Tone in Digital Communication: The Leadership Skill Most Managers Never Practice
In brief
Why digital messages get read as attitude, and what leaders can actually control about it, now that AI drafts so many of them.
More than once in my career, someone has told me that a colleague had an attitude or was refusing to do the work, and the whole case rested on how a message read. One instance stuck with me. I was the manager, and when I reached out to the person directly, I heard a different story: they had been told to wait — the work was not ready for them — and they had already asked what was needed. Once the work was unblocked, they performed well. The negative impression came from how their messages had been interpreted, and it fell apart as soon as I checked it against what was actually happening.
Misunderstandings like this are more likely when people are rarely in the same room. Remote and hybrid work have been the default for more than six years. Most communication for most teams runs digital-first, and a lot of it digital-only. People who returned to offices still spend much of the day on video calls. Cameras are on less often than they used to be. The decline is understandable, since video calls are fatiguing, but video still communicates things text cannot.
This is where most everyday misreadings start. Written messages carry no tone of voice, so readers supply their own. Whatever attitude a reader hears in a message is something they add themselves, which puts the responsibility for clarity on the person who wrote it. Readers fill any gap in a message with whatever they are already worried about. This matters more the higher a person sits in an organization: a short reply from a senior leader leaves room for the person underneath them to fill the gap with their own concerns about what it meant.
Much of this starts with a message that was never meant to sound curt. It made sense to the writer, who wrote it quickly between two other tasks. It did not read the same way to the person who received it. Some of it starts worse. A participant on a call sometimes unmutes with a keyboard still audible, then asks a question that shows they were not listening. None of this is meant as attitude, and nearly all of it gets read as one anyway.
What a sender can do about this is control the substance of what goes out: how clearly something is said, whether it invites questions, whether it checks that the point landed, and whether it adjusts for the person receiving it. Reading a message aloud before sending it is a simple test. A sentence that sounds harsh spoken aloud will read harsher on the page, where there is no voice to soften it. Putting context at the top of a message, instead of at the end, keeps the reader from guessing at the point. This is the discipline Erica Dhawan describes in Digital Body Language, and it still holds up years into remote work becoming the norm.
The receiver has a job here too. Communication is what the listener does, as Manager Tools puts it, and a listener controls their half of that: a reader who assumes positive intent, and asks a question before deciding a colleague has an attitude, prevents most conflicts before they start. Projection runs both ways. A reader supplies their own state of mind onto an unclear message, and careless writing from the sender makes that more likely. A direct question is more useful than a guess.
The newest factor in this problem is artificial intelligence. Many people paste a rough draft into a language model and send back whatever it produces. They skip the revision and iteration that would make it sound like them, and they rarely give the tool clear instructions about what the message needs to say or who will read it. The result is AI slop: text a machine produced that nobody reviewed before it was sent. The work of making sense of the message shifts from the person who wrote it to the person who has to read it.
The problem compounds when both sides of an exchange do this. One person drafts a message with AI and does not fully know what it says. The other person summarizes the reply with AI and does not fully know what they agreed to. Neither person could explain what happened in the exchange if asked, and both are trusting a tool to handle something that still needs a person's judgment. Used well, the same tools can make a message clearer and more accurate, if the person puts in the time and the effort before it goes out.
Clarity matters just as much when the message is difficult. Professionalism includes pushing back. It includes saying no. Both are professional acts on their own. What decides whether they stay professional is the tone they are delivered in. Brené Brown's phrase from Dare to Lead applies here: clear is kind. A boundary stated plainly is kind, whether or not it is the answer the other person was hoping for.
Think about the last message you sent that got a short or flat reply. Before deciding what it meant, it is worth checking whether the problem was in the writing or in the reading. A short conversation is a low-pressure way to look at that with your own team: https://clients.agile-ideation.com/public/fit-conversation