The gulf between enterprise AI spending and adoption is expanding. At the same time, an emerging trend is driving AI pricing. It’s one we’ve seen play out before with chips and storage: capacity is up, costs are going down.

With cheaper AI, there’s a lower barrier for entry, yet data shows it isn't translating into widespread or meaningful use of AI across the enterprise. To uncover the real bottlenecks, I looked into what business leaders are experiencing on the ground. It revealed a two-part requirement in AI infrastructure: technology flexible enough to adapt as the market inevitably changes, and the internal readiness to actually put that technology to work.

Abundance can be a disadvantage

Enterprise technology leaders now face a labyrinth of high-stakes decisions around models, tooling, architecture, and infrastructure.

While long-term commitments to providers like OpenAI or Anthropic may promise stability, enterprise AI increasingly depends on balancing flexibility with reliability to scale adoption. Every choice—from selecting foundation models and orchestration tools to deciding between open and proprietary architecture has implications that extend far beyond today's implementation. These decisions will determine how quickly organizations can scale AI, how effectively they govern it, and whether they'll be positioned to adapt.

I reviewed recent research from global ERP vendor Infor to better understand the factors influencing—and limiting—global AI adoption. Infor’s research found 80% of enterprise decision-makers believe their organizations have the internal capability to manage AI implementation, but nearly half (49%) remain stuck in the earliest stages of adoption.

That confidence gap shows up at the individual level too: a recent survey from my company, Prosper Insights & Analytics , found 53% of executives and business owners say they already use generative AI, compared with just 38% of employees, suggesting the disconnect between belief and execution runs through the org chart, not just the boardroom.

These findings show that AI implementation is as much an organizational challenge as a technical one, spanning infrastructure, governance, security, data quality, and change management.

Security continues to sit at the center of that equation.

Among Infor’s survey respondents, 36% identified data security, sovereignty, privacy, or regulatory compliance as the single greatest barrier preventing broader AI adoption. Another 25% cited insufficient internal expertise to configure and maintain AI systems, while 23% pointed to uncertainty around business value or return on investment.

I spoke with Infor’s SVP of AI Innovation Rick Rider , who argued that these concerns are interconnected.

“Organizations cannot confidently calculate ROI if they cannot confidently deploy AI into sensitive business processes. Likewise, they cannot fully automate workflows without trusting both their underlying data and the systems acting upon it,” said Rider. “Data readiness remains one of AI's most overlooked challenges.”

Infor’s survey echoes Rider’s take: More than one-quarter of respondents said they were unsure—or disagreed—that their organization's data is mature and well-governed enough to support reliable AI. That's significant because even the most sophisticated foundation models depend on high-quality enterprise data to generate trustworthy outcomes, and nearly half of AI-generated outputs still require manual review before organizations trust them in regulated or mission-critical environments.

Trust is as important as innovation

Trust becomes even more important as AI moves beyond generating information toward taking action.

Autonomous AI agents represent one of the industry's most promising frontiers, but enterprise leaders remain cautious. Nearly one-third of Infor’s survey respondents said they were uncomfortable allowing autonomous agents to execute critical business processes. That caution isn't confined to the C-suite – Prosper Insights & Analytics research shows only 18% of the general US population believes agentic AI is a "good idea," and even among executives and business owners, who are notably more bullish, just 34% agree, versus 19% of employees.

According to Rider, that hesitation is understandable.

“Enterprises are doing more than testing model performance; they're evaluating accountability. Leaders need confidence that AI systems will behave consistently, comply with regulations, protect sensitive information, and operate within clearly defined guardrails,” said Rider. It tracks with broader sentiment: 40% of Americans cite AI's tendency toward hallucination and the need for human oversight as top concerns, nearly identical to what executives report.

The need for trust is especially critical in highly regulated industries like financial services, according to Sigma360 founder Stuart Jones, Jr. Sigma360 is an AI platform for risk intelligence, financial crime prevention and compliance, protecting over $2 trillion in assets for banks, fintechs, and payment providers globally.

"There are a lot of time-intensive processes that institutions would love to use AI to free up analyst bandwidth and support the kind of real-time transactions today's consumers are expecting," said Jones. "But they need to know governance standards are being met. Data integrity and stewardship are the key."

Cost predictability is becoming equally important.

While AI pricing continues to evolve rapidly, enterprise budgeting still depends on long-term planning. An overwhelming 87% of respondents from Infor’s research said fixed and predictable AI pricing is important when making long-term investment decisions.

As model capabilities improve and competition intensifies, pricing will likely continue becoming more competitive. But enterprises need transparency as much as they need lower costs. Predictable pricing enables organizations to expand AI across departments instead of limiting deployments to isolated pilot projects.

Ultimately, the winners in enterprise AI will build the most adaptable foundations and use the best model at any given moment.

History offers a useful guide

Every major enterprise technology shift—from client-server computing to virtualization, cloud infrastructure, and modern data platforms—rewarded organizations that invested in flexibility rather than rigid architectures.

AI raises the stakes because innovation is outpacing traditional planning cycles. Frontier models, open-weight alternatives, and new orchestration platforms are expanding enterprise options almost monthly.

In this environment, interoperability is no longer a technical preference; it's a strategic advantage. "When we built our AI strategy, we came from the perspective of, what does an analyst need to move faster without losing accuracy? In the financial services industry, everyone is getting a piece of their risk assessment from different sources, so not only is it inefficient, it's siloed," said Sigma360's Jones. "Interoperability — and the ability to use different models to solve different problems — is key. We're able to bring together different sources of data, so the whole is greater than the sum of its parts."

As Infor’s Rider puts it, “Organizations that win will avoid unnecessary lock-in, prioritize open ecosystems where appropriate, and build governance into every layer of their AI strategy.”

When it comes to enterprise AI, technology decisions will increasingly be evaluated on scalability, adaptability, implementation speed, and long-term business value. There’s precedence here. We’ve seen this play out in chips, storage, and cloud.

But AI is different. The technology is advancing at paces much faster than Moore’s Law . Competition between hyperscalers like Anthropic and OpenAi, and big tech companies pivoting their business models around AI, like Google or Microsoft, is heating up. Emerging threats from open weight and open-source models that provide frontier-grade outputs at zero cost are putting pressure on the entire ecosystem.

Enterprise technology leaders are at the center of it all. Success will depend on moving beyond pilots and building AI strategies around flexibility, reliability, and governance.

Disclosure: The consumer sentiment study referenced above was conducted by my company, Prosper Insights & Analytics . This is the same dataset used by the National Retail Federation, and available from Amazon Web Services, Databricks, and the London Stock Exchange Group for economic benchmarking.