Advice about what to learn follows whatever is currently prominent, which produces a great many people learning the same thing at the same time. A few things hold value independently.
Writing clearly
The most undervalued professional skill there is. Distributed work runs on written communication, decisions are made from documents, and the person who can explain something in writing has disproportionate influence.
It is also learnable, unlike much of what gets recommended.
Working with data
Not data science specifically. The ability to open a spreadsheet, ask whether a number means what someone claims, and check it. Most organisations make decisions from numbers nobody has verified.
Understanding what a business is trying to do
The difference between someone who executes tasks and someone who is consulted is usually this. It cannot be learned from a course and is acquired by paying attention.
One technical skill, deeply
Deep beats broad for employability. Being genuinely good at one thing is more valuable than being adequate at six, and it is what gets you into conversations where the broad knowledge then helps.
How to judge a specific skill
Ask whether it is a tool or a capability. Tools change – the specific software, the specific framework, the specific platform. Capabilities persist – understanding how systems fail, how to structure a problem, how to communicate a result.
Learn tools as needed and invest in capabilities deliberately.
On AI specifically
Being able to use these tools competently is becoming a baseline expectation rather than a differentiator, the way spreadsheet literacy did. Worth having and not worth building a career on.