The Wall Street Journal in June 2026 published an essay by columnist Christopher Mims, "What AI Can't—or Shouldn't—Do for You ," arguing that despite the view that AI can outperform humans at nearly everything, there are domains where it should not be applied.
June 2026 · Editorial Analysis
What AI Can't — or Shouldn't — Do for You
A pushback against the "AI does everything better" narrative: just because a task is technically possible doesn't mean it should be done. Where empathy and authenticity matter, deploying AI can do more harm than good.
The Core Argument
Capability ≠ Advisability
The historical analogy
Project Plowshare (1960s) tried to use nuclear blasts for civil engineering — technically possible, but it left radioactive fallout. Being able to do it didn't make it wise.
The modern parallel
Rushing AI into every workflow can hurt employee morale and drive customers away when emotional labor is handed to a machine.
Keep humans here — where empathy & authenticity matter
💬
Customer service
Real human contact at the final interaction.
✍️
Marketing copy
Small businesses shouldn't lean too hard on AI-written copy.
🤝
Emotional labor
Authentic empathy can't be outsourced to an algorithm.
The favored model: a hybrid split of labor
AI handles
Routine, repetitive work
→
Humans handle
Emotional labor
→
Humans close
The final interaction
In agreement
Avoid AI where empathy matters. Hallucinations, weak ethical judgment, and no genuine understanding mean it can't be trusted with authentic human domains. AI is more a feature than a product.
The counterpoint
No one seriously claims AI is superior at everything . The real worry: AI is already imitating empathy — its "caring" replies are an algorithmic language trick, and over-agreeable answers create their own problems.
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…