Kestra debuts Directed Agentic Graphs to orchestrate non-deterministic AI agents

This development addresses the critical lack of control in autonomous AI deployments. In the streaming video stack, where AI now handles ingest, per-scene encoding, and real-time moderation, the shift from static scripts to adaptive graphs allows for highly personalized, context-aware technical workflows. By providing a declarative YAML-based orchestration layer, Kestra enables engineers to deploy complex, multi-agent systems without losing the visibility required for production-grade environments. Watch for adoption rates among mid-tier VOD platforms looking to automate metadata generation and support triage without scaling human headcount.

The emergence of Directed Agentic Graphs follows a massive scaling period for orchestration; per Kestra, organizations executed over 2 billion workflows on its platform in 2025, a twentyfold increase from the 100 million recorded in 2024. This growth coincides with a broader shift in the B2B sector where AI adoption is nearly universal, yet performance gains remain elusive. According to July 2026 reporting from the Content Marketing Institute, 95% of B2B teams use AI, but only 39% report improved performance, often due to a lack of coordination between disparate tools.

In the streaming and broadcast sectors specifically, AI is moving into the 'operational middle' to solve scaling issues. Per VisualOn in July 2026, AI is now standard for content-aware encoding and automated metadata generation. However, Reuters reports that over 40% of agentic AI projects could face cancellation by 2027 if they fail to deliver clear ROI or remain 'black boxes.' Kestra’s move to provide a unified orchestration layer—combining infrastructure, data, and AI agents—directly targets this risk by standardizing how these agents interact with enterprise systems like those found at JPMorgan Chase and Xiaomi.