The Wall Street Journal published a column by Christopher Mims on June 21, 2026, titled "What AI Can't—or Shouldn't—Do for You," arguing that even as companies rapidly adopt AI, there are areas where the technology should not be put to work.
June 21, 2026 · Opinion
What AI Can't—or Shouldn't—Do for You
Against the mantra that AI beats humans at nearly everything, a survey of lab CEOs, engineers and Nobel and Turing laureates argues: where empathy, authenticity and accountability are on the line, humans remain superior—or essential.
The "Project Plowshare" warning
Cold War engineers detonated hydrogen bombs in the name of "peaceful use"—and spread radioactive harm. Overconfidence in a powerful tool can do real damage. Misused AI can trigger mass layoffs, customer attrition and erosion of corporate value .
One nurse, AI as a tool — not a replacement
15–20
hours/week saved with AI transcription + claims tools
300
patients served by a single solo practice
100%
legal AI output kept under mandatory human review
Where humans still hold the advantage
Empathy-heavy service
Sensitive moments—like a widow canceling an account—shouldn't be left to AI.
Marketing
Firms hire real human storytellers to avoid "AI slop."
Regulated fields
Payments, health, legal—black-box AI lacks explainability and accountability.
Talent development
Replacing juniors with AI loses institutional knowledge and successors.
The broken pipeline — a self-defeating loop
Replace junior staff with AI
→
Institutional knowledge lost
→
No successors left
→
Rehire / add staff later
AI is useful as a tool—but it is not omnipotent.
Practitioners and experts converge on one view: there are areas where humans must supplement and supervise it—against scaled-up bad decisions, skill degradation and data bias.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…