How AI is reshaping career entry and what workers should do

Executive overview

Young workers entering AI-exposed jobs are seeing 16% slower employment growth — a divergence not explained by interest rates or tech sector over-hiring. Experienced workers remain largely unaffected. The risk is structural: AI now overlaps most directly with the book-based, implementation-heavy tasks that entry-level roles traditionally require.

AI cannot yet replicate tacit knowledge, strategic thinking, or social interaction — the skills built only through on-the-job experience. Used as a learning tool, AI could dramatically accelerate skill acquisition and enable workers to move between professions more fluidly.

The career ladder is at risk; the career lattice is the opportunity.

What the research found

• Study tracked millions of US workers via ADP payroll data across AI-exposed and non-exposed jobs
• Overall employment showed no major divergence — but young workers in AI-exposed roles (software, customer service, admin) saw measurable employment declines
• Experienced workers continued to grow on trend
• 16% slower employment growth for young workers in high-AI-exposure roles
• Alternative explanations tested: interest rate sensitivity, tech over-hiring, computer job exclusion — results held
• Structural change, not cyclical: if AI capabilities are driving this, the trend is unlikely to reverse

Why young workers bear more risk

• Entry-level work relies heavily on book knowledge and implementation — the tasks AI handles most capably
• Experienced workers hold tacit knowledge: hyperlocal context, strategic judgment, social interaction — built only through doing
• AI and young workers compete most directly for the same task types
• Firms have incentive to hire some junior staff for pipeline reasons, but less incentive than is socially optimal
• Young workers can leave, so firms under-invest in training them — a private vs. social incentive mismatch

What AI cannot do (short to medium term)

• Physical tasks (absent major robotics advances)
• Strategic thinking: deciding what needs to be done and directing others to do it
• Social interaction and relationship-based work
• Expressing preferences and values — determining what should be built
• Reflection: humans often discover what they want only by thinking it through

The augmentation opportunity

• Augmentation expands the set of tasks a worker can do; automation shrinks it
• A lean startup founder using AI to handle functions they never knew — an example of augmentation in practice
• AI as a personalised learning tool could compress skill acquisition and lower barriers to switching professions
• Historical precedent: education has been the most reliable way to expand worker capability
• AI for math and technical reasoning already augmenting knowledge workers (easier to verify than to generate from scratch)
• Writing is worth preserving as human: it builds understanding that outsourcing destroys

Historical comparisons

• Industrial Revolution: skilled textile workers (Luddites) lost work to new machinery — skilled workers faced more risk
• Electricity and IT revolution: opposite pattern — middle and low-skill work was displaced; educated workers benefited most
• AI's distinguishing feature: the rate of capability improvement is faster than any prior technology
• Open question: will new work created by AI be done by humans, or will AI capabilities advance fast enough to capture that too?

What young workers should do

• Use AI tools as much as possible — build with them, get fluent in directing them
• Focus on developing strategic thinking: the skill of knowing what to build and how to guide AI agents
• Treat AI as a tool for learning, not a shortcut that bypasses it
• Platforms like Khan Academy are building education-mode AI that prompts critical thinking rather than delivering answers
• The goal: understand where you as a human add value, and where the tools fall short

The career lattice vision

• If AI accelerates learning, moving between professions as demand shifts becomes far more feasible
• This requires changes to education systems — universities and earlier — to help people learn faster
• Lower barriers to expertise could compress inequality in the labour market
• A career lattice — flexible, multi-directional — is more resilient than a career ladder under technological disruption