Episode 190: The Agentic Desired Outcome Contract
May 7, 2026 · 1 min read
Agentic workflows need more than broad tool access: they need intent-bound authorization that ties every action back to a declared human outcome. This episode explores Desired Outcome Contracts as time-scoped, cryptographically verifiable agreements that define permitted actions, prohibited actions, consent context, evaluation rules, and execution history. We discuss how DOCs can bring human-in-the-loop consent, MCP integration, OAuth step-up, delegation, and hash-chained auditability into a practical control model for AI agents.
#Speakers
- Host: Mike Schwartz, Founder/CEO Gluu
- Guest: Nick Gamb, Principal Solutions Architect, Strata Identity
- Guest: Guest: Eve Maler, President, Founder Venn Factory
#Homework
#Takeaways
- Some common themes are emerging from the primordial soup of agentic standards. One is that we need both a natural language and machine structured ways to bound the authority granted to agents acting on behalf of humans.
- How do we express intent? Easier said then done, even with natural language. The DOC Protocol tries to solve this challenge in a more wholistic way.
- The use cases for agentic delegation vary across B2B, B2C, and C2B landscapes. For example "MyTerms" might be great to constrain authority for C2B but not for B2B.
- Schema sounds boring, but without some strategy on how to describe entities, resources and actions, interoperability may remain out of reach.