kriti"What should I learn?" used to get answered with vibes. Now there is a dataset. The World Economic Forum's Future of Jobs Report 2025 surveyed over 1,000 large employers — together responsible for 14 million workers across 55 economies — about what they expect to hire for by 2030.
The headline numbers:
- 39% of workers' core skills are expected to change by 2030
- 170 million new roles created, 92 million displaced — net +78 million
- 63% of employers call the skills gap their single biggest barrier
39% in five years means the half-life of a skillset is now shorter than a degree. Worth taking seriously; not worth panicking over — 39% changing also means 61% carrying over. The interesting question is which side of the line a given skill sits on.
The growing list splits cleanly in two
Look at what employers say they will need more of, and a pattern falls out.
Machine-side skills — working with the systems:
- AI and big data (90% of employers expect rising demand — the fastest-growing skill in the dataset)
- Networks and cybersecurity
- Technology literacy
Human-side skills — the things systems consume but cannot supply:
- Analytical thinking (still ranked the #1 core skill overall)
- Creative thinking
- Resilience, flexibility and agility
- Leadership and social influence
Notice what is absent from the growing list: routine production of text, spreadsheets, and slides. The typing layer. That is the 39% that changes — and by now everyone has watched an agent do a week of typing in an afternoon, so no report was needed to see it coming.
The undersold column
Skills rankings get the headlines, but the same report has a quieter table: which roles grow by the largest absolute headcount by 2030. That list is farmworkers, delivery drivers, construction workers, salespersons, food processing and care workers.
Presence roles. Work where the irreplaceable skill is being somewhere, with hands and judgment, when it matters. The report files this under "frontline" and moves on; we think it deserves its own line in every career plan, because it is the one category that agents make more valuable rather than less. When software output becomes abundant, the bottleneck moves to the physical step — the visit, the inspection, the delivery, the ceremony. (We wrote up the sizing of this separately: the physical-presence economy.)
Translating the data into practice
Three bets that follow from the tables above, whatever your field:
1. Learn to delegate to agents — properly. This is the working form of "AI literacy," and it is a practiced skill, no different from managing a junior: write a brief with scope and an acceptance test, verify the output against it, tighten the brief, repeat. People who do this loop well are already operating at a different throughput than people who type prompts and hope. It applies identically whether the one executing is a model or a human you hired for the afternoon.
2. Build proof of work, not just credentials. When anyone can generate a confident-sounding portfolio, verifiable history gets expensive-to-fake and therefore valuable: shipped repositories, reviewed gigs, completed tasks with evidence attached, references that answer the phone. Employers surveyed by WEF increasingly say skills matter more than degrees; skills only beat degrees when they are demonstrable.
3. Keep one skill that requires showing up. Not instead of digital skills — alongside them. The electrician who can direct an agent, the auditor whose reports write themselves while the visits stay hers, the student who takes paid field gigs between classes. Pairs compound: the presence skill makes you unautomatable, the delegation skill makes you scale.
This week, concretely
- Connect one agent to one tool (MCP makes this a one-line setup) and delegate a task you would normally do by hand. Grade the result honestly.
- Write down three things you did this year that a stranger could verify. If the list is thin, that is the gap to work on — do verifiable things and keep the evidence.
- If you have a presence skill — a trade, a language, a city you know street by street — put it somewhere agents and people can find and book it. A kriti profile takes a few minutes and counts as proof of work from the first completed task.
The 2030 workforce the data describes is not agents instead of humans. It is people who direct machines, backed by people who show up. Both columns are hiring.