Can AI agents truly operate above the loop at enterprise scale?

McKinsey (Sukharevsky, Krivkovich, Gast, Storozhev, Maor, Mahadevan, Hämäläinen, and Durth) argues that AI is driving "the largest organizational paradigm shift since the industrial and digital revolutions," a model it names the "agentic organization" that unites humans and AI agents, "both virtual and physical," working "side by side at scale at near-zero marginal cost." The article argues this paradigm will be "built around five pillars of the enterprise: business model; operating model; governance; workforce, people, and culture; and technology and data." Rather than retrofitting AI onto existing processes, McKinsey frames the shift as "a redesign of end-to-end processes with humans 'above the loop' for strategic oversight," with humans and traditional IT systems "selectively introduced back in the loop or above the loop" only where needed.

The mechanism McKinsey describes: traditional organizations are built around functional silos, while the agentic organization instead organizes around "flat networks of empowered, outcome-aligned agentic teams" — mixed teams of humans and agent "squads" aligned to a shared outcome rather than a function. McKinsey offers client vignettes as evidence: one global bank's "agent factory" manages know-your-customer processes with "ten agent squads," which it says "helped achieve a substantial positive impact on the quality and consistency of output"; another bank used humans to oversee agent squads modernizing legacy core systems, "enabling up to 50 percent reductions in time and effort." The underlying logic is that because agents can act at "near-zero marginal cost," cost decouples from growth, and competitive advantage shifts toward proprietary "walled data gardens" rather than headcount or process scale.

This is a more sweeping claim than the constrained picture in Why do production AI agents stay deliberately simple?, whose survey of real deployments finds agents kept narrow and reliability-first; McKinsey's "AI-first" end-to-end workflow vision describes an aspirational trajectory that current production practice has not caught up to. It also sits in tension with Why are companies deploying agents faster than governance matures?: McKinsey's pillars model names governance as one of five co-equal pillars of the paradigm, but Deloitte's survey finds governance maturity lagging deployment ambition industry-wide, which suggests the "above the loop" oversight role McKinsey describes may not yet be backed by the structures needed to execute it. And where McKinsey's bank example assumes agent squads can be bounded into clean workflows — real estate, mortgage underwriting, compliance, contracting, fulfillment — Can workplace AI risks emerge from interactions alone? complicates that assumption by locating risk in the interactions between agents, goals, and environment rather than in any single squad's reliability.

What this excerpt does not establish: it is McKinsey's own framework, drawn from "McKinsey's experience working with early adopters" and a small set of named client vignettes (a bank, an insurer, telcos, a European automaker, a public sector organization), not a representative survey of outcomes — the "up to 50 percent reductions" and "substantial positive impact" figures are reported client results rather than independently measured or replicated findings. The piece also does not specify what "above the loop" oversight concretely requires in skills, staffing ratios, or governance mechanisms; it names governance as a pillar without detailing it. The implication, held no more strongly than the evidence allows, is that the agentic organization is McKinsey's name and sales case for a trend under way at a handful of early-adopter firms, not yet an established organizational model.