AI comes with a promise of increased productivity. There are wild claims about individuals now being 10 or even 100 times more productive because they are using AI. At an organizational level we seldom see these type of wild claims, and data shows that increase in productivity thanks to AI are much smaller, perhaps even as small as 10 % .
That said, assuming that an individual can be 10 times more productive (without opening the can of worm of defining productivity), who benefits from these new superhuman abilities?
You win
Imagine you can complete your 8 hour workday in under an hour. Done, you can now go home. Focus on other things. Spend time with your kids, play video games, run a startup on the side or do volunteer work. In this scenario you get to extract all the value that AI brings. I am convinced that there are already people quietly applying this to their work life. They are using AI, without telling anyone, to quickly and easily complete all their tasks, still getting paid the same salary.
Your company wins
You have a contract for working 8 hours per day, so the company now expects you to complete what used to be 80 hours of work in a single day. Since they are paying for your expensive tokens, they expect you to use them, and be 10 times more productive. You still get the same salary, but the company is set out to increase their margins by being able to ship much more value. Your company gets to extract all the value that AI brings. You are just an employee with new tools.
Customers win
Since all employees at the company now are 10 times more productive, the same products can be sold for less. Customers get more, for less money. New features are shipped faster, problems are addressed quicker and there is always an AI shoulder to cry on in customer support. To beat competition, companies forward all the value extracted from AI directly to their customers.
AI companies win
In a not too distant future, any company that lives off selling software, or digital services, must be heavy users of AI. It will be the only way to stay competitive. As investors demand to recoup at least some of the billions of dollars they poured into creating frontier models, AI companies will start sending the real bill of AI tokens to their customers. No more subsidized infrastructure or huge investment in training new models without a plan to get a return on those investments. The subprime AI crisis finally happened. But it did not end in default. Instead, no company can live without AI, and are willing to pay almost any cost to keep it. Employees are more productive, companies can ship more software, customers get more, but the cost of building on AI is significant. In the end, the price might be so high that the AI companies themselves claw back a big chunk of the value that AI brings.
Reality - it depends
As always, the truth is somewhere in between the extremes. If you are going to be an active part of an organization, 45 minutes per day is simply not enough. Most companies will have common activities, and you will be asked to do work that cannot easily be automated. There is still great value in the interactions between people, the understanding of data, and decisions on what to do next. Companies cannot ask their employees to work with agents 8 hours per day. The cognitive load is simply too big. The constant task switching, the amount of decisions that need to be made and the checking of the result. You will be running on system 2. Perhaps you can get 2-3 hours of this highly productive work done per day. Then you are exhausted. There is still a human limit to how much value AI can deliver. Customers are unlikely to get the full value of AI, as companies will not give anything away for free. The cost and value of software is going towards zero, but companies must still find viable business model to keep operating. The alternative is that customers simply build their own software. AI companies will most certainly raise their prices. But competition will keep the extremes at bay. Companies like Microsoft and Google still have huge piles of cash to pour into AI development, and they are unlikely to be forced to send the full bill to their customers. The value extracted from using AI will be unevenly distributed. The market will regulate itself. Probably not in the way we expect (it seldom does), but it will keep us away from the extremes.
Consultancy prices
Consultants are the first to be hit with these tradeoffs. If a consultant costs 1000 DKK / hour, and a task that used to take 50 hours now is done in five, the consultancy has a few options. They can bill the customer for five hours plus the cost of tokens, and be forced to seek ten times as many assignments to keep the business going (because consultants still earn the same salary). Or they could keep the fixed price of 50 hours, knowing well that competitors will try to underbid them. This dilemma is already happening in consulting services today, and most consultancies will use AI, without telling their customers.
Dynamics
I believe the way we do work and hold jobs will change fundamentally. It has never been as easy as it is today to start a company and launch a product. I believe the way we think about employment will change, and it will become the norm to hold several jobs. Perhaps you can fulfill your full time commitment to your main employer in less than a full weeks work. The rest of the time you spend on projects that might not be paying off yet. Once they need your full attention, and deliver the promise of an income, you switch your full time focus. But many of these side projects will fail, allowing you to experiment and try new things.
This post is inspired by a great podcast with Steve Yegge and Gergely Orosz. You can read the deepdive here or hear the podcast here