Judgment stays human. Routing stays governed.

When AI judgment is not enough

Some moments require human judgment — medical edge cases, legal interpretation, exceptions above a risk threshold, or customer situations that models should not own alone. The protocol recognizes those moments and routes work to people who are qualified to decide — with full context and a durable record.

What Workforce Cloud provides

• →Verified participants. Capability attestations — evidence-backed, not title theater. The router knows who is qualified before the task lands.
• →Intake-to-outcome tracking. Full provenance on every escalation — no black-box handoffs, no context loss between the AI step and the human step.
• →Escalation built in, not bolted on. Routing logic lives inside the workflow manifest — not in a separate ticketing system you have to keep synchronized.
• →Accountability that travels. The human who decides is named, their reasoning is recorded, and the outcome is challengeable. Same provenance standards as the rest of HUMΛN.
• →Review without breaking momentum. Escalations pause the specific gate, not the whole workflow — parallel branches keep running.

How escalation works

1. A workflow reaches an escalation trigger (policy, confidence, or explicit gate).
2. Work routes to a qualified reviewer with the same context the AI saw — no opaque handoff.
3. The human reviews, decides, and records the outcome with cryptographic provenance.
4. The workflow continues — the run stays coherent, inspectable, and contestable.

Operational surfaces

Escalated tasks do not disappear — they land in governed queues where assigned people review with full context, decide, and record outcomes. Same provenance and challenge rights as the rest of HUMΛN.