From AI FOMO to AI hangover: corporate America is taking a long, hard look in the mirror right now

Companies are projected to spend more than $2.5 trillion on AI in 2026, up 47% from 2025, but the article argues that FOMO-driven deployments have produced limited visible business impact, employee pushback and rising token costs. Only about 5% of GenAI-enabled employees—typically top performers—are said to use the technology in ways that materially improve speed or quality. The author argues companies should shift from broad productivity-focused rollouts toward human-led AI use, targeted training and slower implementation to improve ROI and mitigate skill erosion.

Analysis

The investable issue is not AI adoption breadth but whether enterprises convert pilot-seat spending into governed workflows with measurable P&L ownership. If usage shifts from indiscriminate drafting toward approved, role-specific processes, platform vendors embedded in systems of record—ServiceNow (NOW), Salesforce (CRM), Microsoft (MSFT)—should retain budget while standalone “AI feature” vendors face seat rationalization and longer sales cycles. The second-order beneficiary is security/data-governance spend: tighter human-review and data-access controls raise the attach rate for CrowdStrike (CRWD), Palo Alto Networks (PANW), Okta (OKTA), and Snowflake (SNOW).

Near-term, the risk is a CFO-driven pause in incremental licenses and model consumption as companies demand proof of savings; this would matter more for high-multiple software than for hyperscalers with diversified revenue bases. Over 1-3 months, earnings commentary on AI monetization, net retention, and deferred revenue should separate workflow vendors from tools whose engagement is not translating to expansion. Over 6-18 months, AI’s likely value accrues to firms that redesign processes and reduce error/rework, not necessarily to the vendors with the largest raw token volumes; this argues against extrapolating infrastructure demand directly from announced enterprise budgets.

The contrarian point is that disappointing first-wave ROI need not mean AI spend collapses—it can mean spending migrates from generic copilots toward implementation, governance, and vertical workflows. The article's productivity and quality claims are directional commentary rather than independently auditable evidence, so this is not sufficient basis for a broad short; confirmation requires weaker software seat growth alongside rising demand for workflow automation and security controls.

Key Decisions for Investors

• Initiate a 3-6 month pair trade: long NOW / short IGV, sized market-neutral. Thesis: budget consolidation favors governed workflow automation over undifferentiated application-software seats. Falsify if NOW cRPO and subscription growth decelerate materially while IGV constituents show broad reacceleration in net retention.
• Accumulate PANW and CRWD on software-sector weakness over the next 1-3 months, targeting a 10-15% position-level upside versus a 7-10% stop. Human-review requirements and restricted AI data flows increase demand for identity, endpoint, and data-security controls even if discretionary copilot licenses are cut.
• Avoid adding to high-multiple AI application names with weak evidence of paid-production deployment, particularly C3.ai (AI), until quarterly disclosures show durable consumption, renewal expansion, and customer-level ROI. A broad enterprise spending narrative is not a substitute for monetization evidence.
• Monitor MSFT Azure AI consumption and ORCL cloud backlog conversion at the next earnings cycle. If management commentary indicates enterprise token optimization or deferred deployments rather than production scaling, reduce AI-infrastructure beta through SMH/IGV hedges; sustained consumption growth would invalidate the near-term digestion thesis.