With between 79 and 98 percent of jobs in Australia predicted to be automated by 2030, enabling vast productivity gains, how will this change the nature of work? Generative AI (gen AI) will have different effects on different sectors; however, early adopters in any sector can drive positive change—at the company and employee level—not only in productivity but also in the quality of work. Top-down thinking will be critical for a tech strategy that delivers value and enables an organization to scale. But where will this leave the individual? There will be an increasing demand for people with technical skills, social-emotional competencies, critical thinking and complex problem-solving capabilities, and good interpersonal skills; basic digital literacy will be necessary for almost every job.
In this episode of the Future of Asia Podcast, McKinsey’s Debbi Cheong speaks with McKinsey partner Seckin Ungur and associate partner Jacob Johnson to take a closer look at gen AI and the future of work in the Australian context. An edited transcript of the conversation follows.
Debbi Cheong: Hello, and welcome to a new episode of the McKinsey Future of Asia Podcast. I’m Debbi Cheong, and I will be your host for today. I’m delighted to welcome two very special guests, Seckin Ungur and Jacob Johnson, who will be talking about the future of work in Australia and, in particular, the impact of gen AI on the workforce and productivity. I’m very happy to have both of you join us today. Before we start, please introduce yourselves to our audience.
Seckin Ungur: Hi, I’m Seckin Ungur, and I’m a partner in McKinsey’s Sydney office, where I lead our Education Practice in Australia. I’ve done a lot of research on the impact of automation on the future of work, and most recently coauthored a report on gen AI and the future of work in Australia.
Jacob Johnson: Hi, I’m Jacob Johnson, associate partner in the lovely bush capital of Canberra, Australia, where I lead much of our work in the public sector. I’m excited to give a practitioner’s view in terms of helping different clients adopt different types of gen AI and other types of analytics and digital products over the course of the last few years.
Debbi Cheong: To start off, let’s go big picture. Seckin, what do you think the impact of gen AI will be on productivity?
Seckin Ungur: I think the potential is huge. What we know from our report is that even with the technology that’s available to us today, 62 percent of tasks could potentially be automated right now. And, by 2030, it could rise to anywhere between 79 and 98 percent. That offers massive potential for productivity improvements, anywhere in the order of between 0.2 to 4.1 percent per year, with gen AI contributing roughly a quarter to half of that. But it all depends on how bold we are in adoption.
We’ve modeled a couple of different scenarios in our report. In the midpoint scenario, we assume that about 27 percent of tasks will be automated by 2030. So, a bit short of the 62 percent potential. But if we were able to achieve even half of the potential, we would get Australia back to productivity growth of 2 percent per year, which is in line with its heyday post-1990s. So, there is massive potential, but it all depends on our ability to adopt the technology and adopt it in a way that actually drives productivity improvements.
Jacob Johnson: The huge difference here is going to be adoption. Looking at productivity growth in Australia over the past 20 years, most of it has been a mining productivity boom. If we think about the gen AI type of productivity gain, it’s quite different and provides huge potential for broad-reaching consequences across all skill and job types in the economy. I think the challenge here is that the ranges we’re giving are huge. Obviously, we don’t know exactly what’s going to happen. If I look at some practitioner examples, we’ve got cases where people are using free text analysis to get survey insights, and here we’re reducing analysis times from three months to three minutes. You’re seeing a drastic reduction in specific tasks. And the exciting part about this is that you’re seeing improvements not just in the quantity of tasks but also in the quality.
In the example I just shared, there are teams of people who used to spend time trying to synthesize results. Now, with the help of a tool, they can have the prevalence and magnitude of different insights and spend their time thinking about what they should do to change the course curriculum of these survey trainings. The big difference here is that this isn’t going to be a story of everyone improving or becoming more productive at the same rate. This is going to become a highly divergent story, with some early adopters driving radical change and automating these tasks, while others lag behind and don’t reap these benefits as quickly.
Debbi Cheong: I know that both of you talked about how productivity is impacted across different skill and job types. Following on that point, can you explain how different sectors of the Australian economy are being impacted by gen AI? What skill shifts do you see emerging?
Seckin Ungur: If we take three very different sectors in the Australian economy—retail trade, financial services insurance, and the public sector—what we see is that there’s massive potential in all of these sectors, but in different ways. For example, in retail trade, gen AI has the ability to introduce much greater personalization in customer service. It can improve inventory and supply chain management, and augment functions such as customer service and marketing.
Whereas, if you look at financial services insurance, the types of things that gen AI can help improve are risk assessment, fraud detection, software development, and also customer service. Roughly a third of those task hours in that sector could potentially be automated by 2030.
In the public sector, gen AI has the potential to enhance education delivery and improve outcomes; improve interactions with citizens, for example, by augmenting customer call center operations; and improve processing times for welfare or tax, which can help with financial analysis and R&D. This can lead to significant productivity gains but also to an improvement in the accuracy of services and the speed of service delivery for citizens.
If we think about the skill shifts that this leads to, there’s going to be increasing demand for skills that combine technological proficiency with social-emotional competencies. So, for example, healthcare workers will need to develop digital skills alongside their traditional medical expertise. We also know the demand for basic cognitive skills is decreasing, while roles that require complex problem solving, critical thinking, and, more important, interpersonal skills are on the rise. Educational requirements are also shifting. There is a growing need for higher education and specialized training, in particular in STEM, healthcare, and professional services, which are industries and occupations where we expect to see significant growth going forward.
Jacob Johnson: If we think about where there’s excitement around what one could call resilient and growing conventions in science and technology, healthcare, and (not meaning to be self-serving) potentially in professional services, we’re looking at net-demand increases by 2030 of up to a million and a half additional jobs. Again, I think here the nature of the work will be quite different from such conventions.
If we take ourselves [McKinsey] as an example, we have used gen AI to change the way we work with our clients. Our whole knowledge management system is available through an internal gen AI system. We think about how we can leverage each individual analyst who is starting so that they’re not just a person doing something but have the organization’s entire knowledge behind them.
When we think about these jobs and these sectors, even though they’re growing, the type of work that individuals in these jobs will do is quite different. Of course, there’s the stalled but rising category coming after that, things that will continue to grow but maybe not as quickly because they won’t necessarily be automated by gen AI—things like mechanical installation, repair, and rebounding as infrastructure grows.
I think it’s helpful here to take a step back and think that, although we’re talking about gen AI, let’s not forget about traditional AI and exciting things like robotics. Right now, we’re focusing specifically on what gen AI will do to different types of jobs. However, certain types of infrastructure and construction jobs may change radically due to innovations in AI combined with robotics.
I think the last category where we could really see jobs continue to shrink is in sectors where they’re already declining. This is across things that you would expect in different support roles, customer-facing roles, and shifting to e-commerce. Again, importantly, what does this mean for those who remain in those types of support roles? We must change the way we think about how we work—not just as individuals, but also as organizations.
Debbi Cheong: Jacob, what insights have you gained from working with organizations to scale and create value through gen AI?
Jacob Johnson: Some things are surprising to me and are different from assumptions I had a couple of years ago. And then there are other things that have really reaffirmed how I think about digital and analytics and the way they are changing how we work. The one that’s different is that I’ve seen certain gen AI use cases in organizations lead to rapid adoption increases that I haven’t seen in other products. These are things that all of us could use in our day-to-day lives, such as concision and content generation and being able to take large amounts of information and simplify them to share with our bosses or committees.
The reason why I’ve seen these rising quickly in some pockets of organizations is because the value tends to creep to the individual. For example, I’ve seen people who support senior executives being able to get a higher quantity of their work done and to be much faster in synthesizing things and getting them up to their bosses than they could six months or certainly a year or two ago. I think that’s a big difference from how we’ve seen technology adoption in organizations historically, where we were trying to do large-scale digital transformations and it was hard to get individuals to change their behavior to something that might benefit the whole organization.
There’s another category that has reaffirmed certain heartfelt lessons for driving organizational change, and that is that change management still matters—being able to role-model change, being able to help people understand the change that needs to happen, being able to train them in new ways of working, and setting up incentives that change behavior. One example of the way that an organization can rewire behavior and step up the adoption of different types of gen AI is in helping people understand incrementally how it can make their lives better and improve the performance of an organization.
So, if you take a customer-service, customer-facing organization, you can use gen AI to understand historical interactions with customers and then give feedback to customer-service representatives on how they behaved in the past. Step two, you can give them a sense of what customers they might interact with and what interactions they could change in their behavior with those customers. Step three might be that you do that in real time and provide feedback to an individual on how they can change their interaction with the customer. Step four might be that you let those interactions go directly to the customer. It’s not a foregone conclusion that we need to take all these steps, but rather look at what the right thing is for the organization to do and how it wants to leverage this technology to improve its interactions with customers and the way its employees work.
Seckin Ungur: The thing that strikes me is the number of organizations that are passionately rolling out use cases, proofs of concept, and pilots but not thinking strategically about the value that this could add to the organization. How is my organization, my industry being disrupted or has the potential to be disrupted by gen AI? How do I really get value out of that? I’d love it if more organizations were thinking about that top-down strategy behind what they want to get out of gen AI—because, in the absence of that, productivity will come, but you almost leave what happens up to chance and how that productivity is actually deployed. The default could be that it gets absorbed into employee experience.
Employees save time on tasks that they would otherwise have been doing manually. That time can get reinvested into doing more of what they were doing, or doing it better, and providing a better customer experience. Or it can lead not to more or better work but just to a better employee experience. That may well be a legitimate choice, but my push to organizations would be to let it be a choice. Be deliberate about whether you want that time saved through the productivity that gen AI or other automation technology can offer. You need to decide, when you choose, whether you want the result of implementing this technology to go into employee experience or expect that to be more or better output, or fewer people doing those tasks.
In the absence of top-down strategic thinking, the potential for the technology to deliver value for the organization and the customer is limited. Bunches of pilots and use cases that the tech side of the organization is really excited about often don’t scale, because if they haven’t been done in collaboration with the business with the intent of driving business value, it’s difficult to make the business case to invest in scaling that technology and having it drive broader productivity for the organization and, ultimately, for our economy.
Jacob Johnson: For me, Seckin, you’ve hit the nail on the head here. The incremental productivity shifts for individuals, pilots, and use cases will not necessarily lead to business-model transformation. I’ve seen different clients wrestle with this, but that’s not an excuse not to get started. Even though the technology side of the business can get excited about tech for tech’s sake, it is important for the organization to figure out how they’re going to deal with certain concerns that people have about this technology.
They need to look at the limitations of the technology and work through those with certain use cases or clients who are using on-premises gen AI solutions. In this way, they could bring large language models into their own environments and disconnect them from the internet so that they can partition data and information and not share it out again. That’s something that will be important for them to do, whether it’s a small or a big thing, and they need to work through that. It builds a certain level of trust in the organization; they can maintain certain levels of integrity and safeguards that need to be put in place. We’d be silly to think that those things will take care of themselves.
Seckin, to the point you made on the business-model transformation, I think this will vary widely by industry. Again, we’re focusing on gen AI, which is largely in knowledge work and can be powerful in making things more concise and generate content. But if you’re talking about productivity and physical things like mining operations, traditional AI and robotics are going to play much more of a role than gen AI. I think organizations and leaders need to look into this concern—where’s the value going to lie and what role is gen AI going to play in that versus other new and innovative technologies?
Seckin Ungur: Also, what other elements of my operating model do I need to change to reap the value from gen AI? To scale some of these use cases, organizations need to think very differently about how they store, organize, and use data. The traditional concept of data lakes that a data analytics team pulls for you to develop a new product may work for a one-off use case. But if you want the technology to be easily accessible and used, and new innovations to be driven across different parts of the business, then you need to start thinking about what those commonly used data products are that you need to make easily accessible on a self-serve basis to your whole organization.
How do you ensure that they have the right tech skill and abilities, at least to be able to understand what they can do with that data? Because, if that is a skill that we know going forward is going to be an absolute must, basic digital literacy is no longer a nice to have; it’s going to be a requirement of almost every job going forward.
Jacob Johnson: You were implying some limitations here, Seckin, of what this technology can do and what it can’t do. I’ve had clients ask me for gen AI to do things that it should not. And I don’t mean ethically; I mean, it’s just not the right tool. One example was when a client came to me and said, “There are lots of different parts of the organization telling me what to do. I have tasks that are all over the place. How do I prioritize these tasks? Can I ask a large language model to tell me which tasks are important?” And my answer is no, that’s not the right tool for this. It could give you a synthesis of the different types of tasks you’re being asked to do, and it could tell you historically about what types of tasks you have done. But it cannot prioritize them for you and tell you which are the most important things that you should be focused on.
At the end of the day, this technology is just predicting the next token, the next syllable. It’s not thinking through everything and telling me how to run my organization or what the most important thing is. We’ve had other clients who were interested in pulling certain structured data out of semi-structured data—they had a bunch of PDF documents that had information in them, and they wanted to know, “When are we getting this piece of equipment?” But that information wasn’t in a database; it was in a PDF file.
Now you can train through optical character recognition, through natural language processing; you can train models to do these things, but you must be careful when they’re giving you a point-in-time answer. If I ask, “When is this project going to be finished or completed?” it might give me the date it will be finished, or it might give me the date the extension will be finished. I don’t necessarily know unless I investigate that. I think we need to be careful as we’re leveraging these technologies for all the power that they have, as Seckin said, to help educate people in terms of what they’re actually doing, so that we don’t trust them to do things that they’re not really fit for purpose to do.
Debbi Cheong: Seckin, do you have anything to add to that? Are there any other shifts you’re seeing or any potential oversights that could impact gen AI’s role?
Seckin Ungur: Let me give you an example of something that happened recently. We were using gen AI to come up with some futuristic images for client work that we were doing. The client noticed that all of the images were of thin, young women dressed in a particular way. And they weren’t comfortable with that, because it didn’t actually represent the diversity of the population that they were looking to attract. We tried playing with the gen AI tool to generate more diverse images, and it wasn’t very good at doing that.
I think that example showcases one of the limitations of this new technology—because it is new and it’s relying on data that it’s been trained on. And if that data isn’t representative, if it’s got bias within the data, then the results will also have bias. It really highlighted for us the limitations that we need to be aware of and how to address gen AI proactively. Because, otherwise, we do get ourselves into these situations where, to Jacob’s point, perhaps gen AI is not the right tool for what we were trying to do.
Debbi Cheong: Thank you both for such an interesting discussion. Before we close, I have to ask one question. Because we’ve been talking a lot about current obstacles and challenges and also current opportunities, I want to ask you a question on the future. While you can never be too sure about anything, what do you think the evolution of gen AI will look like over the next decade or so?
Seckin Ungur: This is the million-dollar question! What we do know is that the pace of change is very rapid. When we did this automation research back in 2019, we projected out to 2030 and estimated that roughly 40 percent of tasks would be automated by then. Now that figure is 62 percent. For some specific capabilities, like creativity, which one would think would be a uniquely human attribute, we estimated in 2019 that that capability wouldn’t be automated for another 20 to 30 years. Yet, today we know that capability can be automated. And in a couple of years, it’ll be automated to the top quartile of human performance.
The scale of change is so rapid that who knows where it could go in ten years? We know that it will be very different to what we’re experiencing today. But if we get to that point where 79 to 98 percent of tasks are able to be automated, then it’s going to require us to think very differently about what the role for humans is in the future of work. How do we integrate with technology to be able to create better outcomes for citizens by leveraging those aspects of humanity that truly are unique? To me, that comes down to—as we were discussing before—the skill shifts, those social-emotional skills, the ability to use our ethics and judgment to point the technology in the right direction and ensure that it’s being used for good instead of for potentially destructive tasks. I think that’s the future that we need to steer toward.
The technology will continue to evolve, and it will get better, and it will get more accurate. There will be more and more that will be able to be automated. We need to come to terms with the question: what’s the human role in that? How do we continue to skill and upskill to keep up with gen AI and be able to point it in the right direction? We also need to think about what happens in a world where we’ve reached artificial general intelligence. If the machine is able to run AI on itself and do everything that a human could do, what does the future of work look like for humans in that environment?
Jacob Johnson: I think there are two ways to look at this. One is to try to look at how the technology is changing and how far it will get us. We’ve talked about gen AI’s limitations because of what it’s doing. The technology innovation that will take us to artificial general intelligence will be different. But, as has been proven with the difference in our 2019 and 2024 reports, the only thing that’s certain is we’ll get it wrong. We can make our predictions and we can say what we think might happen, but, potentially, what is certain is that the world that my kids will be working in as adults will be quite different to the one that I’m in right now.
I think what we should try to encourage our children to do is dream about what the future could look like and not be bound to a linear path of thinking about where it’s come from and what it might be in the future. I think the great thing about the next generation is they’re not constrained by certain ways of thinking that we were been brought into.
I’ll leave us with one last example, looking back well over 100 years ago, about an innovation that changed things differently from what people initially thought it would. That was when electric motors came in and started to take over from steam engines. The first people who started to implement these electric motors replaced large steam engines in factories with big electric motors because they knew that they would be cheaper, faster, cleaner, and safer. That was the initial thought of how electric motors would transform business, and manufacturing in particular.
It took entrepreneurs like Henry Ford and Ford Motor Company a number of years to figure out that it actually was the simpler and smaller size and scale of the electric motor that could run at different speeds and build a production line. That completely revolutionized the way we thought about manufacturing. The question I will leave us with is this: we all know what the technology can do, and we know we’re replacing certain ways that we do things today, but how is it going to change the fundamental way we think about our society, our lives, and how we operate and work?
Debbi Cheong: Thank you so much, Seckin and Jacob, for this really insightful discussion. To all of our listeners at home or on the road, wherever you are, thank you for joining the Future of Asia Podcast today. If you find this discussion insightful and you want to know more about the topics of gen AI and digital and technology in general, you can visit www.mckinsey.com/FutureOfAsia.
