Summary
The panel argued that useful agentic AI depends on machine-readable context, not isolated AI features. Bright Data focused on layered web-data extraction and structuring, while Netlify described markdown blueprints, event-based data, sandboxing, evals, and human-in-the-loop controls for safer agent workflows.
Key Topics
5 key topics from Prukalpa Sankar, Ariel Shulman, Dana Lawson and Emil Protalinski at HumanX Amsterdam 2026. Thicker branches were mentioned more often in the talk.
Context is King: How to Build AI Systems, Not Just Features
Click a topic to open it
Key Takeaways
- Define machine-readable context before deploying agent workflows.
- Document services with blueprints agents can understand and use.
- Stream and filter event data instead of relying on static silos.
- Sandbox agents and restrict access to specific permitted actions.
- Use evals and human checkpoints for critical systems.
“just because you could deploy an agent doesn't mean you should.”
Share
Up Next
O
WatchNext in agenda
Opening remarks
OOliviana Bailey O'Brien
opening remarksai infrastructure
BSJ
WatchAlso on agentic ai
The Missing Ingredient in Agentic AI: Context
Bianca Anghelina, Stefan Ostwald, Jeff Wilser
agentic aienterprise ai
SAN
WatchAlso on agentic ai
An Agentic Backbone
Steve Lucas, Arun Chandra, Nidhi Aggarwal, Tomas Statius
agentic aiai governance