After years of explosive investments and enthusiasm for artificial intelligence, a new AI Index by fintech company Ramp reveals a noticeable plateau in U.S. corporate AI adoption. This finding marks a key turning point in the broader narrative around enterprise AI, where ambitions for sweeping productivity boosts are now colliding with operational and cultural realities.

Adoption Rates Reach a Plateau

Ramp’s proprietary data, gathered from a financial sample of some 30,000 companies, indicates that corporate AI adoption stabilized at 41% in May 2025, pausing a near year-long trend of continuous growth. Notably, large enterprises are in the lead, with 49% deploying AI solutions, compared to 44% of mid-sized firms and 37% of small businesses.

These numbers reflect significant penetration by modern AI—generative, analytical, and automation apps—across a wide swath of the business landscape. But the stall at the 41% mark suggests the low-hanging fruit may be picked, and companies now face more nuanced challenges in implementation, expansion, and value realization.

Chasing Productivity: Mixed Results and Emerging Disillusionment

Much of corporate America’s AI ambition hinged on promises of cost savings, automation, and smarter decision-making. Yet, recent high-profile cases—such as Klarna’s partial reversal on AI-driven support automation after customer service quality dipped—underscore the limits many organizations are encountering. According to S&P Global, 42% of firms have abandoned most of their generative AI pilots, climbing sharply from just 17% a year ago.

This trend reflects the so-called “trough of disillusionment,” where real-world friction—technical, cultural, and ethical—has begun to temper once-unstoppable enthusiasm. Companies are increasingly wary of the challenges around scaling pilots into sustainable, cross-functional solutions, integrating AI into legacy systems, ensuring data quality and security, and retaining a positive customer experience.

Nuances in the Data: What Ramp’s Index Captures—and What It Misses

Ramp’s AI Index leverages anonymized card transaction and bill data to estimate how widely AI products permeate U.S. organizations. While robust in scope, it is important to recognize its limitations. The index may underrepresent AI investments classified under generic project or IT spending, and could miss some segments—like custom internal AI development—less likely to show up in direct vendor payments.

That said, the index reflects a credible cross-section of commercial AI tools, from major providers like OpenAI, Microsoft, Google Cloud, and Anthropic to specialized SaaS vendors powering sales, marketing, and infrastructure operations. The trend it reveals—that U.S. AI spending now stands at a crossroads—is echoed in related sector studies.

Workforce Impacts and the Human Factor

As organizations hasten or pause their AI investments, workforce implications are surfacing as a decisive factor. A 2025 Gartner report indicated that while AI is expected to create more roles than it eliminates over the next decade, the present transition is deeply disruptive. Reskilling, change management, and preserving organizational knowledge have become top priorities for CIOs and HR leads.

The Klarna example is illustrative: the company had moved quickly to downsize hundreds of support roles with an AI chatbot platform, only to rehire a portion of its workforce after customer experience ratings suffered. Meanwhile, employees in functions such as finance, HR, and marketing are increasingly encountering AI-augmented workflows, requiring both new digital skills and adaptation to evolving processes.

Key Drivers of the Plateau

• Integration Costs: The expense and technical complexity of weaving AI into existing systems and processes can outweigh short-term value, particularly for smaller and legacy-intensive firms.
• Regulatory Uncertainty: Heightened global scrutiny and emergent regulations—like the EU AI Act and several U.S. state-level initiatives—have left compliance teams treading cautiously, especially in sensitive industries.
• Quality and ROI Questions: As real-world pilots replace early hype, tangible ROI remains elusive in many deployments. Sectors such as customer support and content generation have seen mixed results, with measurable gains mainly isolated to data-rich, automation-ready scenarios.
• Security and Data Governance Concerns: High-profile breaches and generative AI hallucinations have triggered stricter internal reviews and slowdowns as organizations grapple with safeguarding data, brand, and compliance.

Where AI in the Enterprise Goes Next

Analysts at firms like McKinsey and Gartner say the current pause in adoption is likely temporary. As organizations digest lessons from pilot failures and invest more strategically in governance, integration frameworks, and employee training, the next phase may see a more sustainable, results-driven approach to AI transformation.

Specific areas showing continuing momentum include supply chain optimization, fraud detection, cybersecurity, and advanced analytics. In contrast, non-core deployments—such as creative content and unmoderated customer support—may see prolonged caution until tools mature further and operational kinks are ironed out.

Ultimately, the data suggests that AI in the enterprise is entering a new, more pragmatic chapter—one focused less on experimentation for experimentation’s sake, and more on building measurable, scalable value while managing risk and organizational change.