AI made teams faster alone, and quieter together

AI assistants made each person faster on their own. They also removed the everyday friction that used to force teams to talk: being blocked, needing to ask and needing a colleague's review. The result is a quiet return of silos, duplicated work and lost shared knowledge. Coordination used to happen by accident. Now it has to be designed.

The friction that connected teams has gone

Before AI tools, people asked each other questions constantly. Someone was stuck, so they asked the person who knew. A change needed review, so a colleague who understood the system looked at it. That friction was slow, but it spread knowledge and kept work aligned.

AI removed much of it. People now ask the assistant instead of the colleague. Each person and team moves faster in isolation. Practitioners are noticing the effect: one widely discussed thread asked plainly whether teams are back to silo engineering.

The business cost shows up later. Two teams build similar solutions without knowing. Decisions that affect several teams get made by one. When a key person leaves, the knowledge leaves too, because it was never shared.

The problem is not people, so the fix is not motivation

It is tempting to see this as a culture problem and respond with calls to collaborate more. That misreads the cause. People did not become less willing to work together. The work changed so that they no longer have to.

When a side effect disappears, you either accept the loss or replace it on purpose. Asking people to collaborate without changing how work is set up rarely lasts.

Design coordination deliberately

Leaders can rebuild coordination with a few deliberate choices.

Give shared AI tools an owner. When every team sets up its own tools, prompts and agents, costs and inconsistencies multiply. One owner keeps shared tooling coherent and stops duplicated spending.

Replace lost conversations with rituals that carry knowledge. Short internal demos of how teams use AI, review rotations across team boundaries, and a written log of important decisions that both people and AI tools can read. Each is cheap and keeps knowledge moving.

Make cross-team decisions explicit. Agree which kinds of decisions must involve more than one team, such as shared data or customer-facing changes, so they cannot be made quietly by one.

Redraw team boundaries for the work as it is now. If AI has changed what a team can do alone, the old team structure may no longer fit. Adjust it, rather than working around it.

What to do

• Ask each team which AI tools and agents they run, who owns them and what they cost.
• Start one cross-team ritual this month: a short demo or a shared review rotation.
• List the decision types that must involve more than one team, and publish the list.

If faster teams are drifting apart, let's discuss how to reconnect them.