Across organizations deploying autonomous agents, the weak link is often the handoff: when to let agents act, when to pause, and how to bring humans in without killing leverage. This episode presents five minimal 'escalation primitives'—simple, composable rules executives can apply to any agentic workflow to ensure reversibility, accountability, and calibrated human involvement. You get one core insight (escalation as a decision primitive), three supporting points (classification of primitives and use-cases; how to trade speed for safety; lightweight observability to detect misfires), and a single action step: implement one escalation primitive this week and measure its effect. The briefing is lean, operational, and framed for executives who need high-leverage controls, not bureaucracy. Expect practical templates you can drop into design reviews, incident runbooks, and agent contracts so autonomy scales without creating unmanageable tail risk.