The detrimental cost of scaling too fast

McKinsey's 2026 global survey found that 44% of organizations are scaling AI across their enterprises, up from 38% a year earlier. Adoption is higher among large companies, with 54% of businesses generating at least $1 billion in annual revenue reporting enterprise-wide AI scaling. The data signals continued acceleration in corporate AI deployment, though the article excerpt does not provide company-specific financial implications.

Analysis

Enterprise AI adoption is shifting the value pool from model availability toward workflow integration, proprietary data access, and measurable labor substitution. The highest-quality beneficiaries should be software vendors with embedded distribution and mission-critical data—MSFT, NOW, CRM, ORCL and PLTR—rather than smaller firms selling undifferentiated AI features. Near-term revenue recognition will lag customer deployment: copilots and usage-based products can lift bookings immediately, but material margin expansion depends on inference costs falling faster than pricing and support expense over the next 12-18 months.

The underappreciated second-order effect is a widening divide between software companies that can monetize AI as a paid seat, transaction, or consumption uplift and those forced to bundle it defensively. This favors NOW and ORCL, where AI can increase workflow volume and infrastructure consumption, while pressuring application vendors with weak switching costs and limited proprietary data. IT-services providers including ACN, EPAM and GLOB face mixed implications: implementation demand supports the next 1-3 quarters, but successful automation ultimately reduces billable labor intensity and could compress longer-duration revenue multiples.

Consensus remains too focused on AI infrastructure capex beneficiaries. If enterprise pilots convert into production deployments, the next leg of the trade should rotate toward application-layer monetizers; if conversion remains weak, elevated AI-related software multiples are vulnerable to guidance resets. Falsify the constructive view if major vendors show rising AI attach rates without incremental remaining-performance obligations, consumption revenue, or operating-margin leverage by the next two earnings cycles.

Key Decisions for Investors

• Build a 6-12 month basket long NOW, ORCL and MSFT versus short IGV: favor vendors with direct enterprise distribution and clearer monetization paths over broad, high-multiple software exposure. Reassess if relative performance fails to improve after two quarterly reporting cycles or if AI-related bookings do not translate into raised FY revenue guidance.
• Use a pair trade long NOW / short ACN over 3-6 months: NOW captures recurring workflow monetization while ACN is more exposed to eventual reductions in labor-based implementation revenue. Target roughly 15-20% relative upside; stop if consulting bookings accelerate materially or NOW's subscription-growth outlook decelerates below the broader large-cap software group.
• Do not add incremental exposure to AI infrastructure leaders solely on enterprise-adoption survey data. Instead, set an earnings watch for ORCL cloud consumption, MSFT commercial remaining-performance obligations, and NOW net-new ACV; verified acceleration in those metrics would justify rotating from infrastructure proxies into application software.