AI investment soars, ROI remains elusive.

Global business adoption of AI has surged, yet returns remain scarce. Nearly nine in ten organizations now use AI in at least one function, but only 37% can attribute any meaningful financial benefits to these technologies ([1]). A mere 6% of companies qualify as 'AI high performers' that see significant profit impact from AI at the enterprise level ([2]). In short, after years of experimentation, most companies are still waiting for AI to pay off.

One recent study by MIT drives this point home: it found that about 95% of enterprise generative AI projects delivered zero measurable return in their first six months ([3]). Just 5% of AI initiatives were generating substantial value in that timeframe ([4]). This disconnect between expectation and reality has rattled executives who bought into the hype of AI’s transformative potential.

The reasons for this ROI gap are becoming clearer. While nearly 80% of employees report that AI has improved their individual productivity ([5]), those gains often remain stuck at the team or task level. McKinsey notes that many companies have failed to redesign processes and workflows to fully leverage AI, so efficiency improvements aren’t translating into broader business performance gains ([6]). Put simply, technology alone is not enough – without complementary organizational change, AI’s benefits stay siloed and do not reach the bottom line.

Skyrocketing AI investments are prompting hard questions about economics. In 2026, global AI-related capital expenditures are expected to reach a staggering $527 billion ([1]). Even tech giants are feeling the strain: Amazon’s plan to spend $200 billion on AI and cloud this year is projected to push it $17 billion into negative free cash flow ([2]), and analysts say Meta’s massive AI outlays will slash its free cash flow by almost 90% ([3]). Meanwhile, about 20% of companies report that rising AI operating expenses – from costly computing infrastructure to increasing model licensing fees – are already constraining their AI usage ([4]).

As AI’s price tag grows, financial scrutiny is intensifying. Nearly 80% of finance leaders say they cannot fully link AI spending to specific business outcomes, leaving them effectively writing blank checks for AI initiatives ([5]). In one survey, 87% of CFOs warned they must demonstrate concrete ROI from AI within 12 months ([6]), but only 22% can currently do so ([7]). Furthermore, about three in five finance chiefs admit they’re already investing more in AI than they can rationalize with actual results ([8]).

Boardrooms are likewise enforcing discipline. Just 26% of corporate boards are still willing to approve AI investments without a clear ROI plan, while 66% now insist on seeing proof of value before green-lighting new funding ([9]). In fact, 22% of boards say they won’t authorize any additional AI spending until existing projects demonstrate real returns ([10]). Investors are also impatient: over half now expect positive ROI from new AI initiatives in six months or less ([11]) – a timeline that 84% of CEOs themselves say is overly optimistic ([12]).

A new twist in AI economics is a shifting 'build vs. buy' calculus. Nearly one-third of enterprises report they have opted against purchasing certain software tools because they realized they could develop similar capabilities internally using generative AI coding tools ([13]). This may save on vendor costs in the short term, but it shifts expenditures to internal cloud infrastructure and talent – and doesn’t guarantee success. Such decisions reflect how AI is reshaping IT strategy and budgets, forcing leaders to carefully weigh immediate cost savings against long-term value and complexity.

Many companies are grappling with AI projects that underdeliver. The issue isn’t a lack of advanced technology, but rather an inability to scale and integrate it effectively. Nearly two-thirds of organizations remain stuck in pilot projects or isolated deployments, unable to roll AI out broadly across the enterprise ([1]). This 'pilot purgatory' means lots of experimentation without material impact. Leaders often initiate dozens of AI proofs-of-concept, yet without a plan to operationalize them, these efforts never reach the critical mass needed to move the financial needle. As one industry expert put it, simply ramping up AI usage doesn’t automatically create value – it requires rethinking business processes to capture the benefits ([2]).

When initiatives don’t show immediate wins, they get shelved. Forty-two percent of companies abandoned at least one AI project last year due to unclear ROI – a jump from 17% the year before ([3]). For many, the initial excitement has given way to impatience. Gartner even found that organizations which cut staff in the name of AI generally saw no ROI improvement as a result ([4]), highlighting that you can’t simply cost-cut your way to a successful AI outcome. Without clear line-of-business impact, CFOs often hit pause – three-quarters of finance leaders who can’t measure AI’s value have stopped or limited funding for AI initiatives, and over a third have cancelled projects mid-stream ([5]).

High-profile stumbles underscore what goes wrong. Volkswagen’s $7.5 billion push to build a unified automotive AI software platform ended in a very costly failure ([6]). The project suffered from fundamental missteps in strategy and execution rather than just technology issues – traditional car manufacturing practices clashed with the need for agile, iterative software development, and competing internal priorities led to bloated requirements and delays ([7]). The result was years of postponed product launches, massive financial losses, and a vivid warning that even deep pockets can’t guarantee AI success in the absence of clear goals, agile methods, and cross-company alignment ([8]) ([9]).

A small fraction of companies are proving that meaningful AI ROI is achievable with the right approach. These AI leaders share a common thread: relentless focus on use cases that drive tangible business outcomes. They treat AI initiatives like any other investment – setting concrete targets for cost savings, revenue gains, or efficiency improvements – and integrating AI into core processes rather than experimenting on the sidelines ([1]). For example, New York Life’s technology team prioritized projects in areas with clean data and clear value potential, reinvesting early AI wins to fund subsequent initiatives and designing AI solutions for reusability across the organization ([2]). This disciplined, value-first strategy ensured each AI project was aligned with the company's financial objectives from day one.

Another hallmark of successful programs is viewing AI as an augmentor of human capability, not just a replacement for headcount. Companies using AI to empower employees – rather than to cut costs by automation alone – are seeing better financial outcomes. One industry analysis found that firms that augmented their workforce with AI achieved roughly 2× higher cash-flow margin expansion than those that focused primarily on labor reduction ([3]). Such results suggest that AI yields more value when it’s embedded into workflows to enhance employee productivity and decision-making, as opposed to treating it as a bolt-on tool or a quick cost-cutting fix.

Even within sectors known for caution, there are breakout successes. Canadian telecom provider TELUS, for instance, rolled out an enterprise-wide generative AI platform that gave its 57,000 employees access to a range of AI models via a secure in-house system. This enabled staff to create over 13,000 custom AI-powered tools for tasks from customer service to software development. The outcomes have been concrete: more than 500,000 hours of work saved and over $90 million in realized benefits so far ([4]). TELUS’s achievement illustrates the payoff of scaling AI thoughtfully – focusing on empowering teams, ensuring data privacy, and measuring results. Such examples can serve as playbooks for other organizations looking to turn AI investments into real returns.