AI Adoption & Business value: a lesson from the Ireland of 1908

Edward Dixon
Nerd For Tech
Published in
4 min readMay 28, 2021


As a founder of Rigr AI, I’m quite literally betting on Artificial Intelligence to feed my kids. So, you might think I’m expecting AI to be everywhere in about, oh, 5 minutes. My actual view is much more nuanced and heavily influenced by the adoption cycle of earlier general-purpose technologies.

To get a sense of just how hard it is to go from “nailed the science bit!” to “serious business value”, let’s take a look at an earlier technological revolution — the electric motor.

A delegation from London experiencing the cutting-edge technology in Ireland c. 1908

Way back in 1908, Ireland wasn’t known for computing or pharmaceuticals, but it was at the cutting edge of ship-building: the Belfast-based Harland and Wolff was the largest, busiest and most advanced shipyard in the world, getting ready to build the Titanic and her sister ships. It was also hosting a group of visitors — some men from Lloyds of London and the editor of the British journal Engineering (founded in 1865, still going today!). The Editor was visiting to see the enormous Arrol Gantry — 840 feet (260 m) long “a machine to build the machine”, which the Yard has installed so they could build the Titanic and the sisters.

However, the Editor was distracted by a much smaller but very pervasive innovation: electric motors were everywhere in the yard — even tiny tools like drills had their own motors!

Even in 1908, electric motors were not new — in 1840s Vermont, Thomas Davenport built one that could drive a small printing press. However, commercial use on a large scale in heavy industry was sufficiently new that it distracted the visitor from the enormous gantry crane he had come to write about. Using his presumably encyclopaedic knowledge of the technology of the day, he thought he could see at once the business value of using electricity to move energy around the huge yard instead of steam, and he was right to think there were savings there:

As the result of electric transmission of power, in preference to steam leads, which were about 1,000 ft long in some cases, the coal bill is half what it formerly totalled.

However, halving the coal bill was not the big win for Harland and Wolff:

But the speeding of machines is a far more significant source of economy. With high-speed tools it is desirable to prescribe a standard speed for the respective metals in all machines and not only does electric driving enable the machinist to exactly regulate the revolutions, but it renders a speedometer check possible. Thus it is easy to get a 10 per cent increase on output, which is of greater value than the reduction on coal. This is a gain not always reckoned.

Why did it take so long for electric motors to become an essential part of a cutting edge engineering business? Infrastructure was part of the problem — Harland and Wolff had to build their own power plant to supply their electricity, and materials were another: they had to employ “one man and a boy” to rewind the coils of failed motors (organic chemistry hadn’t yet produced modern plastic insulators.

The real obstacle to industrial adoption wasn’t materials or infrastructure: it was that the business value (“a 10 per cent increase on output”) was sufficiently non-obvious that it had to be explained to a man who spent his life reporting on the doings of engineers and their businesses.

AI is a new way of building software, with very different strengths and weaknesses. Many of the best tools are open source, and the hardware is available via the cloud — so infrastructure and distribution are very easy by comparison with a motor. However, while a motor is conceptually very simple — give it some power, it spins, more power = faster — this is not true at all for AI. Getting business value from electric motors required having a deep understanding of Harland and Wolff’s business problems and the comparative advantages of steam power versus electric and the ability to get executive buy-in. AI adoption isn’t simple, and the technical bit isn’t necessarily the hardest part.

Getting AI models working has never been easier, but figuring out where to use them is a different story. For that reason, I see huge value for our business in focusing on very well-defined niches. If the electric motor — or hand-written software — are any guide, we won’t run out of niches for a long, long time.



Edward Dixon
Nerd For Tech

#AI guy, Principal/Founder @ Rigr AI, co-author of ‘Demystifying AI for the Enterprise’.