Getting past the pilot: Why so many AI test projects have trouble scaling

The article argues that many AI pilots fail during scale-up because companies lack tight governance, clear business outcomes, documented workflows, and the right data access and stakeholder buy-in. Executives from Amgen, Salesforce, and Thomson Reuters emphasize that successful AI deployment depends more on planning, privacy/security review, and cross-functional alignment than on the technology itself. The piece is advisory in nature and contains no company-specific financial results or quantified market-moving developments.

Analysis

The market implication is less about near-term AI hype and more about a widening gap between AI experimentation and enterprise-grade deployment. That tends to favor vendors that sit in the governance, workflow, identity, and data-access layers rather than pure model exposure: the monetization bottleneck is moving upstream into security, compliance, orchestration, and integration spend. In practice, this is a multi-quarter budget reallocation story, because enterprises usually discover these frictions only after pilots stall, meaning the incremental revenue opportunity for platform vendors should show up with a lag but be stickier once it arrives.

For CRM, the core read-through is not simply “AI demand is strong,” but that customers are increasingly forced to buy more of the workflow stack before they can scale agents. That should improve attach rates for data, automation, and admin tooling, but it also raises the bar for execution: if AI features are perceived as cosmetic, customer tolerance for premium pricing compresses. The second-order loser is point-solution AI vendors that sell above the workflow without owning identity, permissions, or system-of-record integration; their pilots may look good, but conversion to enterprise deployment is where churn and discounting emerge.

For AMGN and TRI, the article is a reminder that regulated enterprises can be early beneficiaries of disciplined AI adoption, but only if they turn governance into a moat. TRI is structurally better positioned than most information-services peers because its products can become the control plane for high-stakes knowledge workflows; that said, any upside likely accrues over 6-18 months rather than immediately. The contrarian take is that this is bearish for companies chasing headline AI launches without process mapping: the failure rate will stay high, and the market may eventually punish “AI theater” more than it rewards feature velocity.

Key Decisions for Investors

• Long CRM vs short a basket of high-beta standalone AI application names over the next 3-6 months: the trade favors companies that can monetize governance and workflow integration rather than those dependent on pilot-to-production conversion. Target 1.5-2.0x upside on CRM relative outperformance if enterprise spend shifts from experimentation to platform consolidation.
• Add TRI on pullbacks for a 6-12 month horizon: the market is underestimating the value of data lineage, permissions, and trusted content controls as AI moves into regulated workflows. Risk/reward is attractive if the company can monetize control-plane functionality without a major multiple re-rating elsewhere.
• Use AMGN as a barometer for enterprise AI discipline rather than an AI beta trade: only add if evidence emerges that AI is tied to measurable operating leverage, otherwise treat any enthusiasm as transient. Best framed as a tactical hold, not a high-conviction growth re-rate.
• Short the most speculative AI software names that depend on rapid horizontal adoption and lack system-of-record ownership, especially into earnings, where management commentary on stalled rollouts can reprice multiples quickly. Time horizon: 1-2 quarters; asymmetry is strongest where valuation already discounts perfect conversion rates.
• If you want options exposure, buy 6-12 month call spreads on CRM rather than outright calls: the theme is real but gradual, and call spreads capture moderate upside if the market starts valuing workflow depth and AI attach without overpaying for a full secular rerating.