<- all tokdocs
2026.06.11@kantrowitz

AI Has the Same Problem Computers Had in the 90s

AI
Watch on TikTok

View on TikTok ->

Boris Cherny, creator of Claude Code, draws a direct line between the AI adoption struggles of today and the PC productivity paradox of the 1990s. Speaking on the Big Technology podcast with Alex Kantrowitz, he points to a specific Harvard Business Review article that keeps shaping how he thinks about AI's real impact on companies.

The Productivity Paradox, Then and Now

In the 1990s, companies were buying personal computers at scale and replacing mainframes. Spending was up. But productivity gains were nowhere. Harvard Business Review published an article asking the obvious question: computers are everywhere, so why is output flat?

The answer was structural. Companies were bolting computers onto their existing workflows without changing how work actually got done. The technology was new, but the processes around it were identical to what came before. PCs sat on desks running the same paper-era routines, just on a screen instead of a clipboard.

The Fix That Worked (and Why Most Companies Skipped It)

The HBR article made a specific argument: to get the benefit from computers, you have to restructure your entire business process around them. Computers need to sit at the center of how you operate, not at the edge. Some companies did this. They went through the painful, disruptive work of rebuilding workflows from scratch. Those companies saw the productivity gains. The rest kept waiting for the technology to deliver results on its own.

AI Is in That Exact Same Position

Cherny sees a direct parallel playing out right now. Companies are adopting AI tools, but most are layering them on top of unchanged workflows. The description of this video adds a pointed example: Amazon reportedly had developers triggering AI automations that ran for hours and then got deleted, just to meet weekly AI usage quotas. That is the 2020s version of putting a PC on every desk and calling it a transformation.

The real work is redesigning how teams operate with AI at the center, not measuring how often people open the tool. Cherny is clear that there is no single right approach. Everyone is experimenting. But the split between companies that restructure and companies that bolt AI onto existing processes will become obvious in the next 18 months.

Key Takeaways

  • Buying the tool is not the same as adopting the tool. The 1990s PC paradox proved that technology alone does not drive productivity. The organizational change around it does.
  • Process redesign is the hard part. Companies that rebuilt their workflows around computers won in the 90s. The same pattern is repeating with AI.
  • Usage metrics are a trap. Hitting AI usage targets (like Amazon's reported quotas) without structural change produces waste, not output.
  • The productivity split is coming. Within 18 months, the gap between companies that restructured and companies that did not will be measurable and significant.
  • There is no single right approach. Experimentation is the correct posture. The mistake is assuming you can skip the restructuring entirely.

Resources

Published June 11, 2026. Writeup generated from a favorited TikTok.