Remote Teams Lost the Hallway. AI Can Give Some of It Back.

Picture a new hire on a remote team. They have a question about how a process works. In an office, they'd lean over and ask. Here, they type it into a channel, rewrite it twice so it doesn't sound clueless, and then wait. Three time zones away, the one person who knows is asleep.

Nothing was wrong with the tools. What's missing is the hallway.

What distance does to a team

MIT researcher Thomas Allen spent years studying engineers and found that communication among them dropped as the distance between them grew. Harvard Business Review later described a steep, exponential decline in interaction as the distance between offices increased. His research, summarized in his book Managing the Flow of Technology, covered more than a decade of work on how communication flows through science and engineering organizations. Distance doesn't only change where we talk. It changes how often we do.

A 2021 study of about 61,000 Microsoft employees found something similar at scale. Firm-wide remote work cut the share of collaboration time spent with cross-group connections by about a quarter of the pre-pandemic level, and people added fewer new collaborators. Asynchronous communication like email and IM went up. The work continued. The bridges between groups thinned out.

And there's a second layer. Psychologist Daniel Wegner described transactive memory, a shared awareness within a group of who knows what, and studies show that groups with stronger systems tend to perform better.

Where AI actually helps

This is where I'd point AI on a remote team, because these are the gaps it's good at filling.

• Catching people up. A meeting transcript is a long document nobody will read. Ask an AI for the decisions made, the owners, and what's still open, and the person who missed it gets the useful part in two minutes. The people who attended can skim it for corrections.
• The quick question. Most remote teams have a pile of docs that technically answer common questions and practically never get found. An AI tool pointed at your approved documentation can answer the first-line, “where does this live?” questions at any hour. It also lowers the social cost. Asking a tool carries none of the embarrassment of asking a person, which matters, as I wrote in the pluralistic ignorance piece.
• Time zone handoffs. At the end of your day, have AI draft a handoff from the threads and tickets you touched: what moved, what's blocked, what needs a decision. You edit, you send. The person picking it up starts with context, not archaeology.
• Onboarding without the osmosis. New hires in an office absorb a lot by overhearing. Remote hires don't. AI can answer their small questions, walk them through the team's vocabulary, and let them rehearse a tricky customer call or a hard conversation before the real one.
• Tone in text. Written messages strip out the cues we rely on, so a short reply reads as cold and a direct one reads as sharp. Asking AI “how might this land?” before you hit send catches a lot. It's especially useful on global teams where directness means different things.
• Who knows what. Searching docs, tickets, and threads to find who worked on something similar is a quick way to rebuild that transactive memory. Check what your company allows first, because this one touches people's work history.

What it can't do

Here's where I'd be careful. Every one of these use cases saves time or removes friction. None of them builds trust. The Microsoft finding wasn't that remote teams communicate less. It was that they stopped connecting across groups, and a summary doesn't fix that.

The point

Remote work didn't remove the work. It removed the hallway. AI can rebuild some of what the hallway did, but only if a person decides what to rebuild and why. The tool is the easy part, and the people side was always the harder one.

Happy Learning,

KP