AI agents are rapidly changing the tech industry as companies automate workflows, reduce manual tasks, and experiment with AI-powered productivity systems. But while many CEOs believe AI agents can replace large parts of human work, experts warn that the technology still has major limitations and risks.

But as companies rush to automate work, cut costs, and deploy AI agents everywhere, a growing number of experts are warning that some tech leaders may be overestimating what AI can actually do today.

The conversation recently gained attention after Box CEO Aaron Levie made a surprisingly honest statement on social media, saying many CEOs may be suffering from what he called “AI psychosis.”

His comments have sparked a much bigger discussion across the tech world about whether companies are moving too fast with AI-driven layoffs and automation.

What Aaron Levie Actually Said

Aaron Levie, founder and CEO of Box, shared his thoughts on X (formerly Twitter), explaining that CEOs are often too far removed from day-to-day operational work to fully understand AI limitations.

According to Levie, executives usually experience only the “happy path” of AI.

For example:

  • Generating reports
  • Creating quick prototypes
  • Drafting contracts
  • Writing simple code

From those successful examples, some leaders quickly jump to believing AI agents can fully replace employees.

But real-world work is far more complicated.

Levie pointed out that someone still needs to:

  • Review AI-generated code
  • Detect software bugs
  • Verify hallucinated information
  • Train AI systems properly
  • Check legal contracts carefully
  • Monitor automation systems

In short, AI still requires significant human oversight.

Even AI Supporters Are Raising Concerns

What makes Levie’s comments especially interesting is that he is not anti-AI.

In fact, he is one of the stronger AI supporters in the tech industry.

He regularly talks about:

  • AI-powered software
  • AI agents
  • Automation tools
  • The future of AI-driven workplaces

He also invests in AI startups personally.

That’s why his warning carries extra weight.

Instead of rejecting AI, Levie argues that leaders should use AI extensively before making major business decisions based on unrealistic expectations.

AI Layoffs Are Increasing Across Tech

The debate comes during a massive wave of tech layoffs.

According to Layoffs.fyi, over 115,000 tech employees have already lost their jobs in 2026 across more than 150 companies.

Many businesses have openly connected these layoffs to AI productivity improvements.

Some companies claim AI can now handle tasks previously done by large teams.

Others say automation allows them to operate with fewer employees.

ClickUp’s AI Strategy Started More Debate

One example that gained major attention was productivity software company ClickUp.

CEO Zeb Evans revealed that the company deployed thousands of AI agents internally and later reduced its workforce by around 22%.

Evans described his vision as building a “100x organization,” where employees mainly supervise and review AI-generated work instead of completing tasks manually.

The announcement triggered intense online debate.

Some people praised the move as innovative.

Others worried it represented a dangerous overconfidence in AI automation.

Research Suggests AI Productivity Gains Are Still Unclear

Despite massive investment into AI, research studies are showing mixed results when it comes to real productivity improvements.

Several academic reports suggest that while AI can improve efficiency in some tasks, the overall gains may not be as dramatic as companies expect.

One study published in the California Management Review found no strong evidence linking AI adoption to major overall productivity growth.

Another report from the National Bureau of Economic Research found that people often feel more productive using AI, even when measurable improvements remain limited.

This growing gap between perceived productivity and actual results is becoming a major topic in the AI industry.

AI Agents Still Struggle With Human-Level Work

Researchers from MIT recently tested large numbers of AI agents across different tasks.

Their findings showed that AI systems still struggle with:

  • Accuracy
  • Reliability
  • Complex decision-making
  • Human judgment
  • Context understanding

The researchers predicted that AI may reach “basic competence” for many text-based tasks by around 2029.

However, outperforming humans consistently could still take several more years.

That means businesses replacing large parts of their workforce today may be moving ahead faster than the technology itself.

The Real Risk Could Be Organizational Chaos

Another challenge is that AI doesn’t remove management bottlenecks.

Research published in Harvard Business Review suggests that when AI helps everyone produce more work faster, executives may actually become the new bottleneck.

More content, reports, code, and decisions still need human approval.

If organizations automate too aggressively without proper systems and oversight, companies could face:

  • Lower quality output
  • Operational confusion
  • Poor decision-making
  • Increased errors
  • Internal chaos

Final Thoughts

Artificial intelligence is absolutely changing the future of work.

There is little doubt that AI tools, automation, and agents will become deeply integrated into modern businesses over the next decade.

But many experts now believe the biggest danger may not be AI itself it may be unrealistic expectations about what AI can currently achieve.

The companies that succeed long term may not be the ones replacing humans the fastest.

Instead, they may be the ones that learn how to combine human expertise and AI tools in the smartest and most balanced way possible.

Read More on VitalStack

Enjoyed this article?

Subscribe for weekly deep-dives on AI and health — straight to your inbox.