Why Meta’s plan to replace employees with AI was rolled back

AI summary focused on the core points of the original article.

1. The rollback of a plan to replace up to 60%

Meta had tried to cut up to 60% of some team headcount through “Project OT (Organizational Transformation)” and replace it with AI agents.

The goal was to quickly transform the organization into an “AI-native” form overall.
However, productivity and stability problems were confirmed in internal data, and the plan was rolled back just before execution.

2. Workload increased, but results did not follow

There was a wide gap between the activity created by AI and actual user value.

Internal code change counts increased 220% year over year.
- Changes that actually led to new or improved features increased only 36%.
Technical and security incidents rose 40%, and the time employees spent dealing with them increased 70%.
More output, in turn, increased the burden of verification, security, and maintenance.

3. Employee trust was shaken as well

The process of teaching AI employees’ work methods triggered strong backlash.

Meta installed tracking software on employee devices that recorded keyboard input and mouse use.
Anxiety grew that they were training their own replacement with AI, and morale-boosting measures such as travel and in-house event support or snack improvements did not have an effect.

4. AI productivity must be measured by results

Even if AI produces 10 times more work output, that does not mean business value is 10 times higher.

Actual improved business results should be checked rather than code volume or processing counts.
After verifying performance on limited tasks, the permissions of agents should be expanded step by step.
Before removing people from work, it should first be proven that judgment, review, and accountability structures operate safely.

XPLORED View

This is a case that shows in concrete numbers the operational cost an organization pays when it mistakes AI activity for productivity.

Good situations for reading

When reviewing AI automation scope and workforce plans together
When designing metrics to evaluate AI adoption results

Key takeaways

Output volume and actual business results must be measured separately.
Before replacing people, first verify the permissions and review structure of agents.

Read original

This article is summarized and interpreted by XPLORED based on publicly available original sources and related materials. Rights to the original source and materials included in it belong to their respective rights holders. If a rights holder requests a correction or removal, we will review and respond promptly.

Get the week's signal by email, every Saturday.

A short read on the tools and articles added that week.