Jack Dorsey Cut 40% and Rebuilt: The Shape of Companies in the AI Era

Block / Square / Cash App's Jack Dorsey did an interview about why he laid off 40% of the company and rebuilt from scratch. Plus Sequoia's Roelof on the "three things a good CEO needs" — once AI can do 80%, the remaining 20% is where humans live.

Jack Dorsey (Twitter founder, now running Block / Square / Cash App) gave an interview about his recent move: laying off 40% of the company, then rebuilding from scratch.

His core thesis: companies used to need armies of middle managers because the boss could not possibly know everything, so layers of people existed to “translate”, “filter”, and “amplify” leadership intent.

AI rewrites the middle-manager role

But AI can do that now.

Decision support, planning, tracking, integration — the “core functions of middle management” — AI handles 80%. The remaining 20% is where actual human judgment lives.

So Dorsey’s logic is: rather than keep an army of middle managers who can only deliver 80%, lay them off and let the remaining people focus on the 20%.

It sounds cold. The logic checks out.

What happens after AI delivers 80%

He also said: when AI can effortlessly produce 80% of something, the remaining 20% is human “taste and judgment” — the only line of defense against products becoming mediocre copy-paste.

This matches what I have been observing. Everyone’s output volume has exploded since Claude Code, but on closer look, the outputs all look very similar. Same tools, same prompt patterns, same template-feel. The median has gone up; the gap between the median and “actually good” has widened.

Three things a good CEO always needs

Sequoia’s Roelof added in the same interview. A good CEO always needs:

Tells the truth (no performance)
Decides with logic
Has taste

All three are things AI cannot do. AI can imitate communication style, but cannot judge “in this situation, is honesty or politeness correct”; AI can list pros and cons, but “deciding under incomplete information” is a human act; AI can generate a hundred candidates, but “which one is worthy of our company” requires taste.

The lesson for individual workers

Dorsey is talking about org-level change, but it applies to individuals too:

When your work is something AI can do at 80%, your value is no longer in “doing that 80%”. It is in “filling in the remaining 20%” — taste, judgment, the courage to decide.

Spend your time on the things that train these. Hand the rest to AI. Do not try to out-speed AI on the 80% — you will lose.