The artificial intelligence (AI) hype cycle has had no shortage of projections, pronouncements and technical forecasts.
But while fears about AI replacing jobs may be overblown, per a recent Wednesday (June 18) report, enterprise adoption of AI has been accelerating in both scale and seriousness over the past year and a half. With purpose-built enterprise models on the rise, AI spending is increasingly no longer a side experiment; a growing number of businesses are treating the technology as a long-term line item, just like cloud infrastructure or cybersecurity.
Companies ranging from financial services to health care, retail and even travel, are no longer asking if AI belongs in their tech stack — they are deciding how and where it fits best.
That “how” varies widely. Enterprises are pursuing multi-model strategies, deploying different AI models for different tasks depending on performance, data sensitivity and regulatory needs. A single provider may no longer be sufficient for the full range of use cases. Companies want flexibility: the ability to mix hosted and on-premises models, the option to tailor tools for specific departments and the assurance that data policies are respected across environments.
In parallel, AI procurement is becoming more sophisticated. Selection criteria have evolved. Capabilities still matter, but so do integration ease, compliance support, vendor transparency and overall operating cost. For many companies, the goal is not maximum output at all times but dependable performance within operational and legal boundaries.
As PYMNTS Intelligence data from The CAIO Report reveals, for today’s CFOs, AI is evolving from a buzzy experiment to a key element of back-office infrastructure.
See more: 3 Ways AI Shifts Accounts Receivable From Lagging to Leading Indicator
AI Budgets Go up as Enterprise Expectations Continue Growing
In the past, enterprise AI initiatives often floundered under the weight of vague objectives, data silos and a lack of skilled talent. The technology was promising but fragmented. Models were generic, one-size-fits-all solutions that struggled to adapt to complex, domain-specific needs. Enterprises were left juggling disparate tools with no unified strategy.
Fast forward to today, and the application of AI technologies, particularly those described as “generative” or “agentic,” is neither following the trajectory of unchecked enthusiasm nor stalling under the weight of its own complexity. Rather, it is unfolding incrementally and through the calibration of capability against operational necessity.
In high-risk areas like finance, procurement or compliance, few companies are ready to let AI systems — particularly agentic AI solutions — take action without layers of review. Even in customer support, where automation is more common, AI tools often work alongside human agents, not instead of them. The gap between what technology can do in theory and what companies are comfortable deploying at scale is still significant.
Still, the idea of “agentic AI,” or systems that can act with minimal human guidance, has started to enter corporate conversations. But it remains more of a future ambition than a present reality. Most AI applications ultimately remain tightly scoped and heavily supervised.
In order to help close some of the governance gaps that may be keeping incumbent organizations from fully embracing AI, IBM on Wednesday launched an end-to-end AI governance tool, watsonx.governance, and a new tool for securing AI models, data and usage, Guardium AI Security.
Read also: CFOs Move From Ledgers to Leaders as Back Offices Become Command Centers
Why AI Is Becoming an Infrastructure Line Item
As enterprises increase their use of GenAI, concerns about pouring dollars into the technology diminish, but new risks emerge. According to PYMNTS Intelligence research, none of the high-automation firms surveyed cited return on investment as a concern. By contrast, half of low-automation firms still worry about whether GenAI is worth the investment. Automation, it seems, validates itself financially only over time.
“AI offers a new path forward with its capability to aggregate and structure internal knowledge across silos, without the need for manual data entry. But it’s not as simple as prompting an off-the-shelf LLM (large language model). These models are powerful, but there’s a need for purpose-built software that understands each fund’s unique workflows and transaction,” Taylor Lowe, CEO and co-founder of Metal, told PYMNTS in an interview.
AI is now becoming a fixture in enterprise planning. But as adoption moves deeper into core functions, the questions companies are asking have shifted. They are less about innovation for its own sake and more about how to use these tools responsibly, repeatably and at scale.
