How Meta Uses AI Agents to Scale ML Innovations Across Ads Ranking Models
How Meta Uses AI Agents to Scale ML Innovations Across Ads Ranking Models
Meta describes "technique propagation," its process for adapting proven ML techniques across its diverse fleet of ads ranking models, and how AI agents now handle the judgment-heavy steps of discovery, generalization, and adoption. Since agents began driving execution in March 2026, Meta estimates they handle over 60% of all technique propagation, with early results showing 3x greater engineering productivity, a median 4x reduction in training compute across select combinations, and an estimated $45M+ in compute savings by the end of the year. Meta says it is ramping agent-driven workflows to be the primary modality for propagation, aiming to cut the time for every proven technique to reach every applicable model from over a year to three months.
Source: Meta Engineering · engineering.fb.com