Latest / 80,000 Hours Podcast / How not to lose your job to AI (article by Benjamin Todd)
Transcript
- Hi everyone, this is Benjamin. This article is called “How not to lose your job to AI.” It tries to answer the question of what skills will be most valuable in the future, given everything I’ve read about AI automation. Then it also has some tips on how to start learning those skills. Check out the full article for all the graphs, links, and footnotes. About half of people are worried they'll lose their job to AI, and they're right to be concerned. AI can now complete real-world coding tasks on GitHub, generate photorealistic video, drive a taxi more safely than humans, and do accurate medical diagnosis. And over the next five years it's set to continue to improve rapidly (as I discussed in my last article). Eventually, mass automation and falling wages are a real possibility. But what's less appreciated is that while AI drives down the value of skills it can do, it drives up the value of skills it can't. Wages on average will most likely increase before they fall as automation generates a huge amount of wealth and the remaining tasks become the bottlenecks to further growth. As I'll explain, ATMs actually increased the employment of bank clerks, at least until online banking automated the job a lot more. Your best strategy is to try to learn the skills that AI will make more valuable, trying to ride the wave of automation. So what are those skills? In brief: 1. Learn to apply AI to solve real problems 2. Personal effectiveness, including general productivity, social skills, and learning how to learn 3. Leadership skills like management strategy and entrepreneurship 4. Communications and taste 5. Getting things done in government 6. Complex physical skills I'll explain more about what these are and how to learn them in what’s coming up. These skills will be especially valuable when combined with knowledge of fields needed for AI, including machine learning, but also information security, data centre and power plant construction, and robotics development and maintenance, and to a lesser extent, fields that could expand a lot given economic growth. In contrast, the future for these skills seems a lot more uncertain: Coding, applied math, and applied STEM; Routine white collar skills, such as the recall of established knowledge and routine writing, admin, and translation; Visual creation such as animation; More routine physical skills such as driving. It's hard to say what effect this will have on the job market overall or how quickly it will unfold. If I had to speculate, I'd guess that in white collar jobs — like finance, tech, law, government, healthcare, professional services — entry-level positions, especially those doing more routine work, will struggle in favour of an expanded class of managers overseeing AI agents. Though in the short run even entry-level wages could increase as productivity increases. Small teams and individuals will be able to accomplish far more than ever before. While jobs that require a physical presence — like the police, construction workers, teachers, and surgeons — will be relatively unaffected, maybe their incomes roughly keeping pace with GDP, at least until robotics catches up. If I had to highlight just one piece of practical advice, it would be to learn to deploy AI to solve real problems. You can likely do this in your existing job, but a career development option to especially consider is working at a growing AI application startup. This not only teaches you about AI, but also lets you gain general productivity and leadership skills relatively quickly. In the rest of the article, I'll explain why automation can actually increase wages for the skills that aren't being automated; and use the existing research, economic theory, recent data, and an understanding of how AI works to identify the types of skills most likely to increase in value due to AI. Then I'll use these categories to identify the concrete work skills most likely to increase in value in the future, and explain how to start learning each one. Then finally, I'll give some closing thoughts on how to position yourself given these points, including avoiding long training periods and routine white collar jobs in favour of roles at smaller or growing organisations, side projects, and learning to apply AI to whatever you're doing. And finally, making yourself more resilient by saving more money and investing in your mental health. Section 1: What people misunderstand about automation In the mid-1990s, ATMs started to show up in banks. At the time, people expected that would put many bank clerks out of the job, and indeed the number of clerks per branch dropped from 21 to 13. That, however, made it far cheaper to run a bank branch. So in response, the banks opened far more locations. Total employment of clerks actually increased for two decades, but the clerks now spent their time talking to customers rather than counting money. So while it's commonly assumed that automation decreases wages and employment, this example illustrates two ways that can be wrong. First, while it’s true automation decreases the wages of the skill being automated (in this case counting money), it often increases the value of other skills (in this case talking to customers), because those skills become the new bottleneck. Second, partial automation can often increase employment for people with a certain job title by making them more productive, making employers want to hire more of them. In this case, fewer bank clerks could give better service to the same number of customers. But here's the final twist to the story: today, bank-like employment is in decline. So while partial automation increased employment, the more dramatic automation made possible by online banking did indeed reduce it. And this is also a common pattern. In effect, there are two competing forces. AI tools make human workers more productive, typically increasing their employment. But if AI starts to replace what they can do wholesale, that decreases their employment. When there's a medium amount of automation, it's hard to predict which force will win. But when there's thoroughgoing automation, the second will tend to. In Britain during the Industrial Revolution, textile production was significantly automated. But this made the industry so much more productive that employment in textile manufacturing dramatically increased, only to decline again several generations later. Today, employment of secretaries, administrators, call centre workers, cashiers, telemarketers, special effects artists, and animators is already in sharp decline, with AI maybe helping to continue long-term trends. Data science employment, however, was still up 20% during 2023, despite AI being pretty good at statistical analysis and visualisation. So far, AI has maybe made data scientists more useful rather than replace them, though it remains to be seen how long this will last. Another analysis found that AI has indeed reduced demand for translators. However, translator employment is still up on net. This might be because the decline due to AI wasn't large enough to offset the increase from general economic growth, so far. The third way automation can actually be good for employment is that employment of one job often creates new kinds of jobs and raises wages in aggregate because society becomes wealthier. Historically most people worked in agriculture, but today in rich countries it's only a couple of percent. So we could say that the majority of jobs in the economy have already been automated. However, today incomes are around 100 times higher than they were back then, showing that in aggregate, people moved into much higher paying jobs. In some countries, like South Korea, much of this transition was accomplished in just one generation. Something similar could happen going forward if many remote work jobs are automated. Epoch AI is a research group focused on the interaction of AGI and economics. They estimated that about a third of work tasks could in principle be done remotely, and that if all of those were in fact automated, it would increase economic output between 2 and 10 times. In that scenario, wages for all the non-remote tasks, those that can't be done remotely, would probably increase by about 2 to 10 times as well. In fact, it's even possible that white collar employment would increase, but the roles would entirely focus on the remaining human in the loop and non-remote bottlenecks. None of this is to deny that automation can be very disruptive for the workers in the jobs being automated; it's just to say that it can also sometimes increase their wages, as well as benefit workers in other jobs. All this is also one reason why I prefer to focus on the skills that will increase or decrease in value, rather than particular job titles, since that's much harder to predict. Now, at this point, you might be saying, what about if AI, combined with general purpose robotics, could automate almost every job? Surely wages would fall then. So what would full automation mean for wages? Just as partial automation of bank clerks increased their employment, but more intensive automation decreased it, maybe the same would happen to human workers as a whole. AI combined with robotics has the potential to be unlike any previous technology in that, in principle, it might be able to do almost every economically productive task better than humans. And although many economists dismiss this possibility, the people who are experts in the technology itself believe that it's possible. And if that does happen, many economic models suggest it could drive wages down, perhaps even below subsistence level. Initially, this would be because a rapidly expanding pool of digital workers would massively increase the supply of labour. And eventually it would be because AI and robotics could convert energy and resources into economic output far more efficiently than humans. To be clear, I'm not saying this is what will happen, but it's one possible scenario. Epoch has made a second integrated model of how full automation might unfold across the economy. With their default assumptions, which involves 10% automation being reached next year — of course, that timeline could be longer — wages initially increase about tenfold, only to plunge in the late 2030s as the final human bottlenecks are removed. If instead humans remain necessary for only a small fraction of tasks, say 1%, then the same model shows that wages increase indefinitely, with every human now doing that remaining 1%. That means the difference between 100% and 99% automation is enormous. We don't know which will happen, but I do think full automation and declining wages is a possibility we should take seriously. If that possibility materialises, what should you do? Well, on the way to full automation, there will be partial automation, and for the reasons I've already given, that will increase wages and give you more leverage for a time. So your next steps should be the same either way: learn the skills most likely to increase in value in the immediate future so you can maximise your contribution and wages in the time between now and full automation. There's also an argument for saving more money so that you don't need to depend as much on government redistribution. See more on my Substack: How to personally prepare for AGI. Section 2: Four types of skills most likely to increase in value due to AI The coming years could be very disruptive for many people, and it's likely that wealth gets more concentrated. This article is not about how we should respond as a society, but rather how you can best position yourself as an individual so you can better help society navigate these challenges. In this section I aim to give you the tools you need to think about which of your skills are most likely to increase or decrease in value, given your unique situation and the huge variety of jobs out there. This is also clearly a moving target, but I'll break it down into four key categories of skills that are likely to increase in value. First, the skills that are hard for AI to perform: data poor, messy, long-horizon tasks where a person in the loop is wanted. Secondly, the skills needed for deploying AI. That means things like organising and auditing AI systems, as well as the skills used in complementary industries such as data centre construction. Third, skills that are used to make things people want more of as they get better and cheaper: like improved healthcare, housing, research, luxury goods. And finally, skills that are hard for others to learn: rarer expertise that matches your unique strengths. For the economists in the audience, these are basically low substitution, complementarity, high elasticity of demand for output, and inelastic labour supply. So first, what are the skills AI won't easily be able to perform? The best way to develop your intuitions about this is to try and use cutting-edge AI tools to do real work, making sure not to use the inferior free models. But I would like to provide some more theoretical grounding to what AI will and won't be able to do based on an understanding of how AI is trained. The first kind of task AI is likely to struggle with is tasks that aren't in AI training data and where that data is hard to gather. LLMs are created by training them to predict internet data. This makes them very good at tasks that are based on pattern matching and recall of data on the internet. And that turns out to be a lot. In 2015, Frey and Osborne wrote an influential paper about which skills are likely to be automated. They assumed that social skills would be one of the areas that would resist automation. But today, therapy chatbots are among the most popular AI applications. Many skills that are difficult for humans to learn, including much of therapy, medical diagnosis, and coding, it turns out can be done pretty well by pattern-matching systems. And LLMs can also clearly make some novel generalisations. For instance, you can ask GPT-4, “If the Leaning Tower of Pisa was swapped in location with St Paul's Cathedral and I stood on London's Millennium Bridge looking north, what would I be able to see?” And it can answer that for novel combinations of locations. However, LLMs remain bad at a lot of things, and typically these are tasks missing from their training data. One example is controlling robotics. While the internet contains a huge amount of linguistic text data, there's no equivalent store of data describing physical movement through a 3D space. And the absence of this movement data is also not trivial to fix because it's hard to create realistic virtual environments that could be used to cheaply generate it. That means the only option is to create huge numbers of real robots and have them move around, which is expensive. So today, AI remains much worse at interacting with the physical world. In contrast, data on how to perform many white collar jobs already exists on the internet. And it will be easy to gather even better data because those jobs are mainly carried out on computers which can track each step. Frey and Osborne also predicted that AI would be bad at creative tasks. But today that also seems too simple. As of 2025, LLMs are good at brainstorming, writing in a huge range of styles, including novel combinations of styles, creating rhyming verse and so on. However, they're still not great at novel conceptual insights. This is probably because the first type of creative task is closer to pattern matching within the training data, but the latter requires a greater extrapolation. AI is also likely to struggle with messy, long-horizon skills. The new generation of AI systems, such as o1 and the latest Claude models, use LLMs as a base model but then they're taught to reason and pursue goals using reinforcement learning. This is a bit like learning through trial and error. AI systems try to do a task, then their accuracy is graded, and then they are adjusted in a way likely to increase their accuracy. Over 2024, this new paradigm unleashed dramatic progress in AIs’ ability to do math, coding, and answer known scientific questions. That's because these domains have objective answers that can be immediately verified purely virtually, making them very suitable for reinforcement learning. In contrast, consider a skill like building a company. This involves many judgement calls with no obviously correct answers and where success is determined over years. So it's much harder to get reinforcement learning to work with this type of skill. There are also no massive datasets showing every step an entrepreneur would take to build a company. Some other examples might be things like starting a cultural movement, directing a novel research project, or setting organisational or political strategy. These skills are messy in that they lack clearly defined instructions and measurable outcomes. And they are long horizon in that it takes time to implement and measure success. This is why, in spite of near superhuman abilities at some math and coding problems, AI is still worse than most 7-year-olds at playing Pokemon. It's also still terrible at many tasks that might seem comparatively simple, such as getting a set of shelves installed in an office. That's because these things involve planning, visual interpretation, hiring someone, checking the work is done. The models can effectively execute short, well defined tasks, but they lose coherence and get stuck in loops over longer periods. This also explains why we've seen so little AI automation to date. Even where AI is strongest, software engineering, it can only do roughly one-hour tasks. But most software engineering jobs are made of projects that take at least multiple days, require coordinating with a team, and understanding a huge codebase. Now, it's also true that AI is improving rapidly, even at messy, long-horizon tasks. And if AI progress is rapid enough, and especially if reinforcement learning generalises well enough, it's possible AI surpasses most humans at even these kinds of skills relatively soon. AI might also be able to start making novel intellectual contributions just via brute force generation of ideas. However, messy, long-horizon tasks are our best bet at what AI is most likely to struggle with. And it's possible the ability to do the most messy, long-horizon skills is still decades away. These remarks could be invalidated if a new AI paradigm is created with very different strengths and weaknesses from current AI systems, or if AI progress accelerates. But I think it's the best assessment we can make today. Next, even if AI can technically do a task, it might not be allowed to do so because people want a person in the loop. Here are some categories I've seen suggested by economists where this could be the case. Firstly, legal liability. If there needs to be a person held responsible for a certain type of important decision, like a court lawyer. Secondly, when very high reliability is required, because AI systems hallucinate and make weird mistakes, so people will want human experts to check their answers. You could imagine a human historian checking huge amounts of AI-generated research of archives. Thirdly, when unions and professional interest groups are involved, lobbyists aim to introduce standards and regulations to protect jobs. This could be things like lawyers who control professional certifications for lawyers, so they could try to block AI being used in their industry. When there's a strong preference for human touch, many people will just prefer humans provide certain services, perhaps as a luxury. This could be things like nannies, an artist with a compelling story, religious leaders. In some jobs, physical presence is needed. Even after robots become effective, people could be reluctant to rely on them for applications like police, teachers. Institutional inertia. I expect AI to be adopted faster than previous technology waves. But still many organisations will be slow to apply AI tools, meaning humans stay in important jobs for longer. This might involve, for instance, many government jobs or companies with strong moats that are hard to compete with. Finally, intent alignment. Even very powerful and accurate AI systems will still need to know what humans want them to do. It's possible that more and more roles could involve specifying preferences to AI systems. The factors I've just listed could remain bottlenecks much longer than the first two, because they could still apply even if we have extremely capable AI systems. On the other hand, we don't know how much they'll bottleneck the use of AI. For instance, people often play classical music at weddings, and most people would prefer a human musician. However, these days most people end up using a recording because it's so much cheaper and more convenient. Likewise, even if people prefer human-produced goods and AI products remain inferior in some ways, they might be so much better in others that they become overwhelmingly what people use. It also remains to be seen whether there will be enough demand for these human-in-the-loop jobs relative to the supply of humans able to do them in order to keep wages high. Even if people still want human produced art, not everyone can be an artist. Finally, there's skills where automation is bottlenecked by physical infrastructure. Suppose general purpose robotics started working great tomorrow. How long would it take to automate the manual jobs? Probably a while. Robot production today is in the millions. To build the one billion or so needed to automate all manual jobs would take time, even if it might be faster than many expect. Relatively slow robot production and the lack of data about physical tasks will create a period where the automation of these physical tasks lags behind cognitive ones. In fact, even AIs’ deployment to cognitive tasks will be somewhat bottlenecked by available computing power, especially if the early systems use a lot of test time compute. This will mean that initial AI automation might focus on the highest value tasks, such as in research and development, somewhat delaying automation of lower wage jobs. So the first category of skills were skills that AI is likely to struggle to perform. Second is the skills that are needed for AI deployment. In 2025, having access to cutting-edge AI is already a bit like having 24/7 access to a team of expert advisors and tutors on any topic, unlimited coding capacity for discrete projects, and unlimited remote workers who can do some short admin tasks. These tools are already giving individual workers much more power to make things happen than ever before. And we can already see this in what's happening with the world's most successful startup accelerator, Y Combinator, who say their current batch is 70% focused on AI and growing several times faster than similar startups 10 years ago. This effect is most visible in the virtual and unencumbered world of software startups, but the possibilities are broadening. You don't need to work at a tech startup to use AI to more rapidly learn new skills, get advice, edit your work, create software, and so on. And true virtual workers would dramatically increase this leverage again. This would likely create a period in which the skill of directing these AI workers would become incredibly valuable. These skills could involve things like spotting problems and deciding what to focus on; understanding the pros and cons of the latest models and how to design around their weak spots; writing clear project specifications; understanding what the end users really want (user experience), designing systems of AI workers, including error checking; understanding and coordinating with the people who are involved; and bearing responsibility. In fact, many of these skills are similar to the skills of managing humans, and there is already evidence that competent human managers are better at managing AI teams. These skills are not only the messy long-horizon tasks that AI finds relatively difficult, but they're also complementary to AI. That means that as AI gets better, they become more needed rather than less. These two effects combine to multiply their value. In contrast, being an artisan maker of bespoke Neapolitan suits descended from a long line of tailors is not something that AI will be able to replicate easily, but it's also not complementary to it. That means the market value of this skill likely roughly keeps pace with global incomes rather than outpacing it. Other skills that might be complementary to AI deployment are those involved in the fields needed for an AI scaleup, such as: 1. Expertise in AI hardware. If AI continues to improve, there will be a huge buildout of chips to run and train these systems. 2. AI development. As AI becomes more valuable, the value of making it 1% more effective again increases proportionally, so remaining bottlenecks in AI R&D greatly increase in value. Though bear in mind working on this also can increase the risks from AI, by speeding up its development. 3. The physical tasks necessary for AI deployment, such as construction of data centres and power plants, as well as robotics development and maintenance. Finally, cyber information security. As robotics and AI get more integrated into everything in the economy, the security of these systems becomes vital. After all, no one wants to get kidnapped by their robot butler. The third category of skill is skills that produce something that society could use far more of. What do I mean? I only need to file a tax return once a year. If AI comes along and automates filing a tax return, halving its cost, I will still only file once a year and I'll save that money for something else. In contrast, ride sharing made taxis cheaper and more convenient, and that made people start using them a lot more, in some cases spending more on taxis than they did before. And in fact, the taxi market has grown a lot in the last decade or two. Looking forward, the same could be true for healthcare, nicer housing, better entertainment, luxury goods, personal development, research, and many other things. In contrast, jobs that are needed to satisfy legal requirements, like licencing, and in sectors where demand is mainly set by the government could have more fixed demand. For instance, healthcare salaries in the UK have fallen in real terms in the last decade, despite demand for healthcare generally increasing with GDP. More broadly, you can think about which sectors are likely to grow faster than the rest of the economy in a world of AI automation. For instance, AI automation might create a huge amount of wealth, probably concentrated in the couple of percent who own the most capital. Increased income inequality will spike demand for luxury goods. Something like providing bespoke tea tasting events in San Francisco would be both hard for AI to do and would likely see increasing demand. Finally, there's skills that are difficult for others to learn. Consider a job like being a server at a fancy restaurant. I expect people to eat out more as they get wealthier, and this is a physical, social skills heavy job where people might retain a strong preference for a human touch. So I expect many manual and retail service sector jobs to see increasing employment and for their wages to generally grow in line with the rest of the economy. However, these jobs might not see unusually large increases in wages because people can enter them with relatively less training. If a lot of other people can learn the skill, that limits how much wages for that skill will increase. The skills that will most increase in value are those where it will take a long time for the labour market to respond to increased demand. For example, if you're a construction worker, you could learn a more specialised trade like becoming an electrician, or focusing especially on areas that will see increasing demand, like data centre construction. People with these more specialist skills are more likely to end up as a critical bottleneck during a period of rapid growth. So which specific work skills will most increase in value in the future? Now we've seen which categories of skills are most likely to increase in value. Let's apply those categories to make an overall guess at some concrete, valuable work skills. We want skills that satisfy at least two of the earlier categories and ideally all four. I'm also going to focus on relatively broad transferable skills. Here, especially feel free to skip ahead to the skills you're most interested in, and you might find it especially useful to check out the links to further resources on how to learn these skills in the original article. So firstly, the skills of using AI to solve real problems. These are the skills required for AI deployment that are difficult to automate, such as understanding the strengths and weaknesses of AI systems, designing systems of AIs and interfacing them with the rest of the world, specifying instructions to AI systems, and user experience (UX) for people using the systems. This is valuable because as AI gets more competent, people who can direct these systems become force multipliers. The messy coordination work that AI can't do and the oversight required becomes the bottleneck. Maybe even eventually a lot of the economy could become figuring out what instructions to give AI systems. Anyone can learn this skill by using the latest AI tools to try to achieve real outcomes at work. You can do this in your current job or in side projects. If you want to switch jobs to somewhere that could turbocharge learning this skill, then consider working at an AI application startup or another growing organisation that's trying to use AI to solve a real problem. In these kinds of roles, you'll not only learn this skill, you'll also learn things like entrepreneurship, management, and general productivity. Make sure to use the most cutting-edge models and also think about what might become possible in the next one or two generations. Secondly, personal effectiveness. I break this down into a couple of categories. The first is learning to be a generally productive, proactive person. This means things like setting goals, having a system to keep track of tasks and hit deadlines, learning to motivate yourself and focus, and good professional habits like running meetings and basic emotional management. These skills are useful in any job, so even if there's a lot of automation, they'll probably still be useful. They're also related to agency and the ability to be responsible for things from start to finish, which is a weak spot for AI, and they multiply the value of your other skills. There's many ways to increase your general productivity, which we list in our article in the career guide called “All the evidence-based advice we found on how to be more successful in any job.” Also see the article “How to be more agentic” by Cate Hall. Within personal effectiveness, I also break out social skills, by which I mean building relationships, coordinating well with others, and understanding other people's emotions. Although AI is already often rated as more empathetic than humans, there will still be cases where people want a relationship with a real person, at least as a luxury. Moreover, as more routine work gets automated, it's possible that a greater fraction of what's left could become coordination among teams of humans. For instance, picture three founders managing a large team of AI agents. Those three founders will need to be extremely well synced up and good at communicating. Or imagine a software engineer who now not only has to update his boss about his own work, but also on the output of 10 AIs. Social skills are also an important input into many of the other skills I'm going to list, such as management. This is hard to learn, but try to put yourself in situations where you can practice a tonne, spend time with people who have good social skills, and also see the notes in the article “How to be more successful in any job.” Thirdly, within personal effectiveness, there's learning how to learn. This means the ability to quickly get to grips with new bodies of knowledge and skills. If the world is changing faster and more unpredictably, the ability to quickly retrain into a new skill becomes more valuable. And at the same time, AI means you can get cheap one-on-one tutoring at almost anything, and many people say this is letting them learn far faster than before. Finally, this skill can also help you with all the other skills on the list. How to learn this skill As I've just said, AI has made it much faster to learn many skills because you can now get 24/7 personalised coaching on almost any topic. There are many practical ways to increase your general productivity, which I list in our article on how to be more successful in any job. The third big category is leadership skills, which I'm going to break down into management, entrepreneurship, and strategy. These are all things that seem hard for AI to do and that benefit from the increasing leverage provided by AI. They're also things society could use far more of and are difficult to learn. However, if you can master them, they can be among the most valuable, and I'll give some tips on that too. So by entrepreneurship I mean spotting ideas for new projects, creating a strategy, proactively coordinating people and resources around them, and also being able to handle the risk. A small team of human founders, as we've seen, can already achieve more than before and may soon be able to instantly marshal large teams of AI workers. Anyone can practice entrepreneurial skills by running a side project or a new initiative at work, such as launching a new product, running a conference, running an online store. AI is going to mean those kinds of projects can move a lot faster than before. If you want to focus on having an entrepreneurial career, see our profile on founding organisations. Joining a new and rapidly growing organisation is also a great way to learn these skills. The next cluster is management, including people management, product management, and project management. Some parts of management are long-horizon, messy tasks where people will want a person in the loop to bear responsibility. And I speculate we will see many organisations get more top heavy with a larger number of human managers overseeing either a smaller AI-enhanced team or eventually large teams of AIs. Today employment in management is growing rapidly, though it's also possible that certain middle management jobs get slimmed down by AI tools. People management skills also help you manage AI systems. To learn this skill, read about management best practices. And then start doing management on a small scale, such as managing a contractor or volunteers in a hobby project. See if you can work under someone who is great at management, and then from there try to progress to management positions. Continue to apply best practice and seek mentorship while collecting feedback from the people you manage. Strategy, prioritisation, and decision making is another part of leadership skills. By this I mean setting the vision and mission and metrics of an organisation, identifying the priorities, and making high-stakes decisions. As AI makes it easier to get things done, the key question becomes deciding what to do in the first place. And that is also a messy, long-horizon task that AI will likely lag on. It's true that AI might soon become better than most humans at certain types of forecasting and decision making, but humans will still need to be in the loop for reviewing the decisions. To learn this skill, try to work with someone who has it. Then focus on finding a domain, even if small, where you can practice developing strategy and applying the best practices in doing so. In the article you can find a list of the most common prioritisation frameworks, popular books on strategy, and our article on decision making. I'd also recommend practicing forecasting as a hobby and tracking your results, learn to use AI tools and prediction platforms as decision aids. While writing is getting automated, it's also one of the best thinking aids, so worth learning for that reason. The final cluster within leadership skills I call true expertise. This means having an expert-level understanding of an important field, research taste, maybe the ability to make novel conceptual insights and do really complex problem solving. Experts will be required to provide oversight of AI systems and key decisions and so will be complementary to them. Moreover, having good conceptual insights and research taste will be among the hardest things to automate because they're the ultimate data poor, messy, long-horizon tasks, even though AI might be good at brute force creativity. And finally, these skills are also hard for most people to learn. Expertise will be most valuable in the sectors likely to grow a lot, such as AI deployment, AI development, robotics, computer hardware, cybersecurity, power generation; and in crucial areas of government policy, such as US–China relations, AI regulation, and defence. On the other hand, the bar for true expertise will continually rise over time as AI gets better. You should only pursue this option if you think you can get to the forefront fast enough and stay there. To develop true expertise, find mentorship under a top practitioner, practice intensely, and then pursue whatever training steps are standard to progress in that field. The fourth work skill is communications and taste. By this I mean having good judgement about design, beauty, what people will like; having a personality, story, unique branding, and personal connection to your audience; as well as messaging strategy, and PR and brand strategy. Although a lot of content creation and marketing seems like it's going to be automated, people will still want relationships with real interesting people, and as it becomes easier to create large volumes of content or design, the skill of selecting what's good — taste — becomes more valuable, as do the strategic aspects of what to create in the first place. Now, it's true that being cool is pretty hard to learn, but you can try to develop a deep relationship with a specific audience such as on a YouTube channel. You can practice using AI to help with content creation and tune your taste by seeing what works over time. Focus on more personality-driven content and storytelling, rather than the type of material people can easily get from ChatGPT. Fifth is the skill of getting things done in government. This means knowing who to talk to and how to frame things correctly in order to get new policies passed or implemented. I also mean things like political strategy and government decision making. Even if much routine knowledge work in government gets automated, the government sector will at least likely keep pace with the size of the economy. People will want decision makers to be real people, and this means the nebulous, long-horizon skills of making things happen in government will remain valuable, especially from the perspective of making a difference to society. Indeed, government might even take on increasing importance as more work is automated. It will probably also be slow to adopt, because it doesn't face much market competition. To learn this, work for someone who has this skill, like becoming the staffer to a congressperson. Or consider the other standard entry routes into policy, which we explain in our article on policy skills. This is especially promising if you think you can make it beyond the more entry-level routine analysis positions. The sixth work skill is complex physical skills, by which I mean the ability to do precise physical tasks, especially in unpredictable and/or high-stakes environments with expanding demand, like overseeing surgery, data centre electrician and construction, semiconductor technician. This is because, as we've seen, robotics development is likely to lag, creating major bottlenecks for manual tasks, especially those necessary for AI deployments and that are harder for robots or other people to do. To learn this skill, apprentice in the standard pathway for this field. OK, now what are some skills with a more uncertain future? The following are some skills where there's a stronger case for the value going down. This is very hard to predict, because as we've seen, partial automation often makes demand for a job go up initially only to fall later, and the same could easily be true for these skills. But here are my current thoughts. First is routine knowledge work — like writing, admin, analysis, advice giving: basically all the research on which jobs are most likely to be affected by the current wave of AI agrees that the largest effects will be on white collar jobs around the 70th to 90th percentile of income, which in the US is about $100,000 to $200,000. AI is already pretty helpful for these kinds of tasks because a lot of examples exist in the dataset, and they involve pattern matching or recall of information. Going forward, it'll be easier to collect even more data, and many of the tasks are short and clear enough that reinforcement learning could work on them. More specifically, this could include skills like many cases of writing and copy editing; carrying out straightforward analysis such as a financial analyst, legal clerk, civil servant, or optician might do; recall of established information such as a medical diagnosis; administration; translation. In each organisation, many of these jobs could be replaced by a smaller number of people overseeing a large number of AI agents or AI-assisted humans, making organisations more top heavy. Luke Drago has called this pyramid replacement. That said, as the economy grows, the total number of organisations expands as new niches become profitable. So even if each organisation might need fewer people doing these kinds of tasks, total employment might not fall, at least for a while. These roles could also evolve so that more time is spent on AI gaps, such as talking over AI-generated advice with clients, checking the results of AI-generated outputs, greater investment in training for a smaller but more productive workforce, giving instructions to AI systems. If there's a lot of gaps, employment might not change very much. Not to mention each worker might be having the output of several in the past, which would further increase demand for them in many areas. Many organisations will also be slow to adopt AI tools, also helping these jobs to stick around for longer. All this means it's hard to say exactly how these changes will translate into changes in employment among white collar professions on net. Here are a few speculations about the intermediate outlook for some specific professions. With investment management I expect a continuation of the longer term trend towards greater use of quant systems overseen by a smaller number of often higher paid workers. In strategy consulting, they could be well placed to advise organisations on how to apply AI and have been growing rapidly recently. Increased demand for advice about AI could potentially offset the automation of jobs currently done by junior employees, and they may still be willing to hire junior employees in order to train them for more senior roles. The outlook for professional services seems similar to strategy consulting, but perhaps a bit worse because they're doing less of the novel strategic work that will be harder for AI. For instance, routine accounting will become more and more automated, leaving a maybe smaller number of accountants to focus on more complex cases. Senior lawyers will likely start to use AI to assist them with research, but will need to review key decisions and also discuss them with clients, perhaps with increasing wages for those able to do that. Routine legal work and research however, will be more automated, and insofar as demand is more fixed, the total size of the sector could fall. In government, civil service positions focused on providing research, proofs, and advice, and doing administration might shrink in favour of a maybe larger class of more senior employees and political positions using AI. Finally, in healthcare, workers might spend less time on diagnosis, admin, and monitoring, but more time on physical tasks like administering treatments. The analysis specialties seem most at risk. Medical training as a whole, which is based significantly on recall of information, becomes a bit less valuable relative to nursing, so doctors’ salaries might not grow as much in recent history. So that was routine knowledge work. The second is coding, maths, data science, and applied STEM skills. Ten years ago, at 80,000 Hours, we told people to learn to code and enter data science just before demand exploded. So it worked out pretty well for the people who followed that advice. However, today, looking forward, the prospects for these skills are a lot more uncertain. Coding is what AI is best at now and where it's improving most rapidly. Since programming is virtual and has quick feedback loops, it's relatively amenable to reinforcement learning. Employment for software developers was flat in 2024, after many years of growth. On the other hand, many people have told us that AI tools have also made it far faster to learn to code in the first place, and the scope of what you can do has gone up. Demand for software could also expand as it becomes cheaper to produce, meaning that projects that weren't worth doing before become worth doing. So it's even plausible that the value of spending one or two months learning to code has actually gone up, even if the value of spending years might have gone down. You might also be able to get much more quickly into a place where you understand coding enough to complement your other skills, such as in entrepreneurship or design. So, as of yet, it's not clear that the value of the skill has declined. But we also need to consider what will happen in the next five years. In that time, it seems quite likely that AI starts to clearly surpass humans at many types of coding, even for some longer and more complex projects. But if that happens, software engineers might be able to move into roles that are more about management of AI systems, using their knowledge of coding but combining it with other skills. Others, however, might struggle to make that shift. The situation for data scientists looks similar, though so far, data science employment has continued to grow rapidly. If you're thinking about going into this field now, then focus more on gaining a conceptual understanding of how to do data analysis, not on implementing the basic skills. We could make similar remarks about skills in applied mathematics and applied STEM, especially those that involve applying preexisting knowledge. AI is already beyond PhD level at answering well defined scientific or mathematical questions with known answers. Third is visual creation. AI is already very good at generating imagery and it's about to crack photorealistic video. It still struggles to maintain consistency and follow detailed visual instructions, meaning there's still a major need for human oversight. But this might get fixed in the coming years as agency and multimodality improves. As noted, there were huge layoffs of special effects artists and animators in 2024, while graphic designer employment was flat. On the other hand, some creators will be able to use AI tools to produce dramatically more than they were able to in the past. Fourth is more predictable manual skills. After many years of predictions, self-driving taxis are finally getting deployed for real and are growing extremely fast. It's hard to know how long this will take to roll out across major cities, but it wouldn't be surprising if we saw a mass wave of layoffs among drivers in the next five years. In general, robotics will find it easiest to do tasks in predictable, simpler, lower stakes environments. For instance, they're already doing a lot of warehouse jobs -- though this hasn't yet decreased warehouse worker employment, perhaps because demand for warehouses has increased even faster with online shopping. But the next couple of generations of robotics could reach a tipping point. The final section is some closing thoughts on career strategy. Given all these developments, how should you approach your next couple of career steps? First, look for ways to leapfrog entry-level white collar jobs. As AI increases the value of leadership skills, it's maybe decreasing the value of the entry-level jobs that previously served as the training path into them. So as a college grad entering the job market, who hoped to get one of these training jobs, what should you do? The ideal might be to find a role that lets you learn leadership skills right away. For instance, in any job where you can work with a good mentor. But what happens if you can't find something like this? Well, first you can start to learn AI deployment and personal effectiveness skills in any job, and those are also high on my list of most valuable skills. Secondly, you might be able to find a way to start practicing leadership or communication skills in your existing role, perhaps just on a small scale, such as by managing a contractor or helping to launch a new product. Otherwise, you might be able to start some type of side project or serious hobby, like running a community project, having a blog or having a side business. These let you practice leadership skills and then by using AI tools you can also achieve a lot more than before. In terms of full-time jobs, roles at small but growing organisations seem more attractive because they let you work on these types of skills faster. In contrast, in larger companies there's more specialisation, which means the entry-level roles often involve more routine work. If you have the option, roles at tech startups applying AI to a real problem seem especially attractive — because they let you learn about AI deployment, entrepreneurship, and generally getting shit done all at the same time. You can also see Luke Drago's article “The case for moonshots.” If you're not able to leapfrog the white collar path, then another option is to focus on sectors where performance is driven by complex physical skills, physical presence, and social skills — like a mediator, events organiser, or luxury tourism. Second, be cautious about starting long training periods like PhDs and medicine. AI automation is already happening faster than previous technological waves, could speed up, and has hard-to-predict effects making long training periods less attractive. This isn't to say you shouldn't spend one or two years training, or even that you should never start long training programmes. For example, graduate study could still be worth it due to a combination of the value of true expertise going up, being able to do useful work during your studies if you think AI progress will be slower, or if it just happens to be the best among your other options. But it's worth thinking harder about alternatives. What about finishing college? For most people this is still worth it because it delivers a large boost in employability. However, the case for dropping out seems better than before, especially if your university doesn't even let you use AI tools. Though, I usually caution against dropping out unless you have an offer to do paid work. However, you could try to get into a position where you might get such an offer faster, such as through summer projects, or just try to finish college more quickly. Thirdly, make yourself more resilient to change. One way to deal with a fast, unpredictable change is to learn the personal effectiveness skills that are useful in every job. But you can also think about ways to set up your life to be more flexible and resilient. That could be not overly tying yourself to a single country, living in a large city with many different types of opportunities, saving more money than you would have otherwise, and doing more to invest in your general mental health. Finally, try to ride the wave. The goal isn't to find a single job that will always be resistant to automation, but rather to stay one or two steps ahead of it. This means keeping on top of what AI is capable of, seeking out people to follow who have insights into what's going on, and continually adjusting to where the biggest bottlenecks lie. So how can you put this into action? This week, think about finding a small new way to apply AI in your current or desired job. This month, choose one of the work skills that I've highlighted and think of one or two steps you could take to start learning it faster. In the next quarter, consider whether you should make a larger change to focus more on one of these skills. So thank you for listening. I hope you found it useful. If you have, please forward it to someone else who you think might benefit, and finally subscribe for more articles on what's happening with AI, what it means for your career, and what society needs to do about it.