AI Will Make Us Even Quieter at Work. That’s Not a Good Thing.
The full value of workplace conversations is easy to overlook.
- By
- September 17, 2026
- CBR - Leadership
We don’t talk anymore.
At least, we speak considerably less than we used to. A study by University of Missouri–Kansas City’s Valeria A. Pfeifer and University of Arizona’s Matthias R. Mehl analyzed ambient audio recordings from the daily lives of more than 2,000 people from 2005 until 2019. Over that period, the average number of words spoken every day decreased by about 28 percent. “This loss of words reflects real spoken conversations, big ones and small ones, that we stopped having with others,” Pfeifer and Mehl write.
Our workplaces are quieter too. The widespread diffusion of digital communication tools means that many do their jobs in relative silence. One in ten employed Americans told Ipsos in 2019 that they “never interact with people through in-person conversations or meetings at work.” Six in ten indicated they do so less than two hours a day. And half of office workers surveyed by KPMG in 2025 agreed that technology has “created a false sense of connection among themselves and their colleagues and has replaced deep conversations with superficial interactions.”
Early evidence suggests that artificial intelligence may quiet work further. A 2026 study from Workday finds that 76 percent of employees have turned to AI instead of a human for advice. It also found that 44 percent chose an AI chat over a conversation with a colleague “because it completes tasks faster without challenging their ideas,” while 37 percent “prefer brainstorming with AI rather than a human colleague for fear of judgment.”
Our quieter workplaces represent a real threat to businesses. Most of these forgone conversations are exactly what they appear to be: a status update, a routine handoff, a request to reformat a document. Automating those is a productivity gain every organization is right to pursue. But a minority of interactions are quietly doing more, building a social framework whose operational importance is easy to overlook. When we automate conversations in the name of expediency, we risk losing more than we may recognize.
The second job of a conversation
Consider an ordinary event in working life: someone asking a colleague a question. Let’s imagine a relatively junior member of a workplace approaches a seasoned coworker for advice on a task the junior person has never encountered. The senior colleague explains how they’d approach it, their young teammate thanks them, and the exchange is done.
The visible job is plain. The person needs an answer, and they get one. If that were all the exchange accomplished, automating it with AI would be pure gain—and often it is.
AI makes it possible to keep the answer while dispensing with the exchange once needed to produce it.
But think of what else happens during that exchange. A junior employee learns how an experienced colleague frames the problem, judgment that lives in no manual. The senior employee learns that something is happening in that corner of the business and makes a note to keep an eye on it. A working relationship is reinforced that will matter during some future crunch. None of that was the point of the conversation. All of it happened anyway.
That conversations do this unspoken work is not a new observation; scholars have understood for decades that the real workings of a company differ from its formal processes. What is new is the power of AI technology. Earlier tools that reshaped how we communicate—the memo, the telephone, email—thinned these exchanges but at least kept a person on the other end. AI makes it possible to keep the answer while dispensing with the exchange once needed to produce it.
Ask a machine, and the answer still arrives. The lesson, the early warning, and the reinforced relationship may not. The visible job gets done; the quiet ancillary benefits simply disappear.
A cautionary tale
One of us (Seth) learned firsthand how costly this kind of lost communication can be while consulting for a large oil company. The company ran its tanker fleet from a room with a large map on the wall. Ship captains phoned in their positions by satellite, and employees moved magnets across the map. Modernizing the arrangement seemed the perfect project for a confident young consultant: A new system could use GPS transponders to report vessel locations automatically; a live dashboard could replace the magnets and the people who moved them. The project was delivered on schedule, and it worked as designed.
Within weeks, ships began arriving later than expected in ports. Docks sat idle. Costs climbed.
The data collection wasn’t the problem; the new system captured location perfectly. The problem was that the calls had been conveying information beyond coordinates. When a captain called in their position, they rarely stopped there. They mentioned weather closing in, an engine running rough, a delay that concerned them. The people in the map room heard the hesitation in the captain’s voice, asked questions, and noticed patterns. They were not simply tracking ships. They were sensing trouble early enough to do something about it.
The phone calls were reinstated. The young consultant was shown the door.
A preview from the pandemic
The abrupt shift to remote work during the pandemic provided a natural experiment that enabled researchers to replicate at scale what the consultant learned in the map room.
Studying a large customer-service operation, Natalia Emanuel of the New York Federal Reserve and University of Virginia’s Emma Harrington find that employees who were working remotely before the pandemic answered about 12 percent fewer calls per hour than their on-site colleagues—and that when the office closed and formerly on-site staff went home, call quality and promotion rates declined across the board. Workers blamed their inability to “quickly consult with coworkers” and having to “wait longer for advice from more experienced colleagues.” They suffered when the conversation stopped.
A separate study addresses why. Microsoft’s Longqi Yang and other researchers from Microsoft and the University of California at Berkeley examined the collaboration networks of Microsoft’s more than 60,000 employees before and after its offices closed. The networks became more static and siloed when work went remote. The ties that bridged groups, that carried new information, thinned. People interacted principally with close colleagues who told them what they already knew. Email traffic increased. Synchronous conversations declined.
Neither study measured morale. Both measured the machinery through which organizations coordinate, learn, and make critical business judgments. As in the map room, that infrastructure degraded when the conversations supporting it thinned. Remote work inhibited those interactions. AI risks replacing them entirely.
Automation’s gains arrive immediately and visibly. The losses they produce come later with seemingly indeterminate cause: junior employees who never learned what their predecessors knew; teams with little in common but their metaphorical jerseys.
What conversation is really for
The social benefits we’re at risk of losing from AI aren’t just produced by the kinds of straightforward interactions we portrayed earlier between the junior and senior colleagues. Sometimes workplace conversations that are frowned upon have unintended positive consequences too.
Consider eavesdropping. Communication scholars Leila Bighash of the University of Arizona and Andrea B. Hollingshead of the University of Southern California explain that eavesdropping provides environmental knowledge that is “not readily available through conventional information-seeking strategies.” University of California at Santa Barbara’s Paul Leonardi has found that through eavesdropping workers “were slowly becoming aware of what and whom their coworkers knew and filing that metaknowledge away for later use.” It provides contextual awareness, helping employees understand relationships, influence, and organizational dynamics otherwise hidden from formal communications.
And then there’s gossip. Not the pernicious kind that can be hurtful and even toxic. But rather, the mundane ways we share information in conversation likely not captured through formal channels. Gossip creates shared understandings and social bonds. Chicago Booth’s Ronald S. Burt describes it as a “prism,” reflecting the values and norms of the organization, and as an “echo” reinforcing those norms through repeated conversations. Gossip also builds reputations. And reputation is a critical predicate for the trust that enables teams to thrive.
Why the danger is easy to miss
If the risk of lost conversations were obvious, careful managers would already be guarding against them. The reason they don’t is structural. The functions that vanish when an interaction is eliminated leave no record and appear on no dashboard, while efficiency is measured constantly and reported eagerly. Automation’s gains arrive immediately and visibly. The losses they produce come later with seemingly indeterminate cause: junior employees who never learned what their predecessors knew; teams with little in common but their metaphorical jerseys.
No one decides to dismantle the social infrastructure a company relies on to sense problems and develop its people. It happens one seemingly sensible choice at a time. Why interrupt a busy colleague when the answer is already in the system? Why ask a senior person to review a draft when a tool can polish it instantly? Each choice is reasonable on its own; together they can hollow out capabilities the organization did not know it had until they are gone.
Start with a question
None of this is an argument against AI. The gains it promises are real. Organizations that hesitate unduly will lose out. The task is to automate with open eyes, to streamline the conversations that are purely task based while encouraging the ones quietly doing more. Take a short pause before replacing a conversation with a machine. Ask a single question: What else may be happening here?
Actively seek out the conversations that carry something—mentoring, an early read on a problem, a relationship that holds two teams together. The point is not to preserve every meeting and hallway exchange, but to recognize that the manifest output of an interaction does not always represent its full value.
It is a lesson those in the oil company’s map room learned only when their ships were already missing their ports.
John Burrows is a senior lecturer at the University of Chicago Harris School of Public Policy and an associate fellow at the University of Oxford Saïd Business School. Seth Rachlin, a sociologist and former Capgemini executive vice president, teaches at Arizona State University. They are the authors of the forthcoming Social Capital at Work: Building the Hidden Asset That Drives Trust, Engagement, and Performance (Simon & Schuster) from which this essay is adapted.
- Leila Bighash, Kristen S. Alexander, Christina S. Hagen, and Andrea B. Hollingshead, “A Model of Social Eavesdropping in Communication Networks,” International Journal of Communication, June 2020.
- Ronald S. Burt, “Bandwidth and Echo: Trust, Information, and Gossip in Social Networks,” in Networks and Markets, eds. James E. Rauch and Alessandra Casella, Russell Sage Foundation, 2001.
- Natalia Emanuel and Emma Harrington, “Working Remotely? Selection, Treatment, and the Market for Remote Work,” American Economic Journal: Applied Economics, October 2024.
- The Human Connection Workplace Index, Workday Foundation report, 2026.
- KPMG Friends at Work Survey, KPMG report, July 2025.
- Loneliness and the Workplace: 2020 U.S. Report, Cigna report, January 2020.
- Valeria A. Pfeifer and Matthias R. Mehl, “Sliding into Silence? We Are Speaking 300 Daily Words Fewer Every Year,” Perspectives on Psychological Science, March 2026.
- Longqi Yang, David Holtz, Sonia Jaffe, Siddharth Suri, Shilpi Sinha, Jeffrey Weston, Connor Joyce, Neha Shah, Kevin Sherman, Brent Hecht, and Jaime Teevan, “The Effects of Remote Work on Collaboration Among Information Workers,” Nature Human Behaviour, January 2022.
More from Chicago Booth Review
More from Chicago Booth
Your Privacy
We want to demonstrate our commitment to your privacy. Please review Chicago Booth's privacy notice, which provides information explaining how and why we collect particular information when you visit our website.