Runtime Org Dynamics — Recruiter, Work-Item DAG, Self-Grown Staff

The company ontology (ontology_company.ttl) models the static org STRUCTURE — :Company / :Department / :AgentRole / :Employee. This subsystem adds the missing runtime DYNAMICS, ported (shape, not code) from OpenOPC's "One-Person Company" (autonomous AI-native company) layer2_organization. It is built as a workflow over the existing orchestrator (Orchestrator.execute_agent → run_agent) — not a new service (the one-core rule).

Three concept-anchored capabilities, all in agent_utilities/orchestration/org_runtime.py:

Capability	Concept	What it does
Recruiter / org-synthesis	AU-ORCH.org.recruiter	From a goal, draft an org chart (departments → roles) and fill each role — reuse an experienced :Employee if one exists, else hire a fresh template. Instantiates/reuses :AgentRole/:Employee KG nodes.
Native WorkItem DAG	AU-ORCH.org.work-item-dag	Derive immutable OrgPlanItem definitions, submit their dependency graph as native :WorkItem records, and run only under engine-issued claims, renewable leases, fencing, and terminal commits. ManagerMode is turn-local context, not another lifecycle.
Self-Grown experience	AU-AHE.org.role-experience	Each item's outcome is written back through the AHE reward loop (FeedbackService.record_action_outcome with a role_experience:<role> action id), growing the :Employee's experienceProfile/experienceScore — which the next recruiter run reads back.

Ontology additions (ontology_company.ttl)

:WorkItem remains the shared BFO process class for the platform's one native work-state authority. The org ontology adds :staffsRole and the :Employee datatype properties :experienceProfile, :experienceScore, and :seniority. It does not define organization-specific WorkItem phases or transition fields.

Sole native WorkItem lifecycle

stateDiagram-v2
    [*] --> submitted
    submitted --> ready: dependencies released
    ready --> leased: ClaimWorkItem
    leased --> leased: RenewWorkItemLease
    leased --> succeeded: fenced commit
    leased --> failed: fenced commit
    submitted --> cancelled: native cancel
    ready --> cancelled: native cancel
    leased --> dead_letter: attempts exhausted

Review and rework happen while the same claimant owns the renewable lease. They do not add states or properties to the durable WorkItem. A human escalation callback returns a decision to the live claimant; the claimant alone commits success/failure. An unclaimable dependency deadlock is cancelled through the native WorkItem authority.

End-to-end flow

flowchart TD
    Goal([goal]) --> Recruiter[Recruiter.synthesize_org]
    Recruiter -->|reuse experienced or hire fresh| Chart[OrgChart: roles + employees]
    Chart --> Derive[OrgRuntime.derive_plan]
    Derive --> Plan[immutable OrgPlanItem DAG]
    Plan --> Submit[submit native WorkItems]
    Submit --> Claim[ClaimWorkItem + renewable lease]
    Claim -->|independent, parallel| Exec[Orchestrator.execute_agent → run_agent]
    Exec --> Review{reviewer role?}
    Review -->|approve| Done[fenced success commit]
    Review -->|rework within budget| Exec
    Review -->|budget exhausted| Human[escalation_cb → human]
    Human --> Done
    Done --> WB["record_action_outcome<br/>role_experience:role"]
    WB --> Profile[("Employee.experienceProfile<br/>+ experienceScore")]
    Profile -.reads back.-> Recruiter

The escalation seam (OrgRuntime.escalation_cb) never writes WorkItem state and defaults to failing closed. A deployment may supply a callback that opens an approval through orchestration/action_policy.py; the native lease holder then commits the decision through CommitWorkItemResult.

Surfaces (MCP + REST, in lockstep)

Both actions dispatch through the focused graph_agents core:

• graph_agents(action='synthesize_org', task=<goal>) — REST twin POST /api/graph/agents.
• graph_agents(action='run_org', task=<goal>) — the same REST twin.

Optional options_json carries {"domains": [...]}.

Self-Grown reward write-back

record_role_experience is invoked by the role_experience:<role_id> branch of FeedbackService.record_action_outcome — the same reward substrate that trains routing (model_route:) and the autonomy ramp (trust:). It accrues successes / partials / failures + per-domain counts into a JSON experienceProfile, recomputes a scalar experienceScore (successes + 0.5·partials − 0.25·failures + 0.5·domain-breadth), and re-bands seniority. The recruiter reads experienceScore when it decides reuse vs hire, closing the loop.