A few months into 2026, a stunning contradiction is emerging in enterprise AI spending. Recent findings from S&P Global show that 42% of companies scrapped most of their AI initiatives in 2025 – up sharply from just 17% the year before. Meanwhile Deloitte’s survey of nearly 2,000 executives across Europe and the Middle East found that 91% plan to increase AI investment this year, even as most report it takes two to four years to see satisfactory returns.

Across decades of tech innovation and adoption cycles, we’ve seen this pattern before. Recent research from data protection firm AvePoint confirms what many in the industry are experiencing: 86% of AI deployments now face delays of up to a year, primarily due to data security and quality challenges. In 2026, tech leaders are learning a painful lesson: the problem with scaling AI adoption isn't understanding the algorithm, it's the data you put into it.

This contradiction reflects a broader pattern in enterprise AI adoption: while the technology's transformative potential remains undeniable, the execution gap has grown wider, not narrower, as investments have scaled. Companies invested, and continue to invest, in expansive new AI technologies for the promise of massive transformations in innovation and productivity—and frankly, for the fear of missing out. Understanding this contradiction requires examining both sides of the equation: the massive financial commitments driving AI forward, and the persistent execution gaps holding it back.

The Story so Far: Big Spending, Unclear ROI

EY’s latest survey found that 21% of senior leaders say their organizations have already invested $10 million or more in AI , with a third planning to spend at least that much next year. On top of this, Google Cloud’s research shows that some early adopters are now dedicating up to 39% of their IT budgets to AI .

With so much investment (and in such a short period of time), you’d expect transformational results. But the reality is more sobering and shows a growing disconnect. AvePoint’s annual AI report also found that 75% of organizations implementing AI have experienced at least one AI-related breach in the past year. A wave of disillusionment is setting in as early adopters grapple with high project failure rates and the stark realization that AI’s current output often falls short of the initial hype.

According to John Peluso, AvePoint’s Chief Technology Officer, “Across the tech sector, we’re witnessing a massive disconnect: organizations are spending big on world-class AI models and tools, but feeding them low-quality, vulnerable and mismanaged data. While companies funnel record percentages of their IT budgets into sophisticated AI models, they are discovering that no amount of algorithmic power can bypass a broken data strategy. If your data is messy or ungoverned, 'big spending' is simply funding a more expensive way to fail.” The Foundation Problem: Avoiding “Pilot Purgatory”

The challenge in achieving ROI is simple. Most organizations are still struggling with the basics: messy data, unclear business cases, and employees who aren’t convinced these tools will help them do their jobs. In fact, according to a recent survey from my company, Prosper Insights & Analytics , 58.7% of professionals are still extremely or very concerned about how their privacy may be violated by AI using their data. As a result, they can’t iterate or experiment with AI because they haven’t laid the proper data security and cultural groundwork.

The data shows organizations are recognizing this reality. The same study from AvePoint indicates that 64.4% of organizations are increasing their investment in AI governance tools and 54.5% are boosting data security tool investments. This shift suggests that smart companies are moving beyond the rush to implement AI and focusing instead on building the infrastructure necessary for success.

Instead of transforming their businesses, they’re stuck in what we can call “pilot purgatory”, watching costs rise while the benefits remain out of reach. But this doesn’t mean AI is failing. It means the approach needs refinement.

Technology leaders across all sectors should be optimistic about AI’s promise – the data suggests that the ROI that businesses are seeking isn’t as far out of reach as the headlines suggest. The key lies in getting three fundamentals right before chasing the latest AI capabilities.

First, it’s important to understand that governance is the accelerator—not the brake. And to that end, it’s important to clean up your data. You cannot build reliable AI systems on unreliable data foundations. Organizations that invest in data quality, governance and security upfront avoid the delays and disappointments that plague so many AI initiatives. Once you’ve laid that strong foundation for success, you’re going to move quicker, with greater success.

Second, stop buying tech, and start buying outcomes. Instead of spending without clear intention—purely out of a sense of “FOMO”—it’s important to find and articulate clear business use cases. AI works best when applied to specific well-defined business problems rather than used as a general solution looking for problems to solve. The most successful implementations focus on clear use cases with measurable outcomes. By spending intentionally, with clearly defined use cases, you can avoid being 1 of them.

Finally, prepare your people. That means going beyond training them on AI tools – and helping employees to understand how AI enhances rather than replaces their work. Education is a critical part of this equation – as 66.1% of employees surveyed by Prosper Insights & Analytics still report having “not heard of” Agentic AI yet.

In 2026, the companies that will win with AI are those using this period of market recalibration to strengthen their fundamentals rather than scale back their ambitions. The AI revolution isn’t slowing down; it’s maturing.

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, Bloomberg, and the London Stock Exchange Group for economic benchmarking.