'We have maybe 20 months' to rebuild for AI agents, Meta's infrastructure VP tells VB Transform 2026

Key Points:

• Meta's VP of Engineering Barak Yagour highlighted a major shift as agentic AI queries to Meta's data systems increased 30-fold in six months, challenging two decades of infrastructure built primarily for human users.
• The rise of agentic AI is breaking core infrastructure assumptions around capacity, identity, and velocity, requiring Meta to develop agent-aware controls, dynamic throttling, and new identity frameworks to manage autonomous agents.
• Meta is implementing trusted data environments to maintain governance and human oversight while granting agents more autonomy in data exploration and decision-making, ensuring outputs are traceable and sensitive data is protected.
• The company is transitioning from batch ETL to real-time streaming data pipelines and schema-aware storage to support reasoning models that demand richer, more timely behavioral data for personalized recommendations.
• Yagour emphasized the urgency for the industry to rebuild infrastructure to support a future where humans and AI agents collaborate at scale, estimating a 20-month window to adapt before current systems become obsolete.