Human Oversight Is 70-75% of AI Agent Costs in Banking, McKinsey Finds
mikeflache · x · 2026-11-04
Rich Turrin highlights McKinsey findings: when an AI agent performs a banking task, token costs are just 20-25% of variable costs while human oversight accounts for 70-75% — the human in the loop is the bottleneck, not a safeguard.
- BNY treats agents like employees: login IDs, employee numbers, directory avatars, Teams/email access, cheeky names like "Payment Pete," and even performance reviews overseen by 100 human managers.
- McKinsey recommends unique agent identities, named human owners, formal registries, risk-tiered controls, audit-ready traceability, and governed memory.
- Design principle: assume agent error — scope tools and permissions, validate outputs pre-launch, and contain failure with rollback paths and blast-radius limits. An accompanying chart breaks down human cost share across in-loop (72%), on-loop (38%) and over-loop (14%) oversight tiers.
More from coding & agent
- Quota reset night: Codex back to 100% while Claude shows just 10% left — rudrank · 2026-11-04
- Self-improving software is inevitable: closing the agent-observability loop — bibryam · 2026-11-04
- Building a 7-GPU 132GB VRAM local inference rig: consolidate on EPYC or split across two systems? — infectiousstupidity · 2026-11-04
- Founder of 50-person, 12-country team says he talks to AI agents more than humans — Scobleizer · 2026-11-04
- Pi launches a minimal agent harness; Zammoth adds AI engineer to Mac music app — solyarisoftware · 2026-11-04
- Reply to Karpathy goes viral again: devs who refuse AI coding will be replaced first — CtrlAltDwayne · 2026-11-04