AI Doesn’t Care What Happens Next. You Do.

Christian Kromme

automation

September 22, 2026

The lights come up after a keynote, the microphone goes into the room, and the same question arrives in a dozen versions. Sometimes it comes from a CFO in the front row, sometimes from a young engineer at the back. If AI makes decisions faster, cheaper and more reliably than we do, why should humans still make them at all. And then the quieter version, the one people ask when they catch me at the coffee table afterwards: are we still relevant in a world where machines do most of the thinking.

My answer is yes. You’ll make fewer decisions in the years ahead. The ones you keep will weigh more than anything you decide today.

The fear under the question

That quieter question is about meaning, and the data shows how far the worry has spread. Pew Research Center found in 2025 that half of Americans feel more concerned than excited about AI in daily life, up from 37% in 2021. By 2026 the share had climbed to 52%. In the same research, 53% expected AI to weaken people’s ability to think creatively, and half expected it to damage our ability to form meaningful relationships. Among adults under 30, those numbers rise to 61% and 58%. The generation that uses chatbots most is also the one most worried about what they’ll take away.

The fear has a basis. A pre-registered experiment published in Scientific Reports this year asked about 270 professionals to complete writing tasks from their own jobs. Those who copied AI output felt close to 20% less ownership of their work and about 10% less meaning in it. The loss stayed when they went back to writing by hand. People who wrote a first draft themselves and then used AI to sharpen it didn’t show the same drop.

So the risk is real, and it depends on how we use the tool. That’s a choice we still get to make.

A digital brain with nobody home

Machines will take over a large share of our decisions. Gartner expects agentic AI to make at least 15% of day-to-day work decisions on its own by 2028, starting from zero in 2024. Routing tickets, approving small orders, adjusting ad spend. Good. Nobody started a career dreaming of those.

Look at what AI sits on, though. Organizations have built a digital nervous system in the same order evolution built ours. Automation came first: ERP, robotics, process flows. That’s the reptile brain of the company, handling reflex and routine. Social media and networks followed as the limbic layer, the place of signalling, grouping and status. AI arrived last, as the neocortex, landing on a stack that was already running.

Every layer works. And in every layer, nobody is inside. Automation keeps a company alive without any wish to survive. Platforms connect billions of people while nobody belongs to anybody. The reasoning layer produces answers, and none of those answers matter to the system that produced them. It doesn’t care whether it’s right. It doesn’t care what happens on Monday morning, when the decision lands on someone’s desk or kitchen table.

You do. You carry the consequences, and so do the people around you.

Consequences land on people

Courts and parliaments keep reaching the same conclusion. In 2024, Air Canada argued before a Canadian tribunal that its website chatbot was a separate entity, responsible for its own mistakes. The tribunal rejected that and ordered the airline to pay a grieving passenger the fare difference the chatbot had promised him. A small amount with a large principle behind it: a machine can’t hold the blame, so the people who deployed it do.

Here in the Netherlands we learned the same lesson at a far higher price. The tax authority used a self-learning risk model to flag childcare benefit fraud, with nationality as one of its risk signals. Around 26,000 parents were wrongly treated as fraudsters, and some estimates reach 35,000. Families lost homes, savings and in some cases custody of their children. A parliamentary inquiry called it unprecedented injustice, and in January 2021 the cabinet resigned. The model didn’t resign. It couldn’t. Ministers did, because responsibility needs someone who can feel its weight.

Fewer decisions, heavier ones

European law is building on that lesson. The EU AI Act requires human oversight for high-risk AI, the systems used in hiring, education, credit and public services. Under the Digital Omnibus, those obligations now apply from 2 December 2027 instead of August 2026. The date moved; the oversight requirement stayed. And the European Court of Justice already ruled in its 2023 SCHUFA judgment that a human who only rubber-stamps a machine’s output doesn’t count as real human involvement.

Put these facts side by side and a pattern appears. Routine decisions move to machines. What stays with us are the decisions where values collide, where the data runs out, where someone gets hurt whichever way you go. Who gets the benefit of the doubt. Which risk you accept on behalf of others. When to overrule the model because the number is right and the outcome is wrong.

Those decisions ask more of you. The machine will have done the analysis in seconds, so “I didn’t have the information” stops working as an excuse. What’s left is judgment, conscience and the courage to put your name under a choice.

How technology makes us more human

This is where I get hopeful. Every wave of technology has taken a function off our hands and handed something back. Machines took over muscle work and gave us room to think. AI now takes over large parts of the thinking, and it hands back what no system has: caring about the outcome, feeling responsible, sensing that something is off before any dashboard shows it.

I call that Humanification. The more digital we go, the more our humanity becomes the thing that makes the difference. It won’t happen by itself. The Scientific Reports study shows the fork in the road. Copy the machine and you lose ownership and meaning. Think first, then work with the machine, and you keep both.

So start with your own thinking. Form a view before you open the chatbot. Ask who carries the consequences of the answer, and whether those people are in the room. When a decision has a human cost, put a human name on it, and give that person the time and the authority to say no.

Machines will get faster, cheaper and more reliable every year. None of them will lie awake over a decision. That sleepless night belongs to us, and it’s part of what makes us human.

Over to you

I’m curious where you stand. Do you expect to make fewer decisions in the next five years, and will the ones you keep feel heavier? Which decision in your work would you never hand to a machine, and why? Share it in the comments. I’d like to hear where you draw the line.