Change takes a long time
It may be a while before generative AI really makes a difference
Last week was my whirlwind tour from Europe-to-America and back again. I was hoping to have time on flights to write more here, but anytime I was not in front of clients I was a zombie. Apologies especially to paying subscribers. I will pick up the pace in the coming weeks. For everyone else I will do my best to get back on a weekly schedule.
In 2006 I was working for McKinsey. One of the areas I focused on (or more accurately had focused upon me) was lean operations in retail. I would do things like shadow employees in different jobs for extended periods of time and record all the things they did and how long those things took (all with the lens of lean operations in mind). Once all the observations were done we would look at how long things took across the entire company (i.e., what needed to happen to pick up a banana from a supplier, get it to a store, put it on a shelf, have a consumer buy and pay for it, and maybe even return it (clearly bananas were not returned very often, but we tried to be extensive and comprehensive in the analysis)). Every time we did an exercise like this we found millions of dollars in potential savings by applying lean principles.
I found out I was going to be on my first lean project on a Friday. On Monday I would be partnered with a director at the client-company and together we would lead a 12-person cross functional team on the initiative. Prior to that Friday I had never even heard the word “lean”. I asked the McKinsey partner for advice on how to best lead a team like that in a subject area I knew nothing about (I used slightly less eloquent wording. Use your imagination). He suggested I read some books over the weekend, including “The Machine That Changed the World” by James P. Womack. Womack’s book came out in 1990 and became the fundamental text in how to run a company with lean principles. Everything McKinsey was doing was a subset or a derivation of the material in that book. A book that had come out 16 years earlier.
By applying the principles in that book we were able to charge clients millions of dollars to create tens of millions of dollars in savings while improving both customer and employee satisfaction at the same time. They say there are no $20 bills lying on the sidewalk. Even the detractors of American business tend to agree that companies are very good at sucking up profit opportunities, but here was a basic “technology” publicly available to everyone a decade a a half earlier, and it still took the most expensive consulting firm on the planet to help the largest retailers on the planet apply the basic principles that a ignorant 30-year old consultant could figure out over a weekend.
The moral is that it takes a long time for companies to figure out new technologies and ideas. From The Economist (July 16th) ($$):
…consider the tractor. Historians disagree about who invented the humble machine. Some say it was Richard Trevithick, a British engineer, in 1812. Others argue that John Froelich, working in South Dakota in the early 1890s, has a better claim. Still others point out that few people used the word “tractor” until the start of the 20th century. All agree, though, that the tractor took a long time to make a mark. In 1920 just 4% of American farms had one. Even by the 1950s fewer than half had tractors.
The article differentiates between countries and organizations that are innovative, from those that are able to implement innovations. It gives the example of Japan:
Japan is unusually innovative, producing on a per-person basis more patents a year than any country bar South Korea. Japanese researchers can take credit for the invention of the qr code, the lithium-ion battery and 3d printing. But the country does a poor job of spreading new tech across its economy. Tokyo is far more productive than the rest of the country. Cash still dominates. In the late 2010s only 47% of large firms used computers to manage supply chains, compared with 95% in New Zealand.
I have written before about how individuals are exploring new technologies, particularly generative AI, faster than companies. It is easy for an individual to spend an hour or two trying a new product and discovering the potential utility. Anyone who writes mid-tier content for a living now who is not using AI is doing something wrong. And yet I have not yet personally encountered a company that is using generative AI to create content at scale (I understand it is happening, I have just not personally seen it). Last week I was in an SEO-workshop with a company that was proud of their ability to create a new blog post every three days. When I challenged them to create a new post every hour they explained to me why it would be impossible. It is the same everywhere.
GPT-4 launched on March 14th, 2023. As of this writing, the latest capabilities have been available for about four months. In some ways that seems like forever: There are dozens of new companies launching daily that take advantage of the AI’s capabilities. There are a half-dozen companies that are investing hundreds of millions of dollars (sometimes billions) to try and catch up on OpenAI’s head start. AI token length has been extended to 100K or more. Image generation is almost indistinguishable from photos. Speech recognition and automated transcript generation is finally useable. Even text-to-video is some-what possible. But it’s still only been four months.
Even as the technology advances at what seems like an unprecedented rate, we will still have the challenge of implementing that technology. The companies that do so will be at a significant advantage, but the ones that don’t will not die. At least not immediately and maybe not ever. Half of the farms in the 1950s were still not using the tractor. The largest retailers in the world were just starting to figure out lean operations 15 years after it was publicly available. It may be a while before the world is transformed by the technology that has been sitting in front of us since March.
Keep it simple,
Edward