Unraveling "AI Takes Jobs"
The binary "AI eliminates jobs / doesn't" doesn't match reality. In fact, impact differs per task, not per occupation. McKinsey and Goldman Sachs forecasts indicate "30-50% of white-collar tasks partially automated"; occupations remain but their content changes is the realistic view.
Impact Map
Strong Impact (Automation Rate Over 50%)
- Minutes, meeting-prep material drafts
- Routine emails, quotes, first drafts of contracts
- Data aggregation/chart creation
- Boilerplate coding
- Translation/proofreading/tone adjustment
- FAQ first response
Moderate (20-50%)
- Market research/competitor analysis
- Proposal-logic construction
- Hiring-candidate screening
- Code-review first check
- Legal research, case-law search
Small Impact (Under 20%)
- In-person negotiation with customers, trust-building
- Internal political coordination/interest mediation
- New-business decision-making
- Final hiring decisions, performance evaluation
- Brand strategy, ethical judgment
Scenarios by Occupation
| Occupation | Change |
|---|---|
| Engineer | Code gen by AI, design/integration by human. 2-3x productivity |
| Marketer | Mass content generation, brand strategy is the differentiator |
| Sales | Proposals/research by AI, focus on customer-relationship building |
| Consultant | Slide creation by AI, framing/consensus-building is the value |
| Legal | Initial review by AI, judgment/negotiation strategy by human |
| Customer support | First response by AI, complex cases by human |
Adaptation Strategy: 3 Shifts
1. "Instructor" Skill
The skill to "instruct AI precisely as a subordinate" is essential. Practice verbalizing the purpose, constraints, and deliverable form, not vague requests. Prompt engineering has come closer to an ordinary manager skill.
2. Productize "Judgment"
Value gathers around people who can explain what to choose and why from the options AI mass-produces. Whether you can back judgment with business knowledge, customer understanding, and management context is the dividing line.
3. "Human-AI Hybrid Design"
The skill of redesigning workflows for "where to leave to AI and where humans intervene" (AI Ops, work redesign) is rising as a new role.
Remaining Uncertainty
- If agent autonomy advances, the "task of instructing" itself may shrink
- A scenario where regulation/ethics suppress evolution
- Areas where AI cost doesn't fully drop and humans remain cheap
- Possibility of encroaching on in-person work via multimodal support
Summary
White-collar work's essence is not "disappearing" but "contents being swapped." It shifts toward routine processing by AI, humans focusing on judgment and relationship-building. "Can you use AI like a subordinate" from now is the career fork 5 years out.
