What happens when you give everyone a tool that makes tasks faster? History says: they don't work less. They just do more tasks.
That's the core argument from economist and Financial Times columnist Tim Harford in a new piece titled "What if AI just makes us work harder?" Rather than treating AI productivity gains as an unqualified win, Harford traces a pattern across decades of digital tools - each one promised to save time, and each one ultimately generated more work.
The PowerPoint Problem
Harford's examples hit close to home for anyone who's worked in an office. PowerPoint meant that "highly paid and skilled professionals started wasting time making their own slides, badly." Before presentation software, you'd sketch a few overhead transparencies or just talk through your points. After it, a one-hour meeting required a 40-slide deck and three rounds of revisions.
Email followed the same arc. Sending a message became essentially free compared to writing and mailing a letter. The result wasn't less communication work - it was "a profusion of low-quality, low-value messages bleeding into the evenings and weekends." Even the humble photocopier gets a mention: it didn't make reading or thinking faster, but it created "a bloated pile of busywork" by making it trivially easy to duplicate documents nobody would ever read.
The economic concept underneath all of this is the Jevons paradox - a 19th-century observation that when coal-burning engines became more efficient, total coal consumption went up, not down. Making a resource cheaper to use increases demand for it. Harford argues AI is the latest instance: making writing, analysis, and communication cheaper per unit means we'll consume vastly more of each.
Greedy Jobs Get Greedier
Harford references Nobel Prize-winning economist Claudia Goldin's concept of "greedy jobs" - roles in corporate law, investment banking, and consulting where disproportionate rewards go to those willing to work the longest hours. He contrasts these with fields like pharmacy, pediatrics, and primary care, where jobs and IT systems have been deliberately designed to let practitioners work reasonable schedules.
The implication for AI is uncomfortable: in greedy-job industries, AI won't reduce hours. It'll raise the baseline output expected per person. If your competitor's associate can now draft twice as many briefs using Claude, your firm matches that or falls behind. The productivity gain goes to the employer and the client, not to the worker's free time.
What This Means for AI Tool Users
Harford's argument doesn't mean AI tools aren't worth using. The productivity gains are real. But expecting them to translate into shorter workdays is probably naive. For freelancers and small business owners, faster output per task often means taking on more clients, not clocking out early. For employees, it means the bar for "a full day's work" quietly rises.
The people most likely to actually reclaim time are those who deliberately set boundaries around output - billing by the project rather than the hour, setting hard limits on deliverables, or working in fields where Goldin's "non-greedy" job design is possible. AI doesn't automatically give you your time back. You have to take it.