AI coordination tools - think automated stand-ups, project summarization, blocker detection, dependency mapping - all operate on the same raw material: written artifacts. A Slack thread, a Notion doc, a pull request description, a comment in a ticket. These systems work by reading the exhaust of how your team operates. They surface what would otherwise require a human to manually stitch together.
Remote-first teams produce that exhaust by default. Decisions get written down because there's no hallway to have them in. Context travels in documents because it can't travel through proximity. The digital trail isn't bureaucracy - it's the substrate that any intelligence layer needs to do anything useful.
Office-first teams produce the opposite: a dense, invisible mesh of verbal agreements, whiteboard sessions that never get photographed, and decisions that live in the memory of whoever was physically present. An AI cannot read a room. It cannot infer what was decided when someone caught the CEO between elevator rides. The information exists - it just exists nowhere a machine can reach it.
This means the gap between AI-augmented and non-augmented organizations will not be evenly distributed. It will roughly track the remote-vs-office divide. Companies that made the cultural investment in async, written-first, artifact-driven work are already sitting on a corpus that AI can act on. Companies that reversed course when the pandemic ended are starting from close to zero - and the habits required to change that are exactly the habits their leadership just chose to abandon.
An RTO mandate isn't just a statement about where people sit. It's a statement about where information lives. And now, that statement means you won't be able to leverage AI as a company intelligence layer.
COVID forced every software company on earth to run remotely for a few years. Most of them did it just fine. Work continued. Products shipped. Revenue held. And then, the moment the pressure lifted, a significant number of them sprinted back to the office and called it culture.
But in some cases it wasn't a culture decision. Leaders who pushed hardest for RTO were often the ones whose working day is most pleasant in an office - walking the floor, being seen, filling calendars with back-to-back in-person meetings. That feels like leadership. It performs as leadership. But it isn't leadership - it's status maintenance dressed up as management philosophy.
There's a second reason RTO may be appealing to people at the top, and it's even less flattering: physical proximity enables the hallway pivot. Someone catches a senior leader between meetings, raises a concern, and a product direction changes by the time lunch is over. No written rationale. No artifact. No trace. The decision either evaporates before it reaches the people executing it or arrives distorted through whoever happened to be in the room. Leaders like to call this "being agile".
But this isn't agility. It's cover. Proximity makes it easy to reverse bad calls quietly, override decisions without documentation, and redirect teams without ever being accountable for the course change that just happened.
The organizational cost of what I call "bulimic agile" is real and underreported. A company that changes strategic direction every few weeks will never build momentum in any direction. It spreads effort across too many simultaneous bets, none of which get enough runway to compound.
The software companies that can benefit from AI-driven coordination are the ones whose work already exists as artifacts. Decisions in documents. Discussions in threads. Plans in repositories. When work is machine-readable by default, an artificial intelligence layer can read the company in real time - what's moving, what's blocked, where the gaps are.
Office-first software companies route their most important information through human proximity. That information is invisible to any system. You cannot automate what you cannot read. AI cannot coordinate around what leaves no trace.
If the strongest forcing function in modern business history - a global pandemic - could not shift your operating model to remote first, you're going to get left behind. You're not facing a remote-work problem. You're facing an AI-readiness problem.
Thoughts? Find me on Bluesky.