Agentic Team Archetypes

McKinsey's three patterns (GCC Philippines Summit 2026 (PHx)) for how future GCC teams combine humans and AI agents — all managed by a human orchestration layer:

1. Human-led, enabled by agents — human is the doer; agents assist. E.g. preparing/executing a structured-finance deal for a corporate client (the end client wants a human accountable).
2. Agent-led, human-in-the-loop — agents do the work; a human stays accountable and handles edge/legal cases. E.g. credit decisioning for standard corporate loans, underwriting, QC memo writing.
3. Fully agentic — agents run end-to-end. E.g. call-center handling of standard requests, standard retail credit decisioning.

Where this sits

This is the operating-model layer beneath the Frontier GCC and the Headcount-to-Value Pivot: as L1/transactional work becomes "fully agentified," humans concentrate in archetypes 1–2 as orchestrators, closers, and fixers. The accountability spectrum is the GCC-scale instance of Human in the Loop; the per-agent autonomy test is Bounded vs Unbounded Tasks.