Against the Grain: The (Positive) Impact of AI on Work

Most of the messages, articles, and other pieces we read about AI anticipate a labor catastrophe, with millions of jobs lost to automation. I am afraid I am going to go very much against the grain of the usual message, as I believe that AI will have a net positive impact on work.

Sometimes I get the feeling that I am being a bit cynical, but I can't help thinking that these doomsday messages have no other purpose than to frighten workers, putting them in a position of weakness in any negotiation with companies, so that they settle for keeping their jobs, giving up on aspiring to any kind of improvement, or to a piece of the AI pie. 💼

It is true that if we focus on certain jobs, they may be threatened by automation. These jobs, without getting into specific roles, are usually repetitive tasks with little creativity, limited decision-making, and low responsibility.

Job destruction is also often blamed on productivity improvements: If I can now do the work of 5 people with 3, then I will destroy 2 jobs. However, we need a little perspective to see the impact of AI on society and the economy as a whole. 📈

If we look beyond the impact on a specific role or type of work, and even beyond the behavior of a single company, we can see several reasons why AI will have a positive impact on the job market.

The first is induced demand. It is now much less costly to automate certain tasks. In fact, many tasks that were not previously automatable now are. This means that a massive amount of software will be generated to implement these automations. This software requires responsibility, maintenance, understanding, and evolution, generating new jobs with new skills to sustain this increased demand for software. ⚙️

This is not the first time that tasks have been automated or that induced demand has been generated.

Take, for example, the automation of the textile sector at the end of the 18th century. Back then, the product, handmade by tailors, had a limited demand due to its cost. By automating and lowering the cost, this demand expanded, allowing more people to access the product. Broadly speaking, we can think of AI automation and the products that benefit from it in the same way. 🧵

The second is the increase in productivity derived from the implementation of AI in companies. But let's understand productivity correctly:

Productivity is not the ability to do more things; it is the (economic) value divided by the time it took us to generate it.

Therefore:

Doing more things, if they are useless, does not increase productivity.
Doing the same thing, if we do it faster, does increase productivity.
Doing something more valuable in the same amount of time does increase productivity.

Let us consider the least obvious case (at least according to what is commonly read): doing more valuable things in the same amount of time. This means we will increase the value of the product generated, boosting business margins. The increase in margins is never entirely kept by the company; part of it is passed on to the worker. This is why high value-added companies with larger margins offer better salary conditions to their employees—those who precisely generate that high added value.

The second case, and the one we see most often in the newspapers, is doing the same thing but in less time. In the short term, this case implies a reduction in costs, which is linked to job destruction.

However, this adjustment does not last long, since the cost reduction (and its corresponding increase in business margins) in the medium term will lead to wage improvements (or better conditions, such as a shorter workweek with the same pay) for the remaining workers, or to a reduction in prices, generating deflation, as Jeff Bezos recently commented. Given the systemic anti-deflationary mechanisms (such as central banks or planned obsolescence), I see the first option as much more likely.

There are other factors that will need to be taken into account: the ease of creating products/companies (see "one person companies"), the ability to reach economies of scale much faster, or the lowering of barriers to entry in many markets, for example; but I would like to end the potential positive impacts with the second-round effects. If workers in companies that increase their productivity improve their wage conditions, consumption will increase, generating work and employment associated with that consumption.

This evolution, which I want to believe is positive, is not without risks. The two main ones I see are the decoupling of corporate profits from wages, and the increase in inequality between wages in different companies or sectors (and their qualifications).

That is why we are already hearing some proposals, as OpenAI did recently, such as increasing corporate profit taxes, promoting the adoption of a 4-day workweek, and, in general, policies that guarantee the redistribution and redirection of the wealth generated by AI back to citizens. ⚖️

Being aware of the risks, let us not succumb to fear and go against the grain; let us seize the opportunity that AI offers us to improve working conditions.