The story: Entry-level job postings have dropped 14% since ChatGPT’s rise. Junior analyst roles may cease to exist. AI agents now handle writing, analysis, coding, customer support, and even interviews (Amazon’s 24/7 AI interviewer).
The numbers:
- Anthropic reports 70%+ automation risk in programming, customer service, data entry
- These are roles often held by educated, higher-paid workers
- Young workers (22-25) hired 14% less into AI-exposed jobs
- Nearly half of London jobs at risk (women hit hardest)
Why it matters: The career ladder is being pulled up. Companies aren’t just automating tasks—they’re automating the foundational work that taught juniors how to think. No entry-level work means no skill development. No skill development means no mid-level pipeline.
The take: This isn’t a temporary disruption. It’s structural. The roles that used to train the next generation of workers are gone. Companies will discover this matters when they need senior talent and find no one to promote. But that’s a future problem. Today’s grads bear the cost.
🎓 The Experience Gap Crisis: Agentic AI’s Hidden Tax
The story: Agentic AI boosts senior productivity by 44%—but slashes foundational tasks (-42%), skill progression (-60%), and mentorship (-55%). New grads face 50/50 odds of landing relevant work.
Why it matters: Seniors use AI to skip the work juniors used to learn from. When you automate the grind, you automate the teaching moments. Mentorship drops because seniors don’t need to delegate—they just prompt. The result: a generation stuck at “entry-level” with no way up.
The take: AI creates a productivity paradox: it makes seniors more efficient while making juniors obsolete. Companies will eventually need to rebuild training pipelines (“simulated foundational learning”) but that costs money. Easier to just hire fewer juniors and wonder why the pipeline dried up.
💼 92,000 Tech Layoffs YTD: The AI Wealth Transfer
The story: Year-to-date tech layoffs exceed 92,000 as companies cut staff to fund $670B+ in AI spending. Meta cutting 8,000 (10% of workforce). Oracle reportedly cutting 20,000. Microsoft offering buyouts.
Why it matters: The math is brutal: compute costs sometimes exceed salaries, but layoffs are immediate while AI ROI is speculative. Companies are betting automation will eventually replace what they’re cutting. Workers can’t wait for that bet to land.
The take: Call it what it is: a wealth transfer from workers to AI infrastructure. Layoffs boost stock prices. Stock prices fund AI capex. AI capex may or may not deliver returns. Workers definitely lose. This isn’t creative destruction—it’s financial engineering with human collateral.
🚀 Rising Demand: AI-Savvy Skills Nearly Double YoY
The story: Entry-level jobs requiring AI skills nearly doubled year-over-year. Success now demands “agentic AI builders” who master workflows, tool-calling, multi-agent systems—not basic prompting.
What’s needed:
- Workflow orchestration (connecting multiple AI tools)
- Tool-calling and API integration
- Multi-agent system design
- “AI-adjacent” skills: ethics, critical thinking, human networking
- Durable skills that AI can’t replicate (yet)
Why it matters: The bar has moved. “I can use ChatGPT” is table stakes. Employers want people who can build AI-augmented workflows, not just consume them. The gap between “AI user” and “AI builder” is where careers live or die.
The take: Upskilling isn’t optional—it’s survival. But here’s the catch: most entry-level workers can’t learn these skills on the job because entry-level jobs are gone. Self-directed learning is the only path. That favors the privileged who can afford unpaid study time.
🛠️ Career Advice for 2026: Build, Don’t Just Prompt
The playbook:
- Master agentic AI now — Learn to orchestrate workflows, not just write prompts
- Focus on human strengths — Judgment, ethics, relationships, context
- Build portfolio projects — Show, don’t tell. Deploy something real
- Network intentionally — AI can’t replicate genuine human connections
- Specialize in AI-adjacent domains — Compliance, safety, evaluation, governance
Why it matters: AI augments winners and accelerates divides for the unprepared. The workers who thrive will be those who treat AI as a tool to amplify their capabilities, not a replacement for developing expertise.
The take: “Learn to code” was the advice for 2016. “Learn to build with AI” is the advice for 2026. But coding bootcamps existed. AI builder pathways don’t—yet. Early adopters who figure this out now will have a massive advantage.
🔍 THE BOTTOM LINE
The career landscape is bifurcating: AI builders (who orchestrate systems) and AI subjects (whose work gets automated). There’s no middle ground. Entry-level roles—the traditional on-ramp—are vanishing fastest.
The brutal truth: if you’re not actively building AI-augmented workflows, you’re being positioned for replacement. Upskilling isn’t career development anymore. It’s job security.
Related Singularity.Kiwi coverage:
- Maharashtra AI Policy 2026 — How AI policy shapes job markets
- Technology & People: Meta’s 8,000 Cuts — The human cost of AI spending