What happened
EY's 21-page analysis (published circa 22 September 2026) argues that AI adoption is advancing more quickly than the operating models required to scale it, trapping many enterprises in a pilot-to-production gap it calls the 'agentic scale trap'. The report attributes failure less to model capability than to missing operating-model elements: process-level value capture, fit-for-purpose KPIs tied to business outcomes, enterprise-wide governance decision rights, AI-ready data lineage and staged autonomy with human-in-the-loop controls. EY recommends treating pilots as staged investments that demonstrate ROI before scale, centralising identity, policy and compliance charters, and gating scale on repeatable data-readiness assessments.
Why it matters
Gives transformation and technology leads a concrete diagnosis for why agentic AI pilots stall plus a five-lever framework for scaling to production with governance built in.
Action needed
Audit your AI pilot portfolio against EY's five levers — process-level value, KPI definition, governance, data readiness and staged autonomy — before committing further scale capital.