CEOs warn against falling into AI’s efficiency trap: The biggest risk is ‘a better version of yesterday’

Fortune’s leadership forum highlighted that AI adoption is widespread—ServiceNow cited roughly 60% of companies having AI agents—but only 5% have redesigned workflows around the technology. Executives from Walmart China, AS Watson, Syngenta, HP, and others argued that AI should drive enterprise reinvention while increasing the value of human judgment and customer insight. Walmart China reported 20.7% quarterly growth, versus 2.6% for Walmart U.S., attributing the outperformance to data, AI, and human decision-making.

Analysis

The investable distinction is not AI adoption but workflow ownership. NOW is best positioned among the named companies because it sits at the orchestration layer where enterprises must connect data, approvals, service processes, and agents; successful redesign raises switching costs and supports both net retention and a higher platform multiple. The risk is that agent deployments remain departmental experiments, producing seat rationalization before enough incremental workflow spend offsets it; watch NOW’s cRPO growth and GenAI-related deal sizes over the next two earnings cycles.

WMT’s China growth should not be extrapolated into a broad AI-margin thesis: merchandising tools can improve assortment and inventory turns, but local competitive intensity means much of the benefit may be passed through via price and fulfillment. The more durable read-through is to Walmart’s retail-media and marketplace economics, where better product matching can raise high-margin advertising yield without equivalent price concessions. Over 6-18 months, AI-enabled demand forecasting also improves supplier bargaining power, pressuring consumer-product vendors with concentrated Walmart exposure more than WMT itself.

HP lacks a clear earnings mechanism from enterprise AI discussion alone. A PC refresh can lift unit demand, but AI-capable hardware risks becoming a lower-margin specification upgrade unless HP captures attach rates in services, peripherals, or commercial fleet management; this is not yet a reason to underwrite multiple expansion. Consensus may be over-rewarding visible AI infrastructure beneficiaries while underestimating that application-layer winners need organizational change, long sales cycles, and clean enterprise data before revenue materializes.

Near-term market sensitivity is likely dominated by yields rather than AI narratives: long-duration software multiples, including NOW, remain vulnerable if real rates rise further. A reversal would come from evidence that agent projects convert into broad production deployments, demonstrated by accelerating subscription commitments rather than management commentary.

Key Decisions for Investors

• Maintain/enter a 3-6 month long NOW versus short IGV hedge: favor NOW’s workflow-control position over the diversified software basket. Target relative upside of 10-15%; reduce if cRPO decelerates below management’s medium-term growth framework or valuation expands without corresponding booking acceleration.
• Own WMT on a 6-18 month horizon, but frame the thesis around retail media, marketplace mix, and inventory productivity rather than a generic AI premium. Add on rate-driven weakness; invalidate if China growth slows sharply while consolidated gross margin and advertising disclosure fail to improve.
• Avoid adding HPQ solely for the AI-PC cycle. Upgrade to a long only if commercial PC demand converts into sustained ASP growth and management shows services/attach-rate expansion; otherwise the likely outcome is volume recovery with limited EPS-multiple upside.
• Use any sharp NOW rally following AI-product announcements to sell upside tactically unless accompanied by disclosed production customer counts, contract duration, and incremental ACV. The missing proof point is monetization, not product availability.