Jelena Racković · · 5 min read · Mindset
When AI is mandatory: focus, memory and the cost of doing six things at once
AI is no longer something I choose to open on a slow afternoon. It sits in every tab, every tool and, increasingly, every brief. On a busy day I can have six threads running at once: a draft here, a summary there, a prototype, a comparison, two replies. By the evening I have produced a great deal, and I can barely recall what I actually decided. This note is about that gap, and about how I try to close it.
Six things at once, and none of them remembered
Doing six things at once feels efficient while it is happening. Every thread moves, every window has something new in it, and the sense of progress is very convincing. The trouble starts afterwards, when someone asks why a decision was made and the honest answer is that I am no longer entirely sure.
Speed was never the problem. The problem is that nothing stays. The work happened, but it happened past me rather than through me.
Why nothing stays
Part of it is simply how attention works. The researcher Sophie Leroy calls it attention residue: when you switch tasks, part of your mind stays with the previous one, so you never arrive fully at the next. Multiply that by six, and every task receives only a fraction of the person doing it.
The other part is new. When a tool writes the first draft, I skim it rather than read it, and skimming leaves almost nothing behind. Text that looks finished invites a nod, not a thought. It is remarkably easy to approve something I could not explain five minutes later.
The slow part is the point
This is where an academic habit earns its keep. Reading properly, from start to finish, going back over what does not make sense, checking where a claim comes from and rewriting the conclusion in my own words: none of that is glamorous, and all of it is what turns information into understanding.
I have come to think of revision as the moment the work becomes mine. Until I have read it, questioned it and changed at least one sentence, it is still the tool’s draft, however polished it looks.
The person owns the facts
A tool can produce something plausible in seconds. It cannot be responsible for it. When a number goes into a proposal or a decision goes to a client, my name is on it, not the model’s. That makes me the owner of the information and of its accuracy, whether I checked it or not, and I would much rather have checked it.
That responsibility is also what keeps the work honest. The moment I start treating generated text as a source rather than as a draft, I have quietly handed over the one part of the job that was never meant to be delegated.
How I keep my footing
Most of what helps is unremarkable. I give each thread its own block of time instead of letting them all run at once, and when I leave a thread I write a single line about where I stopped and what comes next. That one line saves me from rebuilding the whole picture the following morning.
I keep a running log of decisions in plain language, so that the reasoning survives even when my memory does not. I close loops before opening new ones whenever I can, because an unfinished thread is a small, constant tax on attention. And every day I read something slowly, without a tool in between, simply to keep that muscle in use.
Following several processes at once is part of my job, and I don’t pretend otherwise. The difference is between following them and being dragged by them. A short note at every handover is usually what separates the two.
Where AI helps, and where it gets in the way
AI is genuinely useful when it gets me to the starting line faster: a first structure, a summary to orient myself, a set of options to argue with, a dull piece of text I would otherwise put off. It is also good at showing me what I have overlooked, provided I treat its answer as a question rather than a verdict.
It gets in the way when it replaces reading instead of preparing for it, when its output looks so finished that I stop thinking, and when it makes starting a new thread so cheap that I open three more instead of finishing one. That last one is the quietest cost of all, because it feels like momentum.
Momentum without burning out
Momentum is not the same as speed. It is the sense that the work is moving and that I am moving with it, aware of where it is going. Six half-read threads can produce plenty of speed and no momentum at all, and a surprising amount of tiredness.
So I try to work in a rhythm rather than in a rush: a few threads at a time, each one properly finished, read and understood, with room left at the end of the day to remember what happened in it. The tools will keep getting faster. Making sure I keep up with my own work is still my job.