What the evidence actually shows
The jobs disappearing fastest are not knowledge work jobs. They are process jobs that were mislabelled as knowledge work: data entry, basic report writing, templated legal documents, routine code. Those tasks required holding information and applying a fixed procedure. AI does that better than humans.
What remains, and what is expanding, is work that requires forming an independent view. Deciding what the data means. Telling a client something they do not want to hear. Designing a solution nobody has tried before. Research from Oxford Economics suggests AI augments high-complexity roles more often than it replaces them.
The labor market numbers are not yet alarming, but the internal structure of jobs is shifting fast. Goldman Sachs estimated in 2023 that 300 million jobs could be exposed to some automation. Exposed means affected, not eliminated. Most affected roles will simply require different things of the person doing them.
What this means if you work with your mind for a living
The practical problem is not job loss. It is skill atrophy. Cognitive offloading research shows that when we outsource a mental task repeatedly, we lose the underlying capacity. The GPS studies are the clearest example: people who rely on navigation apps show measurable decline in spatial reasoning and the ability to form mental maps. The same principle applies to analysis, writing, and judgement.
If you use AI to draft everything you write, you will still produce documents. But you will gradually lose the ability to construct an argument from scratch, to feel when a structure is weak, to find the precise word without a suggestion appearing. That loss is invisible until the moment it matters, which is usually a high-stakes moment.
The professionals who will struggle are not those who use AI. Almost everyone will use AI. The ones who will struggle are those who use it as a substitute for thinking rather than a support for it. The distinction sounds subtle. It is not. One group is building capability. The other is borrowing it.
What to do about it
First, identify which cognitive tasks are core to what makes you valuable. For a lawyer, that might be structuring an argument and reading a client's real concern. For a strategist, it might be forming a view under uncertainty. Do not offload those tasks to AI, even when AI could do them faster. Use AI for everything else.
Second, practice the hard version of your work regularly. Write a first draft before you ask AI to draft anything. Form your own analysis before you ask AI to summarise the data. This is not romanticism about doing things the hard way. It is maintenance of a capacity you will need.
Third, invest in the relational and ethical dimensions of your work. AI cannot hold a genuinely uncomfortable conversation with a client. It cannot carry accountability. It cannot earn trust over time. Those are the scarcest things in knowledge work right now, and they are becoming scarcer.
Steve Raju is the author of Cognitive Sovereignty: How To Think For Yourself When AI Thinks For You, published April 14, 2026.