Organizations with trustworthy AI practices are 15 times more likely to see strong ROI, per study findings

SAS/IDC finds that 97.2% of users override AI recommendations at least some of the time—most often because the AI can’t explain its decisions—while trust drops from 76% for generative AI to 66% for agentic AI. Organizations applying “trustworthy AI” practices report materially better outcomes: 62% vs. 4% for strong/high ROI (15x higher odds) and 1.85x greater gains across business outcomes. Only 17.5% of enterprises have data infrastructure mature enough for agentic AI, implying adoption and ROI remain bottlenecked by governance, data quality, and auditability.

Analysis

This is less a product catalyst than a budget-allocation signal. The incremental spend is likely to migrate toward the layer that makes AI governable — data quality, lineage, identity, audit trails, workflow control — because the first enterprise failure mode is not model quality, it is human override and compliance friction. That favors platform vendors with control points in the stack, especially Microsoft, Oracle, Snowflake, IBM, ServiceNow, and security/identity names, while pure-play "more agents" stories risk longer sales cycles and lower realized monetization.

Second-order, the report argues for a two-speed adoption curve: a small cohort of regulated leaders will scale, while the median enterprise remains stuck in pilots until the data foundation is fixed. That means the market may be overpricing near-term autonomous-agent penetration but underpricing the service-heavy remediation phase. If this view is right, the relative winners over the next 3-12 months are vendors that monetize governance attach rates and implementation complexity, not just model usage.

Contrarian angle: consensus may be treating trust as a feature when it is actually a prerequisite, which makes the short-run economics messier than the AI bulls expect. The near-term risk is slower ROI recognition, not faster; the longer-term upside is a larger, more durable enterprise software spend pool. The thesis would be falsified if upcoming enterprise earnings show AI deployments scaling without incremental governance spend, or if bookings commentary from infrastructure/data vendors does not improve over the next 1-2 quarters.

Key Decisions for Investors

• Long MSFT / short AI (C3.ai) for 6-12 months: express the view that enterprise AI spend migrates from hype-layer applications to governed platform control points. Risk/reward is attractive if governance becomes a standard procurement requirement; thesis breaks if standalone AI app bookings inflect materially faster than platform attach rates.
• Buy SNOW on weakness over the next 1-3 months as a data-foundation beneficiary. Use it as a relative long against a generic AI-beta basket; if the market rerates "trust" as a prerequisite, Snowflake should see better multiple support. Falsifier: no improvement in enterprise AI data-quality commentary by next earnings cycle.
• Long PANW or OKTA calls into the next 3-6 months as a hedge on auditability and access-control spend. The payoff is asymmetric if enterprises start budgeting identity/logging controls alongside agent rollouts. Falsify if AI governance is absorbed into existing IT spend without incremental security attach.
• Pair long IBM / short a high-multiple AI hype basket for 6-12 months. IBM has the services and governance angle to monetize the messy part of AI adoption, while speculative AI names are most exposed to slower trust-driven conversion. Close the short if the market starts rewarding faster autonomous-agent adoption rather than compliance-heavy rollout.