How AI Changed Our Work Without Asking
The irony of writing an article about how AI has changed the writing process
A few weeks ago, with some coworkers, we had one of those conversations that starts with someone venting and ends with everyone realizing they’ve been experiencing the same thing, but nobody wanted to say it out loud.
Someone mentioned they’d been spending way more time reviewing documents lately because everyone’s now producing more documentation using AI. The discussion turned into how our work has fundamentally changed in the past year and how we are all drowning in each other’s AI-generated documents.
AI tools promised to make us more productive. And in a narrow sense, they delivered. But when it comes to AI-generated documents, they still need to be read, reviewed, and often significantly edited by humans. One of the things I've been doing most lately is trimming down documents that contain unnecessary sentences or even entire sections I haven’t asked for.
When writing required effort, people were selective. You’d work on the analysis that mattered. You’d spend time making it clear and concise because you knew someone would read it. The friction of writing served as a natural filter.
Nowadays, that friction isn’t there. The barrier to creating documentation dropped to nearly zero. And it is super easy to integrate such documentation into the Content Management System you use in your company (I am thinking Notion, Confluence, etc.) and share it with others.
It has happened to me several times that someone spins a proposal about something that you, as a consumer, are obliged to read and comment on, but the proposal will most likely die because it is not backed by a company objective, it is definitely too long for the purpose, and nobody has reviewed it before. It was just a nice-to-have, given little thought about it, and all just because the barrier to writing dropped.
On the other hand, the barrier to consuming documentation stayed exactly the same. I still read at the same speed. I still need the same level of clarity to understand what someone’s trying to communicate. Except now there’s three times as much to read, and most of it is unnecessarily verbose, and sometimes even wrong.
More and more, I have to figure out what the author actually means versus what the AI generated. The productivity gain from AI-assisted writing gets completely wiped out by the productivity loss in AI-inflated reading.
We’ve optimized for speed at the expense of consumption quality. And on top of everything, I’ve realized I’m no longer a writer, I’m an editor. It is a role change nobody announced. And editing is a different skill. It requires you to look at generated text and ask: Is this actually what I mean? Is this the right level of detail? Is this respecting my reader’s time?
I feel this shift in my own writing. I used to write articles from scratch. Now I often start with AI-generated drafts. I spend most of my time removing what the AI added and injecting the specificity that only comes from lived experience. Even this article started as a collaboration with AI.
My complaint is that most people aren’t doing that editing work. Treating AI output as “good enough” and shipping it is unacceptable. Your laziness in reviewing becomes my problem.
This is what the AI that's been helping me write this article calls the “laziness tax.” Every minute I save by not properly editing, I’m taxing someone else’s time when they have to parse through unnecessary words. And the laziness tax doesn’t just cost time. It also costs reputation. When you consistently ship verbose, poorly edited documents, people stop taking your work seriously.
And even if the AI that helped generate this document suggested making it twice as long, I am going to stop it here because I feel I have already made my point.
Don’t stop using AI, but use it responsibly. Review the AI outputs before sharing them!
Enjoyed this post? You might like my book, Data as a Product Driver 🚚.