Jeff Reynar, co-founder of this+that, argues that workplace technology has spent four decades promising freedom while often delivering the opposite. In a recent article, he says the efficiency gains created by personal computers, email, and smartphones were largely absorbed by companies through higher output demands and broader expectations that employees stay available.
His argument does not end there. Reynar says artificial intelligence could become the first real exception to that pattern, but only if businesses choose to use time savings for better thinking, stronger collaboration, and higher-quality work rather than stuffing in more of the same tasks.
The autonomy paradox
Reynar writes that before PCs, email, and smartphones became standard, a worker in a given role might have been expected to produce 10 “units” a week. By 2015, that same role was expected to produce 20.
The tools did not reduce the load. They raised the threshold.
He ties that shift to work by Massachusetts Institute of Technology researchers Melissa Mazmanian, Wanda Orlikowski, and JoAnne Yates, who studied how knowledge workers changed their behavior after adopting mobile email. They described the result as the “autonomy paradox”: workers adopted mobile devices to gain flexibility and the freedom to work from anywhere, only to find themselves working everywhere, all the time.
In Reynar’s telling, no single decision maker had to declare that outcome. The gains were quietly absorbed by the organization.
He also makes clear that he is not arguing technology never removes jobs. Dictation-related roles disappeared, and he does not present that as something people genuinely miss. His point is narrower and more persistent: time saved by new tools is usually redirected into higher expectations instead of being returned to people who were already overloaded.
AI’s early phase looks tense, not lighter
Based on what he sees now, Reynar does not think the current AI cycle looks especially encouraging.
He says AI is making work tighter. That is not limited to early adopters. As role boundaries blur, workers are increasingly expected to absorb tasks that once would have been handed to someone else, which can raise intensity instead of reducing it.
He places this phase within what economists call the productivity J-curve. Major technologies often create disruption and drag on output before the benefits are fully realized later. For Reynar, the real question comes after AI finally starts saving meaningful time: will companies return that time to workers for thinking and collaboration, or use it by reflex to push more of the same work through the system?
A precedent from desktop publishing
Reynar points to desktop publishing software in the 1990s as a case that followed a different path.
At the time, many feared that print shops would close and typesetting work would disappear because anyone could make newsletters or wedding invitations on a computer. But desktop publishing did not simply accelerate the pace of work the way earlier office tools did. It raised the quality bar. Amateur outputs filled with flashy fonts and decorative graphics often looked poor, and people quickly learned that strong work still depended on judgment and skill, not the tool alone.
Some jobs did disappear, and typesetters became far fewer. Yet design work expanded. Citing U.S. Census Bureau data, Reynar notes that the number of designers in the United States rose from 393,000 in 1983 to 788,000 in 2001, while total employment over the same period grew 34%.
That example underpins his broader case that efficiency does not always have to translate into speed. It can also translate into quality.
McKinsey’s 2.5-hour estimate
McKinsey estimates in its “Superagency” report that by 2030, AI could save knowledge workers about 30% of their workday, or roughly 2.5 hours a day.
Reynar cautions against turning that figure into a simple jobs calculation. Automating 30% of work hours does not mean 30% of jobs will disappear, he says, because people are not interchangeable parts. He also says he does not expect healthy companies to cut staff at that ratio.
What he sees as the bigger risk is that some companies will require workers to use those saved hours to produce more of the same output. In markets where demand is already saturated, he argues, that approach would only push prices lower and leave everyone running in place.
The alternative is to trade time for better quality rather than more quantity.
Walking, proximity, and face-to-face contact
Reynar says that when he gets stuck, he walks his dog, runs, or bikes, and often returns with progress. He cites a Stanford University study showing that walking can boost creativity, including walking on a treadmill.
He also refers to Bell Labs’ intentionally designed long corridors and to MIT scholar Thomas Allen’s “50-meter rule,” which found that communication frequency drops sharply when coworkers are more than about 50 meters apart. For Reynar, those examples show that face-to-face interaction is difficult to replace with digital tools.
Open-plan offices were meant to capture that kind of proximity, but he notes that the results often disappointed. Research by Ethan Bernstein and Stephen Turban on two Fortune 500 companies after open-office redesigns found that face-to-face interaction fell by about 70%, with employees shifting back to email.
That, in his view, brings the issue back to a simple question. The problem is not efficiency tools by themselves. The problem is who ends up claiming the time they save.
Whether AI breaks the pattern
Desktop publishing, Reynar argues, showed at least once that a technology upgrade could leave some of the gains with workers by pushing value toward better work rather than more work.
He says it is too early to decide whether AI can repeat that outcome. But one point is already clear in his analysis: if companies take the full 2.5 hours and use it only to make employees do more of the same tasks, then this AI wave will not look meaningfully different from the last 40 years of workplace technology.