The famous thesis is two modes of thinking: one fast, automatic and emotional, the other slow, deliberate and lazy. On that frame Kahneman walks through decades of experiments on how judgement goes wrong in predictable ways.
What I took from it
The planning fallacy and its remedy, which is the part I use most. We estimate badly because we estimate from the inside: we imagine the steps of the work and add them up. The correction is to look from the outside, meaning to find how long similar jobs took and start there, even though this case looks different. It always looks different. Estimating that way, the number I give comes from comparable cases and the task list only serves to distribute it.
Anchoring is the second. The first number in a conversation moves everything after it, even when it is absurd and everyone knows it is. That has an immediate operational consequence: the estimate gets written before hearing the client's budget, not after. Hear the budget first and it is no longer an estimate.
The third is loss aversion, which explains why a company stays with a system that does not serve it: changing has a visible cost and staying has an invisible one, and they do not weigh the same even when the number is identical. Naming that in a conversation is more useful than insisting on the return on investment.
What I left out
The chapters on priming, which rest on experiments that later failed to replicate. Kahneman himself publicly acknowledged putting too much faith in that literature. That the most quoted part of a celebrated book ended up in doubt is, by itself, a lesson in how to treat evidence. What held up is what he researched himself: anchoring, loss aversion, planning. I also left the final section on happiness, which is another book inside the book.