AI Does 80 Percent of the Work.
The Other 20 Are Why You Still Have a Job.

I have been arguing with the 80/20 principle for almost 20 years.

The Pareto principle: 20 percent of causes are responsible for 80 percent of results. In business, this means a small portion of customers, actions, or decisions drives the biggest effect. The rest is effort with little leverage.

But when I talk about it lately, I almost always mean the same thing: AI.

This means we save 80 percent of our work. Honestly, that is not so bad.

Why 100% AI is a massive risk

In 2024, Klarna, the Swedish payments company, announced that AI could replace human employees at scale. 700 jobs were cut, a chatbot handled two-thirds of customer contacts, and customer satisfaction dropped by 22 percent. Answers too generic, complex cases unresolved. In 2025, Klarna started hiring humans again.

That was not a failure of AI. It was a lesson in what is actually automatable — and how much smaller that share is than management assumed.

In parallel: Uber rolled out Claude Code to around 5,000 developers. 95 percent used AI tools monthly, 70 percent of commits were AI-driven. The annual budget was exhausted after four months. Individual developers incurred up to $2,000 in API costs per month.

Microsoft reportedly instructed an entire department to shut down their AI coding tool. Not because of poor quality. Because of costs that were simply too high.

The reason is almost always the same: the 20 percent is missing. Anyone who does not clearly define what is actually needed beforehand gets plenty of output from AI. Just not necessarily the right output.

What AI is good at. And what it is not.

Clearly defined tasks with unambiguous requirements: AI works very well. Code suggestions, simple functions, tests, boilerplate.

Architecture decisions, truly understanding legacy systems, figuring out why a decision was made that way back then, the most important question of all — what should we actually build? Here AI quickly reaches its limits.

And this applies far beyond coding.

The same logic applies to strategic plans. To marketing copy. To campaigns. To every task where AI delivers impressively fast first drafts.

AI writes a strategic plan in ten minutes. But it does not know your history. Not the campaign that did not work last year. Not the client who left because the tone did not fit. Not the internal tensions that quietly influence every decision.

80/20. Not 100/0. And why it will stay that way.

AI takes on 80 percent. Our 20 percent is not the remainder. It is the core.

ChatGPT has existed for almost four years. The 80/20 principle still applies. I am convinced it will remain that way for the next four as well. Less fear of AI. Instead, more clarity about what you bring to the table: your insider knowledge, your context, your judgement. These 20 percent cannot be taken from you.

Companies that use AI sensibly will not be those that reduce human expertise. They will be those that keep their best people, put AI in their hands, and make both faster together.