The Steam Engine Trap
We're using AI like 1890s factories used electricity – celebrating efficiency gains while missing the revolution.
When electric turbines first arrived in factories during the 1890s, industrialists made what seemed like an obvious choice: they simply replaced their bulky steam engines with electric motors while keeping everything else exactly the same – the same belt-driven systems snaking overhead, the same rigid rows of machinery, and the same centralized power layouts that had defined manufacturing since the beginning of the industrial revolution in the 1840s.
For nearly two decades, factory owners celebrated 10-15% efficiency gains while completely missing electricity’s revolutionary secret. Then, around 1913 – the same year Henry Ford perfected his assembly line – a new generation of manufacturers finally unlocked the breakthrough: electricity didn’t just power machines differently; it liberated them entirely from the tyranny of physical power transmission.
By the 1920s, visionary industrialists were tearing down their belt systems and scattering individual electric motors throughout redesigned factories, placing workers where logic demanded rather than where steam power dictated, and organizing production flows around efficiency rather than engineering constraints. It fundamentally transformed manufacturing and created a whole category of new, dominant companies.
I can’t help but think that what we are seeing with AI feels very, very similar in so many dimensions: from LLMs’ ability to write code (which truly is getting better week by week) to its usefulness as human augmentation in many creative endeavors, its prowess in content creation (check out our friend Doug Shapiro’s recent keynote on AI’s disruptive potential in media), and the (gradual, and likely sudden) rise of agents/agentic AI…
So much of the current debate is still dominated by the cost-saving/efficiency-enhancing argument – Klarna, Duolingo, Fiverr, and Amazon all had their CEOs come out and publicly declare their respective companies’ focus on AI to cut costs (ultimately in the form of people). I remember being at a meeting of a big consulting firm where the CEO told his assembled leaders about the “20-30% efficiency gains” they expect by using AI.
Don’t get me wrong; I don’t think there is anything wrong with the argument per se. But I think it is short-sighted – and more importantly, it blinds you to thinking about the (real) future. A future shaped by the question “What impossible thing is now possible?” (a question we like to ask as part of our Killer Competitor workshop).
I had a bit of an aha moment when I listened to my old boss Scot Wingo (whom I interviewed recently for our podcast) talk to Roy Rubin – the former co-founder and CEO of Magento, the open-source e-commerce platform. Roy expects the next billion-dollar direct-to-consumer company to be one that doesn’t have a website anymore – but rather one that serves AI shopping agents directly. Not replacing the steam engine with a more efficient turbine, but a complete rethink of what is possible with an abundance of small, cheap, decentralized electric motors…
Now might be an opportune moment to ask yourself what your business, your industry, and maybe the world at large might look like when you allow yourself to fundamentally rethink, well, anything. Not just “how do I make this thing we are doing 10% faster, cheaper, better” – but “what impossible thing can you do now?”
We (likely) might not be quite there yet in terms of capabilities (I, for one, wouldn’t trust an AI agent to autonomously shop for me in most circumstances) – but now is the time to explore, try, build, learn, iterate, and create the mental map and the muscle memory that allows you to pounce when the time is right.
@Pascal