Picture a waiter at the peak of a Saturday evening service in a busy Indian restaurant. Fifteen tables, the kitchen already stretched, orders coming in before the last ones have gone out. He is not writing them down. He holds the whole picture in his head. Table six has been waiting on their dal makhani, table eleven wants the biryani held back ten more minutes, the family in the corner sent the naan back because it came out too hard, and someone at table three has just realised their lassi was salted instead of sweet. He moves between the dining area and the kitchen without breaking stride, managing the kitchen, managing the mood of a room full of people who are hungry and impatient. Onlookers watch him and see someone who has the whole room under cool calm control.
And then the last order is served. The kitchen closes. The room empties. And if you asked him at the end of the shift how many biryanis went out that evening, or which table had the young couple who kept changing their order, he would not be able to tell you. The moment each order was completed, it was released from his memory entirely. The mental load he had been carrying so impressively in the moment was wiped clean. Not because he has a terrible memory. Because completion is how the mind works. The loop closed when each order was completed, and the mind lets go of the information.
This is what I am watching happen with AI.
The people I work with are producing more than they have at any point in their careers. A report that would have taken three days is done by noon. A proposal that needed two weeks of iteration is drafted, refined, and sent overnight. And at the end of the day, when they close the laptop, something has been wiped clean. It is not the files. It is their thinking.
When AI closes the loop really quickly on tasks, the mind releases the mental effort of thinking. The question that needed to stay open a little longer, the situation that needed to be mulled over, the big picture that is still forming, all of it gets handed off and released from your brain. The processing that would have happened in the discomfort of not yet knowing, in the slow repeated return to something unresolved, does not happen. It looks like efficiency. And it is the loss of something the waiter did not need in his job, but you do. The discomfort of staying with something unresolved is where your thinking forms.
The signals you need to process are not just about the work. They are also about what is happening in the organisation around you.
You need to zoom out and understand what priorities are shaping the decisions above you, where the pressure you are absorbing to do more is coming from, what is shifting in the conditions around you that you have not had the headspace to read. These are not questions that have easy, instant answers, even if AI gives you instant, polished and believable ones.
These questions need to stay open. They need to be carried. The temptation is to answer them with whatever offers comfort and an instant answer, and today that is your favourite 'AI thought partner'.
But these answers come from what you have been sensing, and that cannot be typed into a prompt.
When the working day is a sequence of completed loops, handed off and released by your efficient AI tool, you are not staying with those signals. The dynamics around you keep shifting, and you have stopped observing. Staying able to read those dynamics is the heart of political literacy at work.
There is another hidden cost, and it compounds with time. What I see in my clients who work in technology or management is that AI does not reduce the work they do. It expands it.
More gets produced quicker, so more gets expected. The expectations come from your own self and from the powers that be. You take on more than your role asks because you have shown the people around you that you will do so without dropping the ball. So the work keeps coming. And when you are carrying that much, moving that fast, you will make a mistake. It is only a question of when. When it comes, you will be the one held responsible for it. Your feedback will not say you were given too much. It will say you are not strategic, that you execute well but do not lead. The increasing workload will be forgotten, but the mistake you made will not be. You would be left wondering why something so trivial was blown out of proportion, and why it pulled your performance review down.
The way out is not to produce more carefully. It is to make time to think deeply about what is going on around you.
This is something most people leave abstractly for "the day I don't have many meetings", and it is one of those things where intent does not translate into action unless it is explicitly designed. The thinking does not have to be done alone. It can be a facilitated reflection with a coach. It can be a conversation with a trusted person who knows the terrain. It can be done by yourself. What matters is that you form your own view from your own signals first. You go to technology after you have understood what you are seeing, not before. Outsourcing the production is one thing. Outsourcing the formation of the view is how you go blind.
The waiter's memory was wiped clean at the end of service. It did not matter much to him, because his job was done when the last order was delivered. Your job is not done when the output is sent. Your job includes knowing what is happening around you, and that knowing requires loops that stay open long enough to form something.
Production is becoming cheap and universal. Anyone can generate the report, draft the proposal, clear the queue. The skill that is becoming scarce, and valuable, is the one AI cannot do for you. Reading the patterns around you. Forming a view before you reach for the tool. This is the skill that will separate people in the years ahead, and it is the first one to disappear when the loops keep closing before you have had the chance to think.
When did you last sit with something unresolved long enough for your own thinking to form?
Frequently asked questions
How does AI make you easier to blindside at work?
When AI closes your loops quickly, your mind releases the effort of thinking. The questions about what is shifting around you, where the pressure to do more is coming from, and what priorities are driving the decisions above you stop being carried. You produce more and observe less, so you miss the signals that would have warned you.
Why does using AI to be more productive expand your workload?
More gets produced quicker, so more gets expected, both from yourself and from the people above you. You take on more than your role asks because you have shown you can deliver without dropping the ball. The work keeps coming, and when you eventually make a mistake under that load, the extra workload is forgotten while the mistake is what your performance review remembers.
What is the skill AI cannot replace at work?
Reading the patterns around you and forming your own view before you reach for the tool. Production is becoming cheap and universal, so the scarce and valuable skill is understanding what you are seeing in the organisation first, then going to technology, not the other way round.