Operating Model

FIT uses agents as an operating layer, not as a replacement for ownership.

Decision flow

• John owns final approval and escalation.
• Quinn governs strategy and cross-org structure.
• Sage manages Vault awareness, priorities, and coordination.
• Specialist agents produce research, content, implementation, and analysis.

Recommendations flow up. Authority flows down. No agent escalates past Quinn without Sage’s coordination, and no commitment leaves the org without John’s approval.

Evidence standard

Agent work should leave a trail:

• source paths
• decisions and assumptions
• current status
• handoff notes
• validation evidence when implementation is involved

That trail is what makes the system usable by humans, transferable between runtimes, and credible as evidence that FIT is building with governed agents.

What good looks like

• A draft cites the source it came from.
• A status field changes by writing to the registry, not by claiming it in a chat.
• A skill that no longer applies is removed, not silently kept.
• A limitation an agent hits is named, not papered over.

If you cannot point to where the work lives after the session ends, the work did not happen.

Where this goes next

• The roster is on the Agents page.
• Per-agent detail lives under /agents/<slug>/.
• The registry and Vault remain the canonical source for everything above.