AI can draft the memo, summarize the meeting and offer ten versions of a plan before breakfast. That is useful. It also means we now produce far more text than anyone can read carefully, and much of it says very little. When words get cheap, clarity gets expensive.
You have probably had this experience already. A document arrives that is well formatted, confident and long. You read it to the end and still cannot say what it wants from you. Nothing in it is wrong. It just never commits to anything. Multiply that by every inbox and shared folder, and a large part of the working week goes to decoding text that nobody really meant.
Clarity and simplicity are different things. A clear idea can be complicated. What makes it clear is that you can see its structure: the main claim, the terms, the tradeoff, and what the reader is being asked to do. A clear document can hold uncertainty too, as long as the reader can tell which parts are settled and which are still open.
A good test is whether a reader could repeat your main point to someone else after one reading, in their own words. If they can't, cutting length will not fix it. Plenty of long texts pass this test, and plenty of short ones fail it.
Meetings show the problem well. Two people can leave the same hour with opposite ideas of what was decided, and both will be sincere. The fix is usually one person saying near the end: "Here is what I think we agreed, who does what, and by when." It feels awkward to say. It is also the moment the meeting becomes useful.
Keep the useful friction
The real risk with AI-assisted work is that we hand off the friction that used to show us we were confused. A model can make a vague idea sound finished. It can smooth a contradiction into a tidy paragraph. Without care, we end up managing well-written vagueness.
Writing slowly has always been a way to find out what you think. When a sentence refuses to come, the idea under it usually hasn't formed yet. That resistance tells you to stop and look again. When a tool completes the sentence for you, the resistance disappears, and so does the warning.
That doesn't mean refusing help. It means noticing which kinds of effort carried information. Formatting a table by hand told you nothing. Struggling to state the purpose of a project in one sentence told you a lot. Give the first kind to machines and keep the second kind for yourself.
In practice this can be a one-minute habit. Before opening any tool, write one or two plain sentences: what you need, who it is for, and what would make it good enough. Often this shows that the task is different from what you assumed, or that it doesn't need doing at all. It is easy to skip because it feels like nothing is happening.
When a problem feels tangled, write down what you know and what you are guessing, and what you would do differently if a guess turned out wrong. A model can help gather facts. Sorting knowledge from assumption is the job of whoever has to live with the decision.
Plain language is also a matter of fairness. Organizations often hide behind words like alignment, optimization and transformation, and those words can make responsibility disappear. Clear language puts it back by saying who benefits and who pays.
AI can make this worse or better. It can wrap every hard decision in pleasant corporate phrasing at no cost. It can also translate a policy into plain words, list who is affected, or point out a sentence that hides its subject. Which way it goes depends on what the people using it want.
Plain is different from blunt. You can be clear and kind at once. Kindness here means not making someone guess, and not letting them find the bad news in a footnote.
One habit helps. After writing anything that matters, read it as the person most affected would. A tenant reading a notice from a landlord. A patient reading a letter from a clinic. A colleague reading a reorganization plan that changes their role. Wherever they would stop and frown, rewrite the sentence, whoever or whatever drafted it.
The same applies to personal life. Many people have plenty of information: dashboards, feeds, advice. What they lack is one sentence saying what this period of their life is for. This is a year for getting my health back. This is a year for building the business. This is a year for being present with a small child. One honest sentence like that settles a hundred small decisions. It tells you which invitations to decline and where to put your attention. It also lets some things be mediocre for a while.
AI can help you find that sentence. Ask it to reflect back what you seem to care about, or to point out where your stated priorities and your calendar disagree. The choice itself should stay with you, because a priority someone else picked won't hold on a hard day.
People who are clear will look a bit old-fashioned. They write shorter briefs and ask better first questions. They use tools to see more options, then use their own judgment to pick one. As plausible text multiplies, the people who can remove confusion will be worth more than the people who can produce volume.
None of this takes rare talent. Before you send anything that matters, check what you mean, who needs to understand it, and what they should do next. If you can't answer all of those in a sentence each, the message isn't ready to send.