How *exactly* can AI take your job in the next few years? (Part 2)
Part 2: So...will we actually lose our jobs?
This is part 2 of the full article. To read part 1, click here.
We keep hearing that AI is going to replace a bunch of jobs in the next few years, but no one really knows exactly how that’s going to happen. When will this start, and which exact jobs are in danger? What’s the plan for dealing with this? Right now, it seems like people are vaguely aware of this threat but continuing on with business as normal. If this is really coming in the next 24 months, shouldn’t we be a little more worried, and have a few more answers to these questions?
In the first part, I explored what AI will actually be able to do in the next few years. In this part, I will lay out some actual scenarios that will start to play out in the labor market for white collar workers, and some of the fallout we can expect from these changes.
Part 2: So…will we lose our jobs?
After understanding the capabilities of these remote AI coworkers in part 1, it might seem obvious to assume that many white collar workers will be immediately laid off once this technology is deployed. But there are some factors that make this a less obvious outcome than it might initially seem.
Before we start laying out specific scenarios about what might happen to our jobs, I want to describe some economic ideas that helped make sense of previous times in history when technology transformed economies. These concepts will help us ground our predictions in reality. Despite the AI revolution being different from many of these previous technological revolutions, there are enough similarities to be able to gain insight into what might happen next.
More supply leads to more demand leads to more…work to do
In the early years of car production, without any advanced processes or technology, humans were responsible for assembling each part of the car manually. People who worked in car factories were high-skilled artisans and engineers, as the job required a deep knowledge of each part. Factories could only produce a single car once roughly every 12 hours.
Then, when the assembly line was created, the production of cars started requiring minimal training, and the need for these artisanal workers was no longer needed. You could suddenly start producing many more cars with much fewer workers, and the average time to produce a car went down to about 90 minutes. This productivity was boosted further when automation and advanced technologies were introduced in car factories.
Today, car factories employ more people than ever before. Why is this?
When it became easier to produce cars, it made it much cheaper to make cars, and as a result, many more people started buying cars. The demand for cars increased by orders of magnitude, because it was now so easy to make them! This increased demand meant that employers now needed many more factories to build cars, and started needing many more people to work in those factories, despite their productivity being boosted with the new technology. This massive increase in production also meant that there now needed to be other jobs that supported the production process, like maintenance, quality control, supply chain management, engineering, and much more.
The other thing this boost in car factory productivity did is create entirely new products and services. Since cars were much more popular and owned by a large segment of the population, there now needed to be roads, mechanics, gas stations, auto parts, motels, diners, and so many new things that were impossible to predict before this revolution took place. The world simply changed when cars became a part of daily life, and many of the jobs we have today simply wouldn’t be possible without the technological revolutions that made cars more easily producible.
This pattern has repeated itself endlessly in history. Every time humans introduced a technology that allowed us to produce much more output much more quickly, it led to more of that stuff and also new stuff, which led to…more jobs in ways we could never exactly predict in advance.
Lump of labor fallacy
We have this implicit belief that there exists a fixed amount of work in the economy. In other words, we think that if workers get access to a tool or machine that does half their work, that they’ll only have half as much work to do. In reality, this is not how the economy, or working in general, works. What actually tends to happen when we no longer have to do a portion of work that we usually do, is that other work actually expands to take up the rest of our time.
When ATMs were introduced, they could automatically perform most of the work that human bank tellers did, and many people thought that there would be no more bank tellers in a few years. However, this did not play out as expected. ATMs started to increase the demand that banks faced, because it was now easy to perform most transactions quickly. But since there were still a few things that only humans could do, humans were still needed in banks. So now, more people were visiting banks, and things like complex transactions, identity verification, and financial advice were still needed, which caused human bank tellers to become more valuable! The amount of bank tellers actually increased after ATMs were introduced! Even though a technology could now do most of what they did before, the few things that it could not do, were how they provided value in the new economy, and these things expanded to fill their time.
This isn’t a cherry-picked example to prove my point. Here’s another example.
Bookkeeping and accounting was a slow and manual process for most of history. There were a few bookkeepers that dealt with large amounts of data and it was common to make errors. After tools like computers and spreadsheets were introduced, the job of accountants shifted to become “managers” of these tools, rather than doing the long calculations themselves. They now spent most of their time planning and using technology to perform their jobs. This actually increased the amount of accountants that businesses hired, because it became easier to deal with large volumes of data more quickly, and these professions started providing more value to the economy. It also made the barrier to entry to these fields much lower, and many more people were able to contribute to this field simply by being able to use the tools correctly. Even as accounting tools have become more powerful and more user-friendly, the amount of accountants and financial analysts has grown consistently over the past few decades.
This pattern has also endlessly repeated itself in history. If we gain access to a tool that can perform 50% of the labor we currently do, it’s not true that we will just have nothing to do for the other half of our time. Other work, which we may or may not have already been doing, will expand to fill up the rest of our time.
Back to the question - how will AI impact our jobs?
So, with these ideas in mind, let’s return to the big question: what’s actually going to happen to our jobs once these AI agents are deployed?
I outline three scenarios below that could describe our near future. The goal here isn’t to predict exactly what will happen, but to look at the next few years with more clarity and “higher resolution.” These scenarios are based on my research and reflections over the past several weeks, but of course, I could be wrong.
It’s worth noting that these scenarios aren’t mutually exclusive. One of them might fully play out, or we could see all three happening to some extent at the same time. The latter seems more likely to me.
Scenario 1: More productivity will lead to…more work for us to do
As we have seen, there are many times in history where making workers and organizations more productive actually leads to more workers being needed due to an increase in demand for the products they are creating. As we saw with accountants becoming more in-demand and car factories needing more workers due to their increases in output, AI will have a similar effect on some industries.
Humans with access to remote AI coworkers in their jobs will be able to produce orders of magnitude more output than human workers without this technology. This will also lead to companies starting to increase their output by orders of magnitude.
What would happen if a software game company is suddenly able to create many more games, add more features, and pursue international markets that they weren’t able to before? It would lead to much more demand for their games and their software. And this would require more labor to make those games. Even if humans are only required for a portion of that labor, and AI takes care of the rest, humans will still be needed to fulfill that excess demand.
As we described in the lump of labor fallacy, just because we are not doing the core work ourselves most of the time, it does not mean we will have nothing to do. Rather, we will start providing value through all of the other things we can do that AI won’t be able to do (yet).
This increase in output from companies will lead to more demand for workers in many of these companies.
In this scenario, if you are a worker using these AI agents, the actual job that you are doing will start to drastically change. If you are an engineer, you will no longer be writing most of the code, but rather directing a group of AI agents to produce the correct code. If you are a technical writer or a marketing associate, you will no longer be writing the copy, but reviewing and editing the copy produced by AI agents. We will become “managers” of agents that do the core work for us.
Note
If you’re a bit skeptical about this argument, that’s understandable. It is especially difficult to envision how junior workers or college graduates will be able to provide value in certain cognitive jobs after AI agents can perform their work faster and cheaper than them. This is where Scenario 2 and 3 might be the more accurate ways of looking at the future.
Scenario 2: AI will shift our work to entirely new industries
As we described before, when a new technology enters the economy, it not only increases the productivity of one sector, but also creates entirely new sectors.
The internet revolution created web designers, SEO specialists, social media influencers, and other types of roles that no one could have predicted in advance. The AI revolution will similarly create a plethora of new industries and job types that we will not be able to predict. And even if only some of this work requires humans to perform it, then new jobs will be created.
This is basically the “AI might kill some jobs but it will also create new jobs so we’ll be fine” argument, but laid out in a bit more detail. One of the big questions I always had when I heard this was…well, what exact jobs could it create? Wouldn’t AI be better than humans at doing those new jobs too?
It is impossible to predict specifically what these roles will be, but I’ve explored some possibilities here. I’ve tried to include roles that I think could capture a significant portion of the labor force, not just small roles here and there. It seems likely that humans, not AI, will be needed for these new jobs below for a few years to come.
AI Implementation Specialists for Small & Mid-Sized Businesses
Similar to how web design services exploded after the creation of the internet, it seems likely that as AI becomes a part of our daily lives, most businesses will need help adapting to this new age.
AI Regulation & Compliance Consultants
There are many industries that are incredibly regulated, and will need to navigate the world of AI carefully. Basically every business will need help ensuring that the way they are using AI tools isn’t breaking any laws. Questions like “can I use an agent to handle customer data?” and “can I have an agent reply to my patients for me?” will become common. Since these are new and nuanced situations, humans will be needed to resolve them case-by-case.
AI monitors and trainers
We mentioned earlier that AI will not be able to fully understand human preferences without constant intervention. Small mistakes in their understanding, if not caught early, could lead to disasters, and companies will want to ensure that humans are monitoring the critical parts of their business that are using these AI agents.
As we hinted at the end of Scenario 1, it is inevitable that some workers will no longer be needed to continue doing the work they are currently doing. In this scenario, they will start transitioning to new forms of work. Software engineers will start to become AI monitors, paralegals and junior attorneys might start helping companies transition transition to use AI tools, and each cognitive worker will eventually start using different skillsets to provide value to this new economy. Two of the main ways that humans will provide value in this world is by:
Helping humans transition to start using AI
Helping AI understand what humans actually want
This transition will be messy. It will look different for every single worker. Your particular background, skillset, and preferences will determine how your transition plays out. In some ways, this transition has already begun, however it will really start to build up steam in the next year or two.
These transitions have occurred many times in history before. Agriculture used to employ most of the US population, but now accounts for only 2%. This did not lead to massive unemployment, but a shift to different kinds of work created by the industrial revolution. A similar shift occurred in the digital revolution. It would be unreasonable to expect that at least some of this will not play out when the AI revolution hits.
Scenario 3: Structural unemployment
Despite all this, the fact remains that many jobs that humans do today will simply not be required after remote AI coworkers are introduced. Junior and new-grad white collar workers will be first on the chopping block in certain industries once this technology is deployed.
The lump of labor fallacy is irrelevant if the technology replaces all the things a worker can offer. After elevators became automatic, human elevator operators simply were not required. As these agents become more intelligent and cheap, it will become difficult to justify hiring humans to do much of the technical work we do today.
Some workers will also be fully replaced because some industries simply cannot scale their output to match the increased output that AI provides (i.e. scenario 1 does not apply to them). For example, businesses that are limited by physical constraints cannot just “create more output” if AI makes them more productive. If a paint store can use an AI to maintain its website, it will no longer hire a human for that job. This applies to many other service-based businesses as well.
The demand for some white collar workers will likely start to dwindle within the next 1-2 years if this scenario plays out. In some companies, it has already started to happen. Salesforce has already stopped hiring all software engineers.
There is a real risk that this could lead to structural unemployment in developed economies very soon. If a large proportion of these workers that are laid off cannot find work again, the secondary and tertiary impacts of this could prove destabilizing for our society, as discussed in the final section of this article.
The difference between this technological revolution and previous ones was that previous ones happened over the course of decades and centuries. This gave populations time to adapt and re-skill to the new jobs that are required. This revolution is happening within a matter of a few years. The re-skilling required will be on a timeline we have never encountered before.
The question remains to be seen whether enough of the workers that are affected by these layoffs will be able to find employment due to the new sectors that AI will create (scenario 2) or through their ability to contribute to the increases in output created by AI (scenario 1).
Takeaways
Whatever scenario will play out over the coming years, it’s clear that the nature of the job market is about to change drastically. If you are a white collar worker today, your job is probably going to change in some way within the next 2-5 years. In all 3 scenarios, there is no way most software engineers can go on software engineering like normal until 2030. There is no way a cognitive worker can continue doing what they do normally until 2030. It’s true that some people and businesses will simply be slow to adopt these technologies; this is addressed in the section below. But even with these delays, the odds that your job is not affected by 2030 is slim.
The other major takeaway is that there is a significant chance that Scenario 3 (structural unemployment) plays out for a large segment of the population. That is dangerous news. If this is the case, it will obviously create widespread suffering, but it will be compounded by the political impacts that it will create, described at the end of this article.
Lastly, I want to note that much of this analysis is only focusing on the next 2-5 years. The bigger story is perhaps what AI will be able to do beyond these 5 years. The pool of things that humans can do will shrink more and more, and eventually, there will be no need for human “jobs”, the way we currently understand them. At this point, a new design will be needed to organize society; one we have yet to come up with. These larger concerns about the impacts of super-intelligent AI on society are discussed in this article I wrote in 2023.
I chose only to focus on the near future in this article, because we are almost in this future, and it seems like we still don’t know what it will look like.
Things that will delay AI adoption
Some things will obviously delay the adoption of this technology for many industries. Some people’s jobs will continue as they are today for years into the future without major disruption, simply because of the following factors.
Trust
Many companies and individuals simply don’t trust AI enough yet to hand over important tasks. Whether it’s fear of AI making mistakes, concerns about data security, or just that “gut feeling” that a human should double-check everything, a lack of trust will be a big factor in how quickly we transition to this new age. Industries like healthcare, finance, and legal services tend to be extra cautious, because a single error can have serious consequences. Building that trust often requires reliable performance over time and clear proof that AI can meet (or exceed) human standards of accuracy.
Regulation Is Slow to Adapt
The legal and regulatory landscape around AI is updating very slowly. Some industries, such as healthcare or legal services, may be stuck waiting for specific guidelines or approvals before they can fully adopt AI solutions. Until those regulations are in place, adoption will be stalled.
People Are Slow to Adapt!
Even when the technology is ready to go, people aren’t always quick to change their ways. Companies will need to retrain staff, restructure workflows, and overcome internal resistance to new processes. And on an individual level, many people are simply more comfortable sticking to what they know—especially if they’ve been doing it a certain way for a long time. Change can be intimidating, and a sizable chunk of the workforce will delay embracing AI until they absolutely have to.
Even with all these delays, this transformation will start taking place faster than most people are prepared for. Especially those living in developed countries and working in competitive markets, they will start feeling the effects of this technology very soon.
Political impacts
This is the other big takeaway, but I figured it deserved its own section.
There will come a moment in the next 2-5 years—maybe it happens all at once, or maybe it creeps up on us over a few months—when AI’s impact on jobs will turn into a full-blown political crisis. It might happen when we have the “oh shit” moment when people see how powerful this technology is, or it might hit when the layoffs and jobs transformations actually start to occur. Whenever it happens, it’s going to slam into an unprepared political system.
Governments, by nature, move slowly. We could end up seeing a situation where tens of thousands of jobs start evaporating (or at least radically changing) almost overnight, and lawmakers scramble to come up with quick fixes—maybe new training programs, some kind of universal basic income pilot, or reactive regulations that try to limit AI usage in certain industries. The problem is that right now, almost nobody in office is seriously anticipating that any of this is going to happen in the next year or two. So we will end up with knee-jerk policies, rushed laws, and half-baked ideas getting tossed around in the heat of the crisis.
Beyond policy, there’s also the bigger cultural and social fallout: if people lose their jobs in large numbers and see no immediate path to retraining, you can bet that’ll shape national elections, possibly propelling a new wave of politicians who promise to “protect jobs” by cracking down on AI. We know how this story goes. We have seen it before. Tech-Fuelled Inequality Could Catalyze Populism 2.0
In the 21st century we might witness the creation of a massive new unworking class: people devoid of any economic, political or even artistic value, who contribute nothing to the prosperity, power and glory of society. This “useless class” will not merely be unemployed — it will be unemployable.
What we do in these first few years of AI disruption could shape the political climate for an entire generation.