Become More Human, Not Less
Become More Human, Not Less
Why companies will opt for Strong, not Skinny, in the AI choice
In 2018, surrounded by incredible animal motifs in the Michele-era Gucci offices near Firenze, I watched an older Italian man hammer small nails into the underside of a loafer as I walked the factory floor. I’d been living in Italy for the summer, teaching in a program on artisanal manufacturing and automation related to my book which lauds human skills in the era of AI (already obvious in 2017). When I asked the Gucci manager if they’d tried to automate certain processes he scoffed, talking about all of the inefficiencies it added with greater transaction costs in every hand-off. Whereas one human could perform many tasks fluidly, the machine was far more rigid. The gains in efficiency in one specific task were completely offset by its lack of flexibility.
We lose sight of the fact that technology does not substitute for jobs, but for tasks. And every job is comprised of hundreds of extremely diverse tasks. Human-machine substitution is easy click-bait, but activation costs are high, and efficiencies low.
Because of the inefficiencies caused by an attempt at full technology for labor substitution, in most cases companies will opt for what I call “Strong” not “Skinny,” generating far greater output thanks to AI. While much of the media attention is on job dislocations, we ought to focus more on to obviation of tasks, and rote task dislocations that only take away from us what was minimally human to begin with. What we’re left with is a far better world of greater output, and human work better allocated toward tasks requiring greater flexibility of mind and body.
A year ago I wrote a piece where I framed the future of work debate around AI as one in this parlance of, “Strong versus Skinny.” The basic idea is that new tools like AI enable companies to make efficiency trade-offs to do one of two things:
Generate the same output on vastly fewer inputs (Skinny)
Generate much more output on the same inputs (Strong)
The argument I made is that AI will be net positive for output, because many more companies will choose Strong over Skinny. In other words the impact on labor will be acute, while the benefits to output will be broad-based, generating net gain. This isn’t to minimize acute losses in certain rote and routine processes, and certainly some jobs that contain a high proportion of such tasks may fully disappear, but this is to say that on the whole, AI will create more output on the same labor and capital inputs.
As AI predominately optimizes or makes more efficient specific tasks within jobs, it’s difficult to eliminate full jobs if the benefits of technology are felt only in specific workflows. Let’s say that the benefits of AI could fully replace 90% of a given worker’s tasks, the company may still not be able to fully replace a human with a machine. This is because the gap in those 10% of tasks not being done would generate such transaction cost that it would, in addition to the activation costs of switching, create far greater burden than gain. Understanding that AI obviates tasks not jobs, is a vital distinction that that can help us think through the following hypothetical:
If AI can do 90% of what a human can, and if I have 10 FTE, I can fire 9 and keep 1. This is the false belief in full substitution of technology for human because it takes into account the belief that AI doing 9/10 of tasks is the same as it doing 9/10 full human scale jobs, each job with its own subset of 10% complex tasks. This knee-jerk argument for Skinny, or cutting 9/10 of your workforce because of AI, is one that lacks nuance in understanding how AI will actually substitute for tasks, not jobs. This is the click-bait headline. If a company chooses this path they risk spiraling into downward productivity because of high activation costs, the fact that 9/10 tasks done does not equate to substituting 9/10 jobs done, not to mention cultural erosion or morale for the one employee who lost 9 colleagues.
In actuality, AI obviates 90% of the tasks for all 10 people, making all 10 FTE able to dedicate 100% of their time to the 10% of the tasks that are complex, that are non-routine, that require collaboration or creativity. In other words, each employee just became 10x more efficient by able to spend 100% of their time on only 10% of the hardest tasks. So across your 10 FTE you have 100x productivity increase, meaning that you eliminate the competition who don’t follow suit. This is the non-obvious, and far more optimistic headline that everyone keeps their jobs, and the company just generated 100x the output on identical inputs.
AI will merely supplement those human workers in amplifying ways around specific tasks, making them more efficient, and better able to focus on non-routine, complex coordination, collaboration, and creativity. Because the benefits of AI won’t fully eclipse all functions a person can do, companies making choices around machine for human substitution will be faced with inefficiency if they choose this swap. As illustrated in the above example, if they choose to supplement humans with machines they will be amplifying their aggregate employee output by 100x.
I had the chance to speak on this for Huawei at the Drucker Forum in Vienna at the end of 2023. Our comparative advantage is to become more human, not less.