A Lack of Imagination

When Excel came along it revolutionised the role of accountants and analysts. It saved them time but it did not reduce their working day. My hours as an analyst peaked a decade later working for Goldman Sachs.

The advent of the internet was supposed to be a windfall for newspapers. Distribution costs would collapse while their quality content allowed margins to expand. Yet it turned out that distribution was the competitive advantage. Margins tumbled as news became a commodity and opinions became commonplace.

Analysts and accountants raved about how Excel would change the world. But if you were a lawyer there was little difference in your role. Newspaper editors waxed lyrical about how the Information Superhighway would alter everything. Industries were disrupted, including their own, but not in the way many imagined.

It turns out that a lack of imagination is the biggest blocker to technological adoption. LLMs trained on the history of human thought will not change that any time soon.

The analyst in the first investor meeting I attended had his model on a paper spreadsheet. He was behind others in adopting Excel, but remained top-ranked in the City because of his experience and connections. He didn’t feel the need to rush any changes.

In a couple of years he was gone. Three years later I became a sell-side analyst. My working hours peaked almost a decade later and only declined once I left Goldman.

Excel removed a major bottleneck in producing standard forecasts. It also greatly increased the amount of number crunching I could do. The accuracy of my forecasts did not improve much despite countless iterations, but this did not matter. Clients paid for my services because of my access to people in the industry.

My hours at Goldman were dictated by peer pressure. It was standard practice to be in well before 7 a.m. It was expected that analysts would be available late into the evening, to attend to corporate financiers whose day often began after 10 a.m. My wife called Goldman the bat cave, because all year round I came and went in the dark.

Fast forward to today and AI accelerates the speed at which analysts can update models. They can stress test a wide variety of scenarios and compare many more companies. But this is behind the scenes work and few clients require analysts to show all their workings. Condensing them into a dynamic target price updating in real-time is also of little use. Active investors value analysts who can reach a conclusion that lasts beyond the end of the week.

AI will face this adoption bottleneck across most industries. While it will speed up tasks, it may not make much impression on the other parts of a workflow. Only when that workflow is reimagined will the full benefits be revealed.

The AI Is Normal Technology newsletter describes knowledge work as a “Decide-Execute-Deliver” sandwich. AI compresses the middle layer, such as writing code, but it does not change the decision process or delivery mechanism at either end of a project. The newsletter argues this is why software engineering will change, but the number of engineers will not be reduced.

So if software development remains safe, where will AI’s impact be felt? That depends on the imagination of entrepreneurs, many of whom are yet to emerge. To illustrate this, let’s go back to the dawn of the internet.

An enlightened analyst might have foreseen that changing distribution would commoditise the news. But where were the analysts predicting how the taxi market would change? Outside of online bookings, how could the digital revolution touch such a physical product?

The answer was in the regulatory monopoly that licensed taxis operated under in many cities. The resulting high margins and resistance to change made the industry ripe for innovation. The internet enabled a car to come to you, rather than requiring you to walk to a taxi rank or wave at a passing cab.

It took Uber to imagine how urban transport could be transformed. It took billions of dollars of venture capital investment to realise that vision. AI is faced with similar constraints and its full impact may be a decade or more away.

The lesson I took from the introduction of Excel was that it is important to adapt. Technology will change the winners and losers in an industry, even if that industry as a whole is little changed.

The lesson I took from Goldman is that there is a political pecking order and analysts were near the bottom. The larger the organisation, the greater the need to navigate the hierarchy if you want to be a changemaker.

The lessons of AI have yet to be learned. If history teaches us anything, it is that predictions of great time savings and incredible efficiency rarely lead to people working less.

Perhaps one day AI will be able to decide and deliver as well as execute. But until people redesign whole industries the way Uber reimagined taxis, the predicted job apocalypse and exponential productivity gains from AI will continue to be tomorrow’s story.