AI time savings: six hours saved, four spent fixing

Aug 10

Sam Altman says a post-superintelligence world will leave you busier than you expected, still complaining about it, and secretly happy. The productivity math does not support the first half, and nobody asked you about the second.

Speaking on Ti Morse's Relentless podcast, Sam Altman argued that AI will not deliver the long-promised four-hour work week, and his reasoning is psychological rather than technological. People are focused on how they are doing relative to other people, so as AI makes them more productive they get more competitive and choose to work more. Jenni and I spent a good chunk of Episode 69 on it.

My first reaction was that this might be the very definition of the word delusional. Organizations are still arguing with employees about whether five days in the office is reasonable. The idea that those same organizations are about to hand anyone a four-hour week does not survive contact with a single Monday.

The framing is the part worth sitting with, though. Altman puts the reason for long hours on human nature: competitiveness, relative status, wanting to be seen working. That moves responsibility off the people who set headcount, workload and deadlines, and onto the people living with them. It is a convenient story for the CEO of an AI company to tell.

The math does not help him either. SHRM found that AI saves employees around six hours a week, and that roughly four of those go to fixing what the AI produced. Net gain of two hours. That is not a four-hour week. That is a Tuesday afternoon, and only when the cleanup goes well.

Jenni's objection was about the setup rather than the prediction. Being told in advance that the thing you hoped for is not really what you wanted, and that you will secretly enjoy the alternative, closes the loop before anyone gets a chance to complain. If you end up busier, that was your choice. And nobody is expecting a four-hour week anyway. The four-day week trials that worked, worked because organizations were reimagined, not because the same volume of work got squeezed into fewer days.

Here is the part that costs us something. Internal comms is the function writing the "AI gives you time back" messaging right now, in rollout emails and enablement decks and town hall slides. If the lived experience turns out to be four hours of cleanup and a raised target, that gap belongs partly to us. Go back and read what your own AI messaging promised before the rollout, then find out whether anybody has checked if it held.

The most useful advice in the whole conversation came from the AI creator who posts as Cat GPT: own your own experience of the tools. Ignore the marketing, ignore the PR, and work out what AI does for your job rather than what a keynote says it will do.

Also in this episode:

We opened on the US jobs report published the Friday before we recorded: 23,000 jobs lost, revisions taking more than a hundred thousand off previous months, and labor force participation down to 61.4% against something above 63% before the pandemic. The unemployment rate fell at the same time, which sounds like good news until you remember the rate drops when people stop looking. One percentage point of participation is millions of people. Jenni's UK read was more hopeful inside the comms market specifically, via recruiter Andrew Holland of Comms Leaders, though the wider British conversation is about particular pockets: young people who cannot get in at all, and middle-aged women who cannot get back.
Jenni brought Ana Tkalac Verčič's post on the short-form Internal Communication Satisfaction Questionnaire, written with Nina Pološki Vokić, which cuts a 32-item instrument down to 13 questions scored one to seven. I answered her "would this work in the US" question with something that surprised me when I looked it up: a communication satisfaction questionnaire was built at the University of Kansas in 1977, and the US quietly replaced it with eNPS. My reservation is the word satisfaction, which points at neither impact nor difference. An organization can be entirely satisfied with its internal comms and still be failing as a business. Jenni's counter is that the breadth is the point, since a question about how well organized your meetings are reaches into organizational communication rather than channel performance.
bink.'s Culture at Work study surveyed 300 culture, communications, HR and strategy professionals and landed on a say-do gap: culture is shaped by what leaders reward and tolerate rather than by what they say they value. 91% call that gap one of the most damaging forces inside organizations, and 78% say employees learn the real culture by watching what senior leaders tolerate. The number everyone will quote is the 54% who say the people responsible for improving culture cannot change the systems that shape it. The number I kept coming back to is 14%, the share who say senior leaders view culture as a strategic lever tied to performance. We tell each other culture drives performance. Almost nobody believes their own leaders think so. (ICology was one of six distribution partners on the study.)
Business Insider on Ozempic causing awkward work conversations covers something no HR playbook has caught up with. Because GLP-1s read as a medical topic, people feel free to ask, which has produced scrutiny, speculation and what is being called GLP-1 fishing. One man in the piece overhauled his diet and lifestyle, lost around 20 kilos, and was badgered so persistently that he went to HR feeling harassed and disbelieved. In the US it carries a cost argument too, because shared insurance means coverage decisions land on everyone's premiums. My answer to that one is a benefits meeting I sat in 15 years ago, where a young engineer objected to paying the same premium as the COO. Pooled risk is what insurance is. The reason I care about the rest of it is my father.

Watch or listen to the full episode on Frequency, with timestamps, links to every story and the full transcript.