In the spring of 1992, Steve Jobs stood in front of a lecture hall of MIT Sloan students and opened his talk not with a product demo, not with the NeXT story, but with someone else’s book. The full recording sits on MIT’s official YouTube channel, and about three and a half minutes in, Jobs starts talking about a man called Paul Strassmann.
“There’s a really interesting book that was written by a guy named Paul Strassmann,” Jobs told the room, “and Paul has one of the more interesting jobs on the planet. He’s the chief information officer – CIO – of a very large organization called the Pentagon.”
Strassmann, who died last year at 96, had one of the great careers in corporate computing. He ran information systems at General Foods and Kraft in the 1960s, then at Xerox through the years when its Palo Alto research centre was inventing the modern office. He became the Pentagon’s first Director of Defense Information, with oversight of the largest IT budget on earth, and was later called back to serve as CIO of NASA. And across all of it, he kept building one dataset: company by company, what firms spent on computers, set against what they earned.
The finding never changed. In 1974 he analysed more than fifty Xerox operating divisions selling identical products against identical competitors and found no correlation between computer budgets and profits. By 1985 he had 84 companies showing the same random scatter. By 1990, 292. By 1997, 539 corporations across the US, Europe and Canada. Profitable firms and struggling firms spent alike, and no measure of profitability changed the picture.
The book Jobs had read was The Business Value of Computers, published in 1990. What struck Jobs was a detail inside the data. Successful and unsuccessful companies both spent about 2% of revenues on IT, but they spent it differently. The strugglers put their money into management productivity: reporting, analysis, tools for the people watching the work. The winners put theirs into operational productivity: tools for the people doing the work.
Then Jobs said something you rarely heard from him. “Now, this was not very pleasant for me to read, because I spent the first 10 years of my life on management productivity, which was PCs. PCs and Macs never attacked operational productivity, they just attacked management productivity.”
The most celebrated product mind of the computing age, conceding in public that a researcher’s scatter diagram had exposed the limits of his own first decade. Worth being precise about the history here: Jobs was running NeXT in 1992, seven years out of Apple and five years from returning, and Strassmann’s influence on him is documented at NeXT rather than at Apple. But the concession is on tape, and it tells you how seriously Jobs took the data.
Thirty-four years later, the industry Strassmann studied has changed beyond recognition and the pattern he found has not. Banks and financial services firms are expected to spend around $53 billion on AI this year, heading past $90 billion by 2028. Meanwhile a 2025 MIT report found that 95% of enterprise generative AI pilots were producing no measurable return. The economist Robert Solow’s old line about the first computing wave – that you could see the computer age everywhere but in the productivity statistics – is ageing uncomfortably well.
Strassmann would not have been surprised, because he spent forty years explaining why this happens. Computers are amplifiers. Identical machines with identical software perform admirably in a well-run company and make things worse in a badly run one. His sharpest line still stings: “The best computer technologies will always add unnecessary costs to a poorly managed firm.” Back in 1997 he was already quoting Gartner research that around 70% of IT projects failed to deliver their expected benefits because the results were never integrated into how people actually work. Swap “IT” for “AI” and you have a decent summary of that MIT report, written 28 years early.
Our own industry has run this experiment recently enough that nobody should need reminding. Habito was the most technology-forward mortgage brokerage of its generation, well funded and well engineered, and by 2022 it was posting losses of over £10 million a year and cutting its broker team in half. What turned it around was not another platform. Ying Tan, a mortgage industry operator through and through, took over in 2023, restructured the business, and within months it posted its first monthly profit. Monzo bought it this April. The technology mattered, but it only started paying when someone redesigned the business around it. Meanwhile the brokerages that quietly compound year after year, the L&Cs and Alexander Halls, have always been built on process discipline and specialist people, with technology in service of both.
Which brings us back to the distinction Jobs found so uncomfortable. Translated into a mortgage business, operational productivity is anything that helps the person writing the case: checking documents before they go to the lender, packaging quality, chasing offers, prompting the protection conversation at the right moment, getting updates to clients before they ring to ask. Management productivity is the summary, the dashboard, the AI-generated report for the people watching the work happen. Both have their place. But forty years of Strassmann’s data and one very recent MIT study point the same way on where the returns live.
So here is a question for anyone in our industry sitting in an AI budget meeting this autumn, and there will be many. When the spending proposal lands on the table, does it attack operational productivity or management productivity? Does it make your best adviser faster, or does it make the watching of your best adviser more comfortable? Strassmann’s scatter diagram kept its shape for forty years because most firms never asked. The spend was the strategy. He died in April 2025, just as the largest information-spending wave in history was gathering. It would be a shame if the best evidence about how this wave plays out was published in 1990, praised by Steve Jobs in 1992, and ignored in 2026.
Research notes
Steve Jobs, MIT Sloan Distinguished Speaker Series, spring 1992. Full video: “Steve Jobs President & CEO, NeXT Computer Corp and Apple. MIT Sloan Distinguished Speaker Series”, MIT Video Productions (MIT’s official YouTube channel), 1hr 12min. The Strassmann passage begins at approximately 3:38. https://www.youtube.com/watch?v=Gk-9Fd2mEnI
Partial transcript of the Strassmann passage, hosted on Strassmann’s own site: “Steve Jobs explains Paul Strassmann”. https://www.strassmann.com/pubs/mit/1992-steve-jobs.html – All Jobs quotes are taken verbatim from this transcript. Note: Jobs was CEO of NeXT in 1992; the documented influence of Strassmann’s work on Jobs relates to NeXT’s strategy, not directly to Apple.
Paul A. Strassmann career and dates: strassmann.com biography (https://www.strassmann.com/bio/) and obituaries. Born 24 January 1929, died 4 April 2025, New Canaan, Connecticut. Xerox 1969-1985, first Director of Defense Information 1991-93, acting CIO of NASA 2002-03.
Strassmann’s datasets and quotes: “Will big spending on computers guarantee profitability?”, Datamation, February 1997 (extract from The Squandered Computer, cached at strassmann.com). Source for the 1974 Xerox analysis, the 84-company (1985), 292-enterprise (1990) and 539-corporation (1997) scatter analyses, the “best computer technologies” quote, and the Gartner 70% integration finding. https://www.strassmann.com/pubs/datamation/datamation0297/index.html
Banking sector AI investment estimated at $53.14bn in 2026, on track to exceed $90bn by 2028: Statista, “Artificial intelligence (AI) investment, impact, and returns in the financial services sector”, 2026. https://www.statista.com/topics/51222/artificial-intelligence-ai-investment-impact-and-return-in-the-financial-services-sector/
MIT (Project NANDA), “The GenAI Divide: State of AI in Business”, 2025 – 95% of enterprise generative AI pilots producing no measurable return. Widely reported August 2025.
Robert Solow, New York Times Book Review, 1987: “You can see the computer age everywhere but in the productivity statistics.”
Habito: losses of £10.7m reported for 2022 with annualised losses reduced to c. £2m in 2023; broker team halved in 2022; Ying Tan became CEO June 2023; first monthly profit September 2023 (Mortgage Finance Gazette, 16 October 2023); acquired by Monzo, April 2026. Sources: Sifted, BusinessCloud, Mortgage Finance Gazette, TechCrunch.
