Skip to main content
Adoption at scale needs a clear operating model: who runs the platform, who sets policy, who validates, who builds, and who uses. You define it; the platform makes each right enforceable through roles and permissions.
Five levels: Platform Squad (run, releases, platform guardrails and forced hooks), Architecture and Org admin (org policy, model access, org guardrails), Committees (validate compliance, approve guardrails, grant sharing rights), AI Champions (build and share approved agents), and Everyone (use agents, create for self, no share by default).

​
The squads

Four squads keep the platform alive and industrialize use cases, backed by solution architects.

​
Driving adoption: the Animation squad

Adoption does not happen on its own. The Animation squad keeps the community active with a few concrete, recurring rituals:
  • Use-case webinars: teams present what they built, and others reuse it. Sharing beats rebuilding.
  • Agent bar: a recurring clinic where anyone brings an idea and leaves with a working agent, with help on hand.
  • Agents on demand: simple agents built live in a few minutes, to show how low the barrier is.
  • Shared assets: skills, tools, MCP servers, and agent templates published for everyone to reuse.

​
The rights, level by level

Governance runs on levels, and each level can only tighten the one above. Sharing and publishing rights, and guardrails, are granted centrally, not self-assigned.
  • Platform Squad: runs and releases; sets platform guardrails and can force hooks (for example data classification controls).
  • Architecture / Org admin: sets org policy and standards, model access, and org-level guardrails.
  • Committees (compliance and security): validate the compliance criteria of an agent, approve the guardrails to activate, and grant sharing rights.
  • AI Champions: build agents and, once validated, share them to a group or the organization.
  • Everyone: uses agents and creates agents for themselves. By default, no one shares or publishes their own agent.
For the underlying roles, permissions, and service accounts, see Identity & Access. For how the levels bound each other, see Opening strategies.