Agentic Foundation London Chapter 11 March 2026

Discussed how AI agents can transform knowledge work through automated tagging, semantic organization, and autonomous task execution. The core insight: agents work better when information is structured with domain-specific ontologies and markdown blocks, enabling them to find relevant knowledge and curate personalized feeds without manual effort.

The group demonstrated practical examples, including an automated video portal that generates chapter summaries and a fashion-design workflow that produced 3D assets and VFX pipelines in hours using Comfy UI, SAM 3D, and Blender. However, they identified critical unsolved problems: fine-grained access control for agents accessing corporate files remains largely unaddressed, and autonomous agents sometimes deviate from creative briefs when proper constraints aren't embedded.

Key focus areas include building ontology-driven systems for knowledge organization, designing agent-facing products rather than consumer tools, and addressing security concerns around agent-to-agent communication and provenance. The team also discussed the infrastructure challenge: running complex workflows requires substantial compute resources, though renting cloud GPUs offers a practical path forward.

Practically, they're planning London workshops and lab-based immersion sessions to operationalize these approaches, with emphasis on combining agent automation with human oversight to maintain both efficiency and control over outputs.