30,000 seats is not the same as 30,000 changed workflows

WHAT HAPPENED

Uber moves ChatGPT Work from a pilot to its whole workforce

Uber is rolling out ChatGPT Work, an enterprise AI agent, to more than 30,000 employees. It started as a pilot of 1,000 people earlier this year. Uber says it will track how employees use it and evaluate the outcomes.

THE NUMBER

80%

of the early pilot adopters were non-engineers

Finance, operations and product were among the teams in the pilot.

THE REAL PROBLEM IN COMPANIES

What I see in companies after an AI rollout

1. 01Access gets announced, but nobody defines which workflow should change
2. 02Usage gets counted in logins, not in work that actually got done
3. 03Each team experiments alone, and the learning never reaches the others

WHERE THIS BREAKS IN PRACTICE

Uber has not settled how employees will use it

Uber's VP of engineering says the approach is to give teams room to experiment. Uber did not disclose what the rollout costs. OpenAI says the companies plan to share outcomes as the rollout progresses. No results have been reported yet.

MY TAKE

Measure instead of assuming.

Seats activated is an easy number to report. It says little about the work. In practice, the companies that win will name the workflows that changed and compare them before and after. Uber says it plans to evaluate outcomes. That is the part worth watching.