uh, sort of leading in the space of control towers?

Speaker D: Oh so a few years ago it was emerging and we had to, in the organization I was in we had to partner with multiple stakeholders. So we had, it was a censure but effectively they have built the AI refinery that is their example. Um, but actually here in Australia we have a startup that is building an incredible tool that I, I not going to try and plug but I'm generally think it's amazing which is agents here. Um, and they have this whole concept of know your agent and you know it effectively is an independent platform that monitors the agents that you build in your own organization plus the agents that you partner with and by. And so if it's a Salesforce agent, a OpenAI agent, a Claude agent, Databricks agent, whatever the agent is, it'll tell you exactly how that agent is performing. They're you know still early days but they're a fantastic organization.

Speaker B: Great top tip there. Uh,

Speaker C: Ola. So if we step back there is so much happening. If you had just one piece of advice for leaders now in terms of capturing value and advantage from AI, what would that be?

Speaker D: I guess it's knowing what AI is for in your organization. It is very much not just about the tools. It's around understanding where will AI deliver the greatest value. Not all value gets delivered by AI being uh, deployed into your organization. But really choosing those bets more effectively which comes back to that leadership decisioning and that whole of business strategy. Um, and I think that's the, the biggest opportunity for the advantage is actually understanding that but then also making certain that you have the conditions in place and that's the work that you know I really focus on. Those conditions are your risk and governance and oversight. Do you have a robust risk, oversight and governance in your organization? Your people have your people got fluency and do you have co learning opportunities in your organization? So you know there's a real need for leaders to, to stop just focusing on once off training. It's actually this co learning that's required and you know creating the psychological safety in the organization and the need for people to be curious. These are all parts of the conditions that are needed. And then of course as I said it's the designing the work, redesigning the processes and then it's the technical capability that needs to be in place. So you know, it, it is a layered cake that has to be baked to actually deliver that real advant in the organization.

Speaker C: Yeah, no, I think it's really helpful just to have you sort of sum all that up because there is so much and it's. Yeah, it's, it is a sort of a tough strategic question that leaders have to ask themselves for sure. If we change tack now, how about how you use AI in your own working day? Are there certain things that you just love? How AI has improved what you do or any tools you single you'd single out for loving working with them.

Speaker D: For me, I'm building an AI native company, so I use AI every day probably a little bit too much sometimes in the sense that I'm constantly exploring new tools and new capabilities. But um, for me again it's the same question that I ask the leaders that I work with is actually what's AI for in my organization or my company? Effectively it's an AI advisory company, are a uh, consultancy company. But we're flipping the model for what consultancy is currently and that is, you know, we have, obviously we're building our own agent platform, but on top of that is we look at what are the roles we need in the organization. So there are some roles that I can have as an agent or AI only role. So um, I have research that's constantly running in the background. So for example, all of the recent annual reports for the ASX 100, I, you know, have those summarized. And then I've deep dived into them myself where I've actually read them. So I've read more annual reports this year than I probably would have previously, which is lucky you. I know, but I'm an accountant, so I still love that stuff.

Speaker B: Okay.

Speaker D: So, you know, there's research, there's agents, I've got analysts that I have working as an agent. And then effectively, you know, we have all of that, that capability. And then I also look at where AI supports. So it, you know, when I'm looking at capturing evidence, when we're doing our assessment tools, you know, that helps speed that up. And then again, of course there's some roles that we have where we don't have AI uh involved, which is very much the judgment roles. So we've, we've looked at that across every single thing we do and we've also looked at all of the workflows. So as much as possible, when I, I'M thinking about building something. I'm um, now thinking about it as a workflow as opposed to say previously if I was in a corporate I would have hired somebody in to do the role. So now I think differently around how I'm designing it. And then for me in my own day to day, you know, some of the things I love doing or uh, at the end of the week is I run a retro. So I ask the tools that I work on, what, what's worked well, what hasn't worked well, what do I need to improve. And that's actually a learning moment for me as well. So I find that really helpful. Obviously I use the major tools that are out there but I'm also, I'm very creative so I love all the image generators. So I have way too many licenses with figma and MID Journey. And um, I'm constantly having to rationalize my subscriptions that I tend to definitely um, sign up for way too late at nighttime. So

Speaker B: I know that feeling.

Speaker D: Yeah, so there's a constant uh, agent looking at my subscriptions and rationalizing them constantly. But um, you know, I think for me it's just constantly looking and learning differently with what's available in the market.

Speaker B: Just out of curiosity, what's your favorite sort of platform in terms of you know, you're using agents, are you using them in one ecosystem or have you got lots of different agents from different platforms

Speaker D: at the moment? Um, we have of one major provider that we're working with. However, if I think long term about our strategy and um, what I advise into other clients is you, you really have to think about that concentration, risk. So when people ask me this question, you know, I'm less, I think we, we think too much more of the user experience when we're thinking about the tool as opposed to what lies underneath it. So I also sit on an advisory for it, an um, AI Frontier lab here in Australia. And one of the things that we're looking at is obviously AI as infrastructure. So the big concerns that I have around AI is that concentration of risk. So what we look at when we're talking to organizations and even for my own company is where do I need to have that bespoke intelligence that I own that sits within my company versus where am I comfortable to have general models and tools available. And my business partner, he's the technical um, person, he, you know, we are looking at open weight models, we're looking at GPUs on our laptops, we're looking at all different kinds of opportunities. We're not necessarily going all in with one provider because it's a concentration risk.

Speaker B: Yeah, yeah, that's really interesting. I think that this debate's going to get greater as we go forward.

Speaker C: Yeah. And you have to think more and more, especially as capabilities improve, that companies will have locally hosted models staying fully on premise where they have their most sensitive data.

Speaker B: Yeah. And I think that, uh, also there's that interesting sort of layer of infrastructure around sort of open router, you know, that will use different models for different tasks based on different costs, based on different levels of intelligence, et cetera. You know, interestingly, I heard the other day a stat that 80% of all token usage is Chinese open models right now. And so this is kind of happening. And it's not just what I think what we hear in Australia anyway is that frontier models are the ones everyone's using, but it's. That's actually globally not the case.

Speaker D: Yeah. And I think that this is probably, you know, my other big passion and what I spend a lot of my time on is actually infrastructure of AI. So I'm very much focused on not just advantage from AI for organizations, actually. What is it for Australia?

Speaker C: Yeah.

Speaker D: So the work I do with i14, which is an AI frontier lab that's spun out of University of Melbourne, is, you know, we're looking at specialized intelligence. We're building a different architecture for AI, which, you know, currently there's the transformer models we actually. And um, that organization is based on cognitive computational AI, so it's actually building far more efficient, effective training models. You know, it is one of Australia's greatest, you know, opportunities, I think, to really have some of our own capability where we have sovereign capability, but also specialized capability for certain industries at a national level and at an organizational level. We really need to start to think about where we're going to source this intelligence going forward. We can't be completely reliant on, um, one part of the geography of the world. Um, and then, you know, the implications of that. And that's a big part of what I've spend my time focusing on that.

Speaker B: Yeah. Or two geographies,

Speaker D: multi geographies. I think it's fairly. I think Australia, um, you know, has a role to play globally on this. Um, you know, if I look at say, Canada from a Commonwealth perspective, what they rolled out with their AI for all, but the UK has rolled out, um, and you know, even Australia, what we've done previously with social media, etcetera, I really feel that Australian values, um, Australia really can play a Much bigger role than what we're playing currently.

Speaker B: Yeah, agree, agree. And really, really hope to see this. Fingers

Speaker C: crossed, but keep up the good work. Yes, exactly.

Speaker B: Now, uh, just switching gears a little bit. We like to ask all of our guests, you know, thinking a few years ahead, I don't know, three, five, probably no more than that because it's moving so fast. What are you concerned about in the future in this new AI world?

Speaker D: Yeah. So I guess like when I took my year out, I actually was like, let's look 10 years out. And I know people will say right now, oh, we can't look further than, you know, two months or six months. But actually if you take a moment and step right out and zoom out and go, well, what does 10 years look like? And my daughter will be 18 in 10 years time.

Speaker C: Yeah, yeah.

Speaker D: So I actually, and this is the reason why I set up overstory, because overstory is the canopy of the forest. It sets the conditions for everything beneath it to grow. The whole concept there is that if we actually step out and think 10 years from now what we want society, organizations, people to think, feel and do and to live, then that's the world that we should be designing for. So when I hear all these conversations now about is AI going to destroy humanity, et cetera, et cetera, I'm like, it's the wrong question.

Speaker C: Uh,

Speaker D: the question should be, what do we want this to look like? And let's design for that. And therefore, you know that 10 years out requires a far more diverse group of leaders thinking about that outcome. Gender, different ethnicity, people from all different parts of society. And, um, I guess what I'm most concerned about is when I constantly turn on my LinkedIn feed or my news feed. But it's, you know, it is seven men in the US who are ultimately kind of designing this. You know, it is a design. This is a design decision that's happening. It's not something that, you know, we are creating. It is a design decision. And I think that that is my biggest concern. So If I think 10 years out, what we need is more leaders getting involved in the conversation. More leaders is thinking about what is this going to be to the next generation? Because that's the responsibility that we all have. And I think, you know, we've been blessed with generations before us developing a world that hopefully is a little bit better. And I think that's what we need to do as well.

Speaker B: Yeah. And so with that in mind, if you're looking 10 years ahead, what is the thing that scares you? The Most, you know, what could happen if we don't design things in a way that's really focusing on the values that we hold dear? Yeah.

Speaker D: So I've read so many books on this topic. I've gone down all of the rabbit holes on this. See, there's probably two things I worry about. One is ultimately, at the moment, AI is being predominantly generated for military. That's where the investment is coming from. It is a military capability globally. And I. Therefore, that's what concerns me is actually how we're designing it to be a warfare tool. And that concerns me. The other side is actually then not so much what AI will do. It's more like when we actually have all of these AI systems integrated and suddenly it gets out of control, then what happens to our core infrastructure? So has it the ability to hack into our energy infrastructure, our financial services infrastructure, our telecommunications infrastructure? And then when those things happen, what do humans then do as a result of that? So, ah, we saw the outage recently from telecommunications, and no one's getting the train. And suddenly it was an uproar. You saw when Covid happened and we couldn't get access to toilet paper. Everyone was going mental. So, you know, humans have this crazy way of how we, you know, deal with. And I think that. So therefore, it's like, it's more, for me, it's less. Less what it'll do to us and more what we will do as a result of.

Speaker B: Yeah, really interesting. And if we flip that on its head and we say, what are you most excited about? What's that?

Speaker D: It's funny, I actually asked AI recently this question. I was, uh, like, let's find all the good things AI has done. Like, I want an Excel spreadsheet. I want an audit list around the world. Find me everything. Give me all of the reference points. Then I came back with, like, probably I was a little bit underwhelmed. Oh, no. And then I answered around all the things that I got wrong, and it was like, straight away able to list them all out.

Speaker B: Interesting,

Speaker D: isn't it? That was a bit of an oh, no moment. But, you know, I think if we. If we design it right, if we design it so that it can enable greater efficiency, uh, in the sense of, like, it can. You know, you get your ability to have. Be more creative, more curious in your thinking. You can eliminate sort of all a lot of the mundane stuff that we've ended up creating for ourselves. And that's really good. Obviously, the one everyone talks about is medical, et cetera. I think that there's opportunities around in particular the energy transition. I know there's a lot of concerns around data centers and the increase of energy as a result of AI, but I also think that it can also help speed up the energy transition. And I think, think someone who's worked in energy, you know, when you think about the renewables and the grid and what you need, uh, it's ultimately a big massive data system. So you need AI in that world. So there's a lot of good positivity that we AI has the ability to do. It's just that again, it's around ultimately it's around leaders in organizations and in governments who are making decisions today. How do we make the decisions that design AI to deliver the right outcomes for us?

Speaker B: Yeah, absolutely. So it's really every leader leaning in, um, and thinking about that. And I like your challenge on the 10 years because I think we have sort of defaulted back into thinking more short term, um, just because things are moving so fast and so it's hard to imagine. But you're right if you don't think 10 years ahead when you're designing

Speaker C: because it is easy to uh, maybe even not consciously use AI as an excuse about the fast, you know. Oh no, we can't have a strategy that looks beyond the

Speaker B: next

Speaker C: quarter. But actually that is a cop out for genuine deep thought.

Speaker B: Yeah, it really is. So that's, thank you for that. That's a good pushback. Well, Orla, uh, it's been such a great conversation. Thank you so much. We've really, really enjoyed it. Now if our listeners want to find out more about you and more about overstory, where would they go?

Speaker D: So our email, our website is theover storygroup.com

Speaker B: Great, perfect. And so we'll put that on the show notes page. So if people uh, would like to find out more and contact you, then they can go there.

Speaker C: Thank you so much Ora.

Speaker B: Yeah, it's been really brilliant.

Speaker D: Thank you. It's been a pleasure.

Speaker E: M. Your website can do a lot more than you think with base 44. Describe what you want in your own words and build a website that looks your way and works your way. Let people book with you or buy from you with hosting already included. Then help more people discover your site with built in search tools, social content and Google Ads. Build it, get it live and start sharing. Go to base44dot.