EY Global Vice Chair: AI’s biggest paradox comes down to what AI can’t do

Companies may forfeit up to 40% of potential AI productivity gains because of insufficient talent investment, according to the EY commentary. The article argues that AI fluency, human judgment, and integrated human-AI workflows are essential to scaling adoption; Lion reportedly achieved 75% faster customer response and about 30% lower people-function operating costs through ecosystem collaboration. Daikin’s hybrid AI-human ERP pilot accelerated delivery by roughly 30%, improved counter efficiency 10%, and shortened financial close by 20%.

Analysis

This is not a direct earnings catalyst; it is a directional signal that enterprise AI value will accrue less to foundation-model providers alone and more to vendors embedded in governed workflow redesign. The bottleneck shifts from inference availability to implementation, controls, data access and change management—supporting durable services attach rates for ACN, IBM, GLOB and large platform ecosystems such as MSFT, NOW and CRM. The cited productivity outcomes are vendor-case-study claims rather than independently auditable evidence, so they should not be extrapolated into near-term revenue estimates.

Over the next 1-3 months, investor attention is likely to remain concentrated on AI capex and model releases, leaving governance and workflow-integration beneficiaries relatively underappreciated. In 6-18 months, disappointing ROI from stand-alone copilots should drive buyers toward integrated deployments, increasing demand for ServiceNow workflow automation, Microsoft’s enterprise stack and cybersecurity/identity controls from PANW, CRWD, OKTA and CYBR. The second-order beneficiary is ERP modernization: SAP and ORCL can monetize AI only where customers consolidate process and data layers, while fragmented legacy estates raise implementation friction and favor consultants.

The contrarian risk is that "human-in-the-loop" becomes a cost center rather than a monetizable feature if regulation remains light and enterprises accept lower assurance for low-stakes workflows. A broad AI spending slowdown, weak consulting bookings, or evidence that customers deploy internally with hyperscaler tooling would challenge the services thesis. Watch ACN bookings and utilization, NOW subscription cRPO, SAP cloud backlog, and enterprise AI project conversion rates through the next two reporting cycles.

Key Decisions for Investors

• No immediate directional trade from this commentary alone; treat as a watch signal for Q4/Q1 enterprise-software and consulting results rather than a standalone catalyst.
• Build a 6-12 month basket long NOW and SAP versus a short legacy IT-services basket proxy (IGV is not appropriate as a pure short); prefer selective underweights in labor-arbitrage-heavy consultancies if bookings reveal AI-led project cannibalization. Thesis requires accelerating workflow/ERP backlog, not merely AI product announcements.
• Monitor ACN: initiate a tactical long only if next-quarter new bookings and utilization stabilize while management quantifies AI-related implementation revenue. Upside is multiple re-rating from AI displacement fears; invalidation is renewed pricing pressure or falling headcount utilization.
• For defensive AI exposure, favor PANW or CRWD on 6-18 month pullbacks: autonomous-agent deployment expands identity, audit and policy-control requirements. Exit/avoid if enterprise security platform consolidation materially reduces net-new spend or regulated-agent requirements fail to emerge.