When async is the wrong tool
Async has real limits, and teams that treat it as an ideology spend weeks on things a call would have settled in twenty minutes.
Async-first is a good default and a poor absolute. There are situations where writing is slower, worse and more damaging than talking, and a team that cannot recognise them pays for the principle. For broader context on distributed asynchronous work, GitLab’s all-remote handbook is a useful reference.
The cases that need a conversation
- Disagreement with any emotional content. Written conflict escalates, because tone is inferred and the least generous reading tends to win.
- Genuinely open problems — where the question is not yet clear. Exploration is fast out loud and painfully slow in writing.
- Anything with consequences for a person: performance, pay, restructuring, bad news. Delivering these in writing is a failure of nerve dressed as efficiency.
- A discussion that has already produced more than about three rounds without converging. That is a signal, and the correct response is a call.
- The first contact with someone new, where the purpose is to build enough context that async can work later.
If a written thread has gone three exchanges without resolving, stop. The cost of the call is now lower than the cost of continuing, and the thread is probably about something other than what it appears to be.
Async is slower for divergence and faster for convergence
The useful distinction is between opening a problem and closing one. Generating options, understanding an ambiguous situation, working out what the real question is — these move quickly in conversation.
Evaluating a proposal, reviewing a document, committing to a decision — these are better in writing, because they benefit from consideration and produce a record. For a practical software-side reference on the distinction between productivity and efficiency, this page adds useful detail.
The productive pattern is therefore often a short call to open, writing to work, and a written record to close. Teams that insist on writing throughout spend a fortnight discovering what an hour would have surfaced.
Written conflict has a specific failure mode
Two people disagreeing in a thread each read the other's messages at their least charitable. Each response is calibrated to the previous misreading, and by the fourth message the disagreement is about the thread rather than about the work.
The intervention is mechanical rather than emotional: whoever notices moves it to a call. Not because feelings need managing, but because the medium is actively generating the misunderstanding.
Be honest about which one you are avoiding
Sometimes async is chosen not because it suits the task but because the conversation is uncomfortable. Writing lets you deliver something difficult without watching it land.
That is worth noticing in yourself. Difficult things delivered in writing are received worse, remembered longer and answered less honestly.
Record the outcome afterwards, always
The cost of a call in a distributed team is that people who were not there have no access to it. Any call that produces a decision needs a written record afterwards, in the usual place.
Without that, calls become an invisible layer of decision-making available only to whoever happened to be in the right time zone — which is the specific inequity async was meant to remove.