AI helps firms with efficiency, but most don’t trust it to drive growth

EY-Parthenon surveyed 271 growth leaders and found 63% say AI is improving efficiency and productivity, but only 14% use it to stay ahead of competitors, 8% to reach new customers, and 7% to diversify revenue streams. While 78% expect AI to accelerate growth, just 34% trust it for growth-related decision-making, highlighting a trust gap amid more challenging business conditions cited by 80% of respondents. The report frames AI as a strategic growth catalyst, not just a productivity tool, but the article is primarily survey-based and unlikely to move markets materially.

Analysis

The market is underpricing the distinction between AI as a cost tool and AI as a growth engine. That gap matters because the first use case drives margin expansion, while the second can change revenue mix, customer acquisition cost, and product velocity; the equity re-rating is much larger when AI starts showing up in top-line KPIs rather than just operating expense. The likely winners are the software, data, and infrastructure vendors that sit closest to decision workflows, because they can monetize budget shifts from experimentation into embedded production use.

The biggest second-order effect is competitive compression. Once a few firms use AI to shorten product cycles, personalize offers, and redeploy pricing faster, laggards won’t just lose efficiency — they’ll lose pricing power and share in high-margin segments first. Legacy ERP/CRM vendors are vulnerable if customers start separating “decision intelligence” from transaction systems, because that disintermediation can reduce switching costs and push value toward standalone AI orchestration layers.

The key risk is a trust bottleneck that can persist for quarters, not weeks. Governance, auditability, and compliance are likely to remain the gating factors for adoption in regulated verticals, which means the near-term beneficiaries are more likely to be companies selling control, observability, and workflow integration than pure model providers. If macro volatility eases and growth confidence rebounds, AI-for-growth adoption could accelerate abruptly; conversely, a recessionary budget freeze would keep AI confined to productivity use cases and delay revenue impact.

Contrarian takeaway: consensus is probably too focused on frontier model quality and not enough on the plumbing required to make AI decision-safe. The better risk/reward may be in picks-and-shovels names that enable auditable deployment, not in the names selling generic “AI transformation” narratives. The market may also be overestimating how quickly incumbents can convert AI pilots into durable revenue gains; the transition from pilot to scaled rollout is where most projects will stall.

Key Decisions for Investors

• Long MSFT / short legacy horizontal software basket over 3-6 months: Microsoft benefits from embedded workflow distribution and governance tooling; short names with high exposure to generic copilots but weak differentiation. Risk/reward favors MSFT on recurring monetization versus multiple compression in slower movers.
• Long NOW on any post-print weakness, 2-4 month horizon: workflow control is one of the few areas where AI can be tied directly to measurable business outcomes. Upside comes from seat expansion and higher ACV; downside is limited if execution remains steady.
• Long PLTR as a high-beta beneficiary of AI decision intelligence over 6-12 months: if enterprises move from productivity pilots to governed decision systems, demand should shift toward data integration and audit layers. Use call spreads to cap premium if trust adoption remains slow.
• Short INTU or pair short legacy workflow incumbents against AI-native automation names if management commentary suggests slower AI monetization over the next 1-2 quarters: the risk is not immediate revenue loss, but margin pressure as customers demand AI features without paying proportional uplift.
• Buy 6-9 month call spreads on SNOW or DDOG: both can benefit if enterprises need observability and data plumbing to make AI auditable. The trade works best if adoption broadens beyond experimentation; risk is that projects stay contained and spend remains lumpy.