I wake up, eat my breakfast, sit at my computer. The first thing I see on Slack is a question from my colleague: “Why did you not comment on the document I shared with you last week?”

Why I haven’t read your doc

I feel a bit guilty, for I have not even read it. I value being a reliable person and co-worker, so I open up the document. It is AI-generated.

The reason I haven’t read it is that I get at least 5 of these requests a week. Suddenly, everyone is a prolific writer. Most of the documents are verbose, generic, and clearly unread even by their senders. I do not want to read it.

How we got here

While AI enables more people to share their ideas through text, it has created a new problem: it’s harder than ever to filter what content deserves my own attention, versus what should be handed to my AI assistant.

Sure, we could all have our agents on Auto-mode, generating and reading each other’s documents ad infinitum. But the quality of work — any kind of work — increases dramatically with human effort, and effort is just attention applied over time. Attention requires focus. And focus is the first thing this sheer volume of documents destroys.

We are stuck in a vicious cycle: more people lean on AI to skip the effort, more content lands in everyone else’s inbox for review (or, ironically, their AI-assistant’s review). The less focus any of us has left, the more tempted we are to skip the effort ourselves. And so it goes.

What I do instead

I believe the solution is cooperation. When I need a human to review my thought process, I do not present them with a verbose, long-form document written by AI. That’s for their AI-assistant to read.

Text that starts with human thought and ends with AI polish deserves our attention. Text that starts with an AI prompt and ends with human forwarding does not. From the outside, both might look similar. Only one carries thought.

So here’s my process:

  1. I write the first draft. Even if the language or the sentence structures are not the best. I must put my own thoughts down. I will often catch my own cognitive dissonance, and improve my ideas in the process.
  2. I review it by making sure all of the main ideas are there. I make sure to not drift too far beyond the main point I am trying to make. I make sure it follows a context-first approach. One thing is a must: my review MUST be smaller than the first draft. If I don’t feel like reading the text myself, I cannot count on another human to want to read it.
  3. Finally, I let my AI assistant give me an opinion. Not rewrite it completely. Not add a bunch of paragraphs. Not make it more verbose. If I want AI to add anything I always end my prompt with “be direct and concise”.

If we all did this, my colleagues wouldn’t have to chase me down after breakfast.