Work
300 million jobs exposed. Knowledge work transforms. Tasks disappear before jobs do. The question isn't whether your job will change — it's whether you're ready for how fast.
The core dynamic nobody is talking about
Not jobs disappearing. Tasks unbundling. Every role is a bundle of tasks. AI eats them one at a time.
Here's what actually happens: a job is not a single thing. A radiologist reads scans, writes reports, consults with doctors, manages patients, runs a practice. Each of those tasks has a different vulnerability to automation. AI doesn't replace the radiologist in one shot — it takes the scan-reading first. Then the report drafting. Then the referral triage. Each task gone changes the economics of the role.
This is already happening in every knowledge sector. The people who notice first aren't the ones who lose their jobs — they're the ones who find their role shrinking to the tasks AI can't do yet. Which is worse, in some ways. You keep the responsibility but lose the parts that made the job viable.
Benedikt Evans calls this "the jellyfish" — a role that still exists but has lost its substance. It's happening everywhere.
The numbers
Goldman Sachs research in 2023 estimated 300 million jobs exposed to AI automation. That was before GPT-4. Before Claude 3. Before the capability jumps of 2024-26. The actual number is likely higher now. The shape is clearer: every knowledge role gets compressed, and the compression accelerates.
Who gets hit first
Not a simple binary of "safe" and "not safe." It's a cascading wave that hits different roles at different times.
Data entry, basic coding, customer support tier-1, translation, content production, accounting clerks, legal document review, transcription, scheduling, basic design production. These aren't future scenarios. These roles are already being unbundled. The AI-native companies of 2024-26 are built around not hiring for these at all.
Software engineering (junior to mid), marketing operations, consulting analysis, teaching (content delivery, not classroom), journalism (reporting and drafting), project coordination, paralegal work, medical triage and diagnostics. These roles still exist but the entry-level pipeline is drying up. Companies hire fewer juniors because AI does what juniors used to do. Mid-career people find their teams shrinking.
Physical trades (electrician, plumber, mechanic), healthcare bedside (nursing, physical therapy), creative direction and strategy, human relationships (therapy, coaching), leadership and high-stakes judgment. These aren't immune — they'll be augmented, transformed, made more productive. But the core human presence remains the bottleneck. For now.
What actually changes about work
The entry-level collapse
The most underreported story in AI and work. Junior roles have traditionally been where people learn the ropes — grunt work that builds intuition. AI can now do most entry-level work better than a new grad. Companies respond by hiring fewer juniors. The career ladder's bottom rungs are being removed. The apprenticeship model of knowledge work is dying, and we don't have a replacement.
The one-person team
A senior engineer + AI can now do the work of a team of five. A single designer + AI handles what a studio used to. This isn't efficiency — it's structural. Companies discover they need far fewer people to achieve the same output. Hiring freezes follow. Layoffs follow. The survivors are productive but precarious — one bad quarter and there's no team to absorb the work.
Skill half-life collapses
What you learned two years ago may already be partially obsolete. Not because the knowledge is wrong — because AI can do it better. The premium shifts from "what you know" to "what you can figure out with AI." The specialists who spent decades building deep expertise find themselves competing with AI models that have read everything and forgot nothing.
Remote work goes hybrid-AI
Companies already struggled with remote collaboration. Now add AI agents into the mix. Is that Slack message from a human or an agent? Which parts of the project are human and which are AI-generated? The boundary blurs. New protocols emerge — AI-handled meetings have different rhythms, different expectations, different accountability. Nobody has good norms for this yet.
Timeline
These aren't predictions. They're structural trajectories based on what already exists.
How to prepare
Become fluent with AI tools
Not "learn to use ChatGPT." Learn to build workflows with AI. The people who thrive aren't the ones who can prompt well — they're the ones who know how to decompose problems, feed them to the right tools, and validate the outputs. This is a meta-skill that compounds.
Invest in judgment
AI can generate. It can analyse. It struggles to decide when the right call isn't the optimal call. Judgment — knowing which problem to solve, when to ignore data, how to weigh competing values — becomes the premium human skill. This is earned through experience, not courses.
Build relationships before you need them
Network effects have always mattered. In an AI-augmented world, they matter more. The roles that survive longest are the ones embedded in human trust networks. People hire people they trust. AI doesn't have that advantage. Nurture your professional relationships while you have the time.
Stay adaptable
Your 2030 job probably doesn't exist yet. The best career insurance is the ability to learn new domains quickly. Depth in a single skill is riskier than breadth plus the ability to go deep on demand. Think "portfolio of capabilities" rather than "career ladder."
The optimistic case
It's not all grim. There's a genuine upside if we handle this right.
Boring, repetitive work disappears. Everyone gets PhD-level assistance for everything — learning, planning, creating. The workweek compresses to 4 days, then 3. Solo founders build billion-dollar companies from their living rooms. Humans are freed for creative pursuits, community building, and the things we actually want to do with our limited time.
The optimistic case requires policy choices: shorter workweeks, UBI, retraining infrastructure. But it's not fantasy — it's what happens if we treat the productivity gains as collective benefit rather than concentrated wealth.
The bottom line
Work isn't going away. But the structure of work — the career ladder, the 40-year trajectory, the implicit social contract — is already dissolving. The people who adapt fastest are the ones who treat AI as a partner, not a threat. The people who wait for things to stabilise will be waiting a long time.
Work becomes optional for some. For most, it becomes different. The difference matters.