GitLab cuts staff and shifts to autonomous AI agents

GitLab is transitioning to AI agents for automation of internal operations and announced a workforce reduction. This is one of the most concrete examples of how a large company is restructuring itself under the assumption that intelligent agents will become the primary work tool.

How GitLab is transitioning to agents

The company acknowledges: software development and DevOps are the first candidates for automation through AI. Instead of hiring people for code reviews, PR approvals, and cross-team synchronization, GitLab is betting on agents. The idea is that agents will be able to take tasks, write code, review each other, and deliver results — faster and without negotiations.

What's changing in the structure

The reorganization will affect three key areas:

• R&D is divided into approximately 60 small autonomous units instead of large team groupings
• Management hierarchy is simplified — flattening management layers means fewer coordination levels
• Geographic presence is reduced by 30% — likely closing offices in regions with lower activity
• AI agents take over code reviews, approvals, and handoffs between teams

This essentially means the structure becomes more grid-based and less hierarchical — centralized management is replaced by agent-driven management.

Why this is happening now

GitLab sees that DevOps and development have already advanced far enough in automation. Git, CI/CD, containerization — these are no longer manual work. The next step: automate the development organization itself. Agents can read code, see which parts are interconnected, suggest steps, even write parts of solutions.

"This is the first real scenario for deploying agents in production," analysts in the

DevOps community believe.

What this means

For developers: tools will shift toward agent-first design. For companies: reduction in code review and DevOps positions, but growing demand for people who can configure, verify, and guide agents. For the industry: this is a signal that the age of agents is not in the future — it is already beginning in development tools.