Explicit-request-overrides rule

Definition

When a user explicitly asks for a human, the agent escalates immediately — no investigation, no attempt to solve first — even if the request is trivially solvable.

Key points

• Explicit request for a human → escalate immediately, no investigation/attempt first. This overrides everything.
• Do not ask for further clarification and do not try to resolve it "to save the human's time."
• In the Escalation Handoff decision tree this is the "I want a human NOW" branch → Immediate Escalation, honored instantly via escalate_to_human (no account-context gathering first).
• Contrast with the Complex Policy Issue branch, where the agent first calls account-context tools (e.g. get_customer) before escalating.

Why it matters for the exam

• A favorite scenario: user says "give me a human" for a problem the agent could clearly fix — the correct answer is escalate now, not resolve.

Common gotchas

• "It's a one-click fix, so solve it anyway" is the wrong answer — the explicit request wins.
• Only the explicit request triggers immediate escalation; mere frustration does not (see Wrong escalation signals).

See also

• Escalation triggers
• The Escalation Handoff
• Wrong escalation signals
• Structured handoff summary

Sources

• netlight-study-guide
• architects-playbook