Investors are questioning AI spending even as executives promise efficiency gains. Amazon CEO Andy Jassy told employees in 2025 that he expected AI to reduce the company’s corporate workforce. For managers, the temptation is to maximize utilization, leaving no idle hands, no empty calendar blocks, no wasted minutes.
But when Microsoft Outlook and Teams went down in November, something unusual happened in a Vancouver office. People actually had time to talk. An engineer-in-training told reporters that as colleagues stepped away from their screens, the office grew talkative. Online, workers joked to the scrambling Microsoft engineers: Take your time. Monday would be fine.
The joke landed because an hour without new demands feels like a rare victory. But it should serve as a warning to anyone managing people.
Here is the paradox of modern management: Squeezing every minute out of a team can make it slower. In a standard queueing approximation, increasing utilization from 80 to 90 percent more than doubles expected waiting time, holding task variability and average duration fixed. When employees are booked to absolute capacity, there is no room to absorb a question, a surprise, or a fix. The cost of a conversation that never happens is invisible, but it is steep. A new employee spends three days trying to solve a problem a senior colleague mastered years ago. Two departments trade defensive emails for a week because neither has ten minutes to pick up the phone.
Companies pay top dollar to acquire expertise, but their obsession with hyper-efficiency can leave employees without time to share it.
Economists Natalia Emanuel, Emma Harrington, and Amanda Pallais recently examined a related trade-off in a study of software engineers at a Fortune 500 company. They found that sitting near teammates increased feedback on code by 18.3 percent and improved overall code quality. Younger and less-tenured employees benefited the most.
But there was a catch: Experienced engineers wrote less code when they sat near teammates.
If a company evaluates that senior engineer only by the code they produce, the time they spend helping looks like a drop in productivity. Under pressure to hit their own optimized KPIs, the experienced worker may put on noise-canceling headphones, leaving the junior worker to figure it out alone.
When everyone looks busy, even a good question feels like an interruption. Teams need spare capacity for unexpected demands and regular opportunities to exchange knowledge. I call those planned exchanges structured serendipity.
In my family science classroom at the University of Hawaiʻi at Mānoa, I train the next generation of your workforce. I see students drowning in this same pressure to be constantly “on.” To counter it, I protect five to eight minutes of a two-and-a-half-hour class for what I call “lounge time.” I project a prompt about the week’s topic on the screen and let them talk. Sometimes I have them rotate partners in short rounds, like speed dating, to bounce project ideas off one another.
It acts as a break, but it is actually a highly effective information exchange. In a follow-up interview, one student recalled feeling stuck on a project until classmates shared ideas and asked questions that helped them find a direction. Because we do it consistently, asking for help becomes an expected part of class.
Employers can easily borrow this format. Take ten minutes from a routine status meeting. Pair people who rarely cross paths and ask each to bring one problem they are still trying to solve. If there is a customer queue to cover, stagger the conversations. Leave some time unassigned for problems that arrive unexpectedly.
Research shows what structured advice exchanges can accomplish. In a field experiment at an inbound sales call center, Jason Sandvik and colleagues gave randomly assigned pairs worksheets to discuss difficult calls and seek advice, plus an opportunity to continue the conversation over lunch. During the four-week intervention, the structured-meeting group’s revenue per call rose about 24 percent relative to comparison groups. Benefits remained evident throughout the 20-week follow-up after the intervention ended.
If you expect experienced employees to help newer colleagues learn the job, you must give them the time to do it, and you must reward that time when you evaluate their output. Adjust their workload to account for that responsibility. Look at both sides of the exchange: Did the newer employee stop making the same expensive mistake? Can they now handle the next case independently?
As companies pursue AI-driven efficiency gains, managers still have to decide how much time to preserve for helping and learning. But leaders must realize that social capital—knowing who to ask, trusting they will help, and having the breathing room to do so—helps sustain long-term output.
Before you fill Monday’s agenda, ask yourself: Who is spending a day on a problem a colleague could help solve in ten minutes?
The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.
