This week: a special on artificial intelligence in Australian organisations and AI fluency with Deloitte’s AI Lead, Dr Kellie Nuttall.
Sandra Peter (Sydney Business Insights) and Kai Riemer (Digital Futures Research Group) meet once a week to put their own spin on news that is impacting the future of business in The Future, This Week.
Follow the podcast in your favourite app
Join us for the AI fluency sprint, our exclusive online experience and build an understanding of the opportunities, risks and limits of AI.
Our guest this week
Links for the curious
Follow the show on Apple Podcasts, Spotify, Overcast, Google Podcasts, Pocket Casts or wherever you get your podcasts. You can follow Sydney Business Insights on Flipboard, LinkedIn, Twitter and WeChat to keep updated with our latest insights.
Send us your news ideas to firstname.lastname@example.org.
Music by Cinephonix.
Disclaimer We'd like to advise that the following program may contain real news, occasional philosophy and ideas that may offend some listeners.
Kellie Nuttall So you could say 'look, there's a lot of focus on how much investment is going into AI but you know, what about how many companies are actually adopting AI rather than just investing into it?'
Sandra This week, a special on artificial intelligence in Australian organisations and AI fluency with Dr Kellie Nuttall, Deloitte's AI Lead.
Intro From The University of Sydney Business School, this is Sydney Business Insights, an initiative that explores the future of business. And you're listening to The Future, This Week, where Sandra Peter and Kai Riemer sit down every week to rethink trends in technology and business.
Sandra There's been a recent string of articles in the AFR and other publications around the low adoption of AI across Australia. There's been an article last week talking about underinvestment in AI leaving Australia vulnerable towards becoming even a low-skilled economy. And today, there's been an article, again, around Australia being a laggard, saying that only about 34% of firms are using AI across their organisations.
Kai And to be clear, here, we're talking about organisations using AI, not necessarily the startup sector or investments in AI firms, but companies adopting it as a way of improving their operations, as a way of improving their business models.
Sandra But companies lagging behind in AI, we've been talking a bit about AI on the podcast, but we've been talking less about AI in organisations, so I think we should get some help talking about this.
Kai And there couldn't be anyone better than Kellie Nuttall who, you know, heads the AI Institute at Deloitte.
Kellie Nuttall Thank you guys. Thanks for having me.
Sandra Kellie, you've been involved with AI for a while, you're running AI for Deloitte. Just for our audience who might not be familiar with your work, tell us a bit about how you've come across AI and what you do these days.
Kellie Nuttall So, my career has always been in the world of how you use data to solve problems. And over time, data moved from survey-collected data into big data, and then you'd start to do analytics on it. And then obviously, the next iteration on analytics was AI. So, I've been at Deloitte for 10 years now working in the artificial intelligence domain around helping organisations get value out of data. That's moved far beyond just business intelligence tools and basic analytics into much more analytics-at-scale using these technologies. So just to be clear, when we're talking about artificial intelligence, what we mean is using a suite of different technologies to mirror human skills, human intelligence attributes. Basically being able to solve problems in a much more scaled and efficient way than humans can. And there's quite a few different ways to apply AI to solve different problems in business, it's kept me very busy helping organisations build AI to solve problems or help them build their strategies to become AI-fuelled organisations.
Kai I want to go to the data aspect, because you're putting it right in line of a progression from simple data analysis, data analytics, big data to then AI. And I think many people come to AI in there from the side, they think AI, magic technology, thinking machines, magic is happening, when in fact, it's just a way of deriving patterns from large amounts of data, and data is really key here. It's not like traditional computing, where someone sits down and builds a procedure, tells the computer what to do. We're really talking about computers deriving insights or deriving patterns from data that can then be used to do all the kinds of things that we are now familiar with and talking about in that field as AI, as you said, simulating skills that we normally associate with human intelligence.
Kellie Nuttall Yeah, absolutely. Data is basically the lifeblood of any algorithm and it's only as good at producing a prediction or an output based on its inputs. The great thing about machines is that they can see patterns connecting hundreds of variables, and how they interact with each other and the relationships in them that the human brain cannot. But if that's based on not that many data points, it's unlikely that it's going to be highly predictive or accurate. So, you know, when we think about things like healthcare, you know, being able to predict clinical conditions that have comorbid effects with other clinical conditions, even one hospital's worth of data probably isn't enough to show one patient that has three of those conditions together, you might actually have to go across the world's hospital population of data to really actually be able to predict that in the future. So data is really hugely important, and AI is really good at predicting things when there is good data, but it's only able to predict things that have happened before in some ways too. So if you think about things like COVID, you know, a lot of the time people thought, 'could AI have helped us?' AI can help us respond to COVID, but it might not have been able to predict it because something of that nature had never happened before so we wouldn't have seen that trend in the data. It's all about pattern recognition, where it's seen it before and where it can predict it playing out in the future.
Sandra In this case, AI could actually help a lot of organisations because there are many things that we have seen before, like COVID happens, you know, once every 100 years, or once every two years, remains to be seen. But there's a lot of data that theoretically companies could use to implement AI. Yet, all these articles we've been seeing over the last couple of weeks say that Australia is really in the very early days, that implementation in many Australian companies and in many Australian industries are lagging far behind everyone else. What's going on here? What are some of the barriers? Why aren't companies using it?
Kellie Nuttall So we do, we think it's the largest AI survey around the globe every year in Deloitte, called our State of AI report. And what we see is, in general, most countries are now moving into the scaling AI phase. They've experimented with it, they've tried different use cases, they're now fully into, 'how do we take our AI strategy and scale it to become AI enabled across our whole business.' We don't see that trend in Australia, we still see a whole lot of experimentation, I would say we're at that stage, just before scaling. We're seeing some of the larger organisations starting to think about strategies for how they become more AI-enabled across their business. But largely, we still see so many proof-of-concepts going on that don't go into production. You know, I've used this stat before, that, I think, you know, in Australia, it would be fair to say that only two out of five, maybe one out of five, proof of concepts ever goes anywhere. And that's okay, it's okay to fail and try something and not take it into production. But I think a lot of the time, it's because they've proven it and they think that's the answer, and they can just keep running it in this proof-of-concept type environment. But the skills that it takes to actually operate AI and put it into production and continuously model and monitor it and feed it new features, that takes a different skill set, and that's often not what we see. in Australia. It's led by data scientists and maintained by data scientists. And that could be an issue as well, like it really needs to go into a production environment and run like any proper IT solution, or product does. But other barriers, analytics strategy in general. So what we see is, you know, I think about a third of companies in Australia have an AI strategy. And to really become AI-driven across your business, you need to be understanding 'how does this technology support me to achieve my strategic objectives?' It needs to closely align to 'we are a business that delivers value in X, Y, Z ways, and how can these technologies help us deliver those things?' The strategy then needs to help them understand what foundations do they need to put in place to execute on those value use cases. So when I'm talking foundations, do you have the right data? Do you have the right people capability? Do you have the right governance and operating models to scale it? And you know, thinking about AI for the enterprise, not just for my part of the business. We see this so often where people have built a really interesting tool that maybe supports a compliance use case. So you know, maybe potentially monitoring calls in the call centre using natural language processing to see whether compliance statements have been read out, excellent use case. But could you use the same logs then to detect customer complaint themes and start to help the marketing people or the customer service people, not just the compliance people? So when we think about scaling AI, we need to think about building things that can be evolved in scale to help other use cases across the environment in a connected way, as well. We see a lot of focus on fluency being important. So when I talk about AI fluency, you know, universally, I would say the majority of senior leaders in most organisations do not understand artificial intelligence. They've heard of it, they're afraid to admit what they don't know. We do this thing in our AI fluency workshops, where we ask people what the first thing that comes to mind is when we think about AI, and by far, the most common response is Terminator.
Kai We do the same in class.
Sandra That's exactly what we're seeing, you know, in our Executive MBA classes and in the cohorts that we teach. It's the Terminator and it's Westworld.
Kellie Nuttall Yeah, well, it's wild, right? Because at the end of the day, where we are with AI is it's still very narrow and it's very good at doing one thing very, very well. If you've trained an algorithm to detect an anomaly in an x-ray, it can do that incredibly well. If you try to get it to then navigate to your house, you know, a child could probably do that better. It doesn't transfer to different contexts yet, you know, and there's been a lot of discussion in the media lately about foundational models as well, where we're starting to see more common datasets used to train multiple algorithms. And I think that's a step of where we're starting to see the shift from narrow AI until a bit broader, but for the most part in most organisations, AI is about very narrow focus to one thing that can be incredibly helpful. So, I'll give you an example, if you've got someone auditing tax receipts that have been uploaded into a system, and it's like hundreds pf 1000s, or millions, or whatever, of those every year that you'd have to check, you don't, you send a human to do an audit. But now with something like computer vision and natural language processing, you can literally apply that to the entire population of receipts to detect whether it's been compliant. And so it's just doing its one task, but it's doing it incredibly efficiently. And it's doing it across an entire population of data rather than a human having to do that 10 times a day. But it's not then predicting whether you're going to make your cash flow predictions for the month, it sticks in its tracks that it stays there.
Kai Kellie, before we talk fluency, let me go back to that you said: 30%, or about, companies in Australia have some kind of AI strategy. Now, as a keen listener, I would immediately jump to, 'oh, these must be the big organisations because the small and medium sized, they have no clue.' Who are these companies who have strategies in your experience?
Kellie Nuttall If you said is it one industry that stood out is one type of size of business.
Kellie Nuttall It's not any of those things. It's the organisations that honestly just see value in these technologies and they normally are led by someone who is quite a pioneering leader, who understands that this technology is going to be transformative in the same way that the internet was for business. And they know that they have no other choice than to embrace it.
Kai So we can't say that, you know, the big ones are good at it, and the small ones need help, right? There's quite a number of large organisations who have not really made any inroads into applying AI, where you would think surely there must be enough of the kind of data that would benefit from automation, from pattern recognition, from the deep kind of analytics, to improve efficiency across business processes.
Kellie Nuttall So certainly, the larger organisations are by no means more progressed at all. In some ways, it's actually a harder ship to turn, because they're so big, you know, so even banking organisations are probably at the forefront in terms of an industry around progressiveness and AI and it's largely driven by the fact that the regulations are going to require them to get on the front foot with this stuff in terms of anti-money laundering, and cyber, et cetera. But in general, everyone still struggles with legacy systems with, you know, when it comes to personalisation in a bank, that would be the next place to focus AI. But then if you think about the way banks have been traditionally set up, they've been set up in silos of businesses, for products, for insurance versus banking, versus wealth management. Whereas with personalisation at scale, you need to be thinking about a horizontal, and a single view of the customer across all those products. Now, that's a challenge when you've got politics and hierarchies and structures set up around those verticals.
Kai That's the crux, right, the integration of the data to make data from different divisions, but also from different business processes talk to each other, integrate them. So you can actually run them into a machine learning algorithm, for example, or a deep learning algorithm. And large organisations are sometimes behind the eight ball because they have so many legacy systems that are held together by spaghetti duct tape code, where you know, little overnight patches, you know, pump data from one system to another, is not necessarily the kind of environment where you just come in and, you know, pull the data and run it into a deep learning algorithm. So sometimes I think smaller companies are actually at an advantage here.
Kellie Nuttall I completely agree and one of the greatest quotes I ever heard from a client that was in one of my workshops was said, "Kellie, you know, the reason that God created the world in seven days was because there was no legacy systems." Such a great way of summing up the situation we see, and it's so easy. So when we work with organisations around where are you at in your AI maturity, and there might be pockets in your siloed parts of the organisation, but where are you at, like an enterprise level? And, you know, AI is a number five out of five, AI-fuelled, and it's transforming every part of the business, you know, who's doing that, oh it's Netflix, it's Spotify, it's okay not to get there it costs a bit of money, costs, you know, a bit of social capital and political will. And it's okay, just to move up the curve. So how can you take it from being siloed to being a few use cases that are more enterprise wide?
Kai Yeah, well, my favourite example, you know, ‘look at TikTok, look what TikTok is doing’. That's great if you're TikTok but doesn't really help a large energy business in Australia.
Kellie Nuttall Totally.
Kai The fact that, you know, we have a TikTok algorithm.
Kellie Nuttall No, absolutely. And the other thing is, it's not the technology that's often the hard bit, it's the change and the people side of things. And you know, you can provide the most interesting AI application and put it in the hands of someone that's never dealt with AI in their life, and it can be amazing, or it can actually be like the best algorithm, and they completely reject it because they haven't been taken on the journey and they're not used to using that sort of tooling in their day-to-day work. And I think that's where organisations are going, the really high value use cases for AI or applications for AI, are for people who are in frontline roles, operations staff, traffic managers, call centre agents, it's giving them you know, a little AI sidekick to help them do what they do better. But if you don't design it with those people in mind from day one, and you just have data scientists working in isolation, you don't often get the buy in in the adoption. And that is actually then a foul on many parts because then those people are saying 'we tried AI, it wasn't very helpful,' it gets a bad name in the organisation, no one wants to invest in it again. User design and user experience design should be front and centre with AI and I think that's another blocker. We see people building AI and little vacuums and because it's more of a cool science experiment than something about what would actually make that job better for that person, take the robot out of the human rather than everyone's like, 'Oh, they're gonna take my job.' Well, no, actually, this little sidekick of yours, he's gonna take the robot part out of your job and allow you to be more human, add more value, spend more time with your customers, getting more interesting information. So yeah, there's a lot of value in it. I think, a lot of the organisations we work with underinvest in the change required to support these big transformation programs, they focus on the technology, they don't focus enough on the change required. It does sound to me like it's coming back to strategy, then. Yep, good strategy is not just technology, that it's people, process, technology. data. So all of those things are foundations, right? You got to have the right operating model, the right governance structures, the right data quality, the right data coming in, all foundational. But if you don't actually then execute on that strategy, in line with use cases. So what we would always recommend, and it works really well, is have a, you know, two three-year transformational roadmap of where you're going, but stage it by use cases, where are the highest use cases that you can deliver value for your organisation? And then get the foundations right to support that, then move to the next one. As you put that first one into production, what did you learn about your foundations? What do you need to invest in? What do you need to reprioritise or change? Move to the next use case. And so what we're also seeing is, it's a bit of a different investment strategy for execs and boards and just going up like an SAP transformation and saying, I need, you know, this many million dollars to do this ERP replacement. It's an untested technology, execs don't know exactly what they're gonna get. So how can you deliver a more incremental approach to investing in this technology that allows you to test, learn, go back and get the next bite sized chunk of investment to build the confidence? I think that works a lot better. And I think that's also where we kind of first got together because, I think, in terms of blockers, we were both seeing the fact that these execs who were meant to implement and lead these strategies and get involved with AI really had little understanding of just how much goes into it, let alone what the technology can or cannot do. But then what would it take for their organisations to actually implement AI, and implement the AI at scale, and not just in a couple of pilots here and there. Yeah. And imagine the power of being a senior executive that actually can determine I would like you to focus it here, because this is the biggest lever to us delivering value for our stakeholders, shareholders, customers, rather than having the other end of it where you just basically have someone come up to you and say, 'we want to spend $20 million on an untested technology' There's power in understanding, there's power and being able to challenge, and I think executives don't need to know how to code, they need to understand, in general, what skills, or what are these technologies good at. And the reason it's called AI is because there's many technologies that sit under there, which also makes it a bit nebulous and hard to understand, but it's ultimately about, you know, can it predict, can it categorise things, can it see things and turn it into categories, can it automate something? They're all like little human skills and there's different technologies for every one, if you understand, essentially, what these skills are, and the problems that they could help you solve, and point them at the problems that actually matter to your business, that's where true transformational value comes from.
Kai Executives, people who have a busy day job, they have to focus on whatever industry they're in, they're on top of their own business, competitors technologies in their field. It's not easy to grasp what is new and what is novel about AI. And there's a double thing here, right? On the one hand, it's a new computing paradigm that starts with data and not procedures. And on the other hand, we have all these pop culture references, you know, like Terminator, Westworld, and the like, which are not helpful to get a head start on the topic. So you have to unlearn that and then you have to wrap your head around, at least the basics and the conceptual foundations of what this new kind of computing is like.
Kellie Nuttall Exactly. The thing I think is still really interesting is, you can look at some of these systems, and because they do such inherently human characteristics to think that they've got cognition in mind. And they don't. There's still just a set of coding algorithms that are very like, if then, if this do that, if this categorise this, if that predict this, you know what I mean, but because they look human, we think, 'Oh, they're going to take my job.' They're not, they're just doing a very, you know, neatly programmed set of things to do, and they do it incredibly well.
Sandra So I'm conscious Kellie that often when we talk about articles like this one, or the state of organisations, you know, not really doing enough about AI or the opportunities that are missed in Australia and so on. I'm reminded Erica on your team always says, you know, 'we spend a lot of time looking at the shovel and complaining about what's not being done.' So how do we get better at this? How do we pick it up?
Kellie Nuttall Yeah, no, let's pick it up, because it's worth picking up. And I think the thing at the end of the day is AI is so incredible, when we can do it well, and we can point it at the problems that make a huge impact in our lives. And if you just think about your day-to-day life, now you wake up, you open your phone with your facial recognition, your emails are being filtered from spam, because of natural language processing. I get to drive to work in a non-congested way, because of Google Maps, I get some great discounts in my clothing purchases, which my husband argues I spent too much money on, but it's because they personalised my marketing deals. It's valuable to me, right? So how could we apply the same superpowered awesome skills to our work life to deliver value for ourselves, for our customers, for our stakeholders, I just think it's too important not to do. And the other thing is, everything that is a barrier can be overcome, we can develop a strategy to help you execute on your AI roadmap in a way that delivers value incrementally and quickly. We can help you understand ethical frameworks to put around the way you do AI to protect you and your reputation, we can help you become more fluent. And this is the thing, you know, we are so passionate about, which is, how I came to meet you find people in the first place, right? Because what I was learning is that executives don't understand AI, so how could we partner with a university like yourself to actually help executives understand AI and not people in the world of AI, but people who are in proximity to AI that need to understand it to drive better strategy?
Sandra And what better way to do it really than, you know, having a university and a consulting company getting together. We used to say, you know, we're going to make this rigorous, practical and no bullshit, because on the one hand, it does need to be rigorous, but it does need to be very, very practical.
Kai And it is, unfortunately, a gold rush kind of sector that is full of snake oil, and bullshit.
Kellie Nuttall Full of hype.
Kai Yeah, we have to cut through that.
Kellie Nuttall Totally. And consultants love to speak a lot of consulting language, so you're helping me cut my bullshit as well and make it more pragmatic. But, I think, you know, at the end of the day, we're so excited about this, because we want it to be practical and applicable to your roles to really be able to understand these technologies, what they are, what they're good at, what they're not good at, but how they'll deliver value for you and your business.
Kai So the idea is really not just to shovel, but to actually, metaphorically speaking, sell the shovels too so other people can shovel themselves, right? Bring AI fluency to executives, to the people in business who need to make decisions about AI, who have great ideas about their business, but not necessarily about the capabilities that this new form of technology can deliver for them.
Kellie Nuttall The work you guys do here is so important because artificial intelligence is one of the emerging techs coming this way, you know, for executives. The fact that we're now in a world of lifelong learning and being able to actually really understand these new technologies and how they apply to the work we're seeing is not just for someone at the start of their career, it's for lifelong, because they're coming at us all the time. And so I think that's an awesome opportunity.
Sandra And it's not just for technical people, I mean, a lot of the courses that are out there on AI around programming, are around data, are around how do you model. We found, as did you guys, that there's very little out there that helps executives and organisations make inroads into AI that go beyond the technical applications, that try to build an understanding of opportunities, of risks of limits, and AI and…
Kai Basically AI for everyone in business, not just the tech people, right?
Kellie Nuttall No, tech, as you say, they have a lot of streams that support them anyway. This is really around AI applied, right? AI applied where it delivers value. And I think that is, unfortunately, something that we don't see enough of, is AI can be a game changer but let's point it at the things that matter most your organisation and it's not just about differentiation anymore, it's about relevance. All companies are going to need to be thinking about artificial intelligence if they want to stay relevant. Doesn't matter whether you're in government, or private sector, because every citizen expects the same level of service, no matter who they're dealing with. You know, AI is definitely here to stay, it's here now, it's here in our private lives. It could be so impactful in our work lives if we could just think about how to build that fluency, have a good strategy around it and let's get Australia back on the map in terms of AI innovation.
Kai And you do see some really good examples of successful use cases of AI. So I'm really curious, what are your top applications that you've been seeing in recent times?
Kellie Nuttall Yeah, I think where we see the most interesting opportunities are in a few domains. So intelligent operations is becoming a really, really big theme. Whether you're someone responding to elective surgery backlog following COVID and you're trying to use AI to optimise patient flow to try and optimise getting as many people through the backlog as possible, through to intelligent operations through the court system. So basically, if you think about anything that's a value chain, how do you unblock value chains using AI and data along that?
Kai We've just spent almost three hours this morning to get from a plane seat out of the airport, everything clogged up, you know, that reminds me very much of the sort of problem that you're talking about.
Kellie Nuttall Yeah, and I mean, that's definitely my favourite sweet spot to talk about is digital twin technology and how we're starting to see platforms where we're bringing all of this information into one system to try and actually get a situational view of what's happening in a system or a network, and then using AI on top of it in real time to optimise that network. So we don't have to sit in congestion anymore, or the plane's not delayed, and we can actually predict that's going to happen in advance and do something about it.
Sandra Any industries that you'd say are kind of ahead or doing particularly well, and any that you're seeing have the biggest potential to, to make moves?
Kellie Nuttall Yeah, I mean, I think potential is everywhere. But look, as I said, the banking and financial services sector has been more on the front foot just due to the regulatory requirements to be more proactive around that, but are still now on the wave of, 'okay, how do we take it into customer experience?' It's been mostly compliance and regulatory driven.
Kai And also the banking industry in Australia just recently had to really go through a big digital upgrade, build new infrastructure, many of our banks in the last 15 years have put in place the kind of infrastructure that now benefits them in doing these kinds of things. That's not the same in other industries, though.
Kellie Nuttall No, well, and we've we've also seen, you know, Commonwealth Bank come out and take a stake in h2o.ai, which is a really easy to use AI platform that's really around trying to democratise personalisation at scale across their bank to make them feel like a small bank and be more personal. So I think they're definitely on that cusp of being really AI-driven. Mining and resources is another one that again, they can't afford not to if they want to remain competitive. And obviously with a big shift to decarbonisation, they need to be thinking about how they use these technologies to optimise every part of how they operate, not just through a throughput or safety, now through a carbon lens as well, right, so that's another interesting driver. And then retail business is another one that are really, really starting to hit this straps in AI and starting to make some real progress in the personalisation agenda around customer experience. People are always like surprised, but some of the best applications of AI I see are in government in Australia, and when I talk about Australian government, I mean, compared to globally, I see some really, because I'm connected into our global network, some of our coolest, most exciting applications of AI in terms of predicting resilience after natural disasters, the digital twin work I'm talking about, all of this stuff is happening in governments. And that's super exciting.
Sandra And you will be sharing some of those insights and you know, good use cases throughout the course, throughout the online sprint that we've got planned. Given we started this conversation around the fact that only 34% of firms in Australia are using AI across their operations, there's a huge opportunity here.
Kai So let's put it out there. The first one of those starts on the 12th of August. There will be one more later in the year and we're super excited that we're doing this with you guys.
Kellie Nuttall Yeah, we're delighted. We can't wait, it's gonna be great. We're going to teach you actually how to understand really where to focus your AI not just 'how do I do computer vision,' no actually, what problem are we trying to solve and where may computer vision, or another technology actually helped you solve it?
Sandra So come sprint with us from the 12th of August. Designed, you know, for the busy executive.
Kai Exactly. Move the dial a little on getting Australia to adopt this transformative, game-changing technology. Pick up the shovel!
Kellie Nuttall Yeah, pick up the shovel, and what you'll find is it's actually not that scary, it's actually a lot of fun, and we're gonna have some fun with us. Please come and join us.
Sandra Kellie, as always, thank you so much for having a chat to us today.
Kellie Nuttall So welcome. Always a delight.
Kai Thank you so much.
Sandra See everyone online.
Kai See you soon.
Outro You've been listening to The Future, This Week from the University of Sydney Business School. Sandra Peter is the Director of Sydney Business Insights and Kai Riemer is Professor of Information Technology and Organisation. Connect with us on LinkedIn, Twitter and WeChat and follow, like, or leave us a rating wherever you get your podcasts. If you have any weird or wonderful topics for us to discuss, send them to email@example.com.
Sandra Also, why in God's name are they writing this article at 1am? It was published at 01:08am!
Kellie Nuttall There was probably an AI that wrote it.
Sandra It does seem to have been an AI that wrote it.