AI adoption surges, ROI lags: the reality behind the hype.
Adoption versus outcome.
Uptake is high but payoff is low. In the UK, for example, 78% of firms report using AI tools, yet only 31% say they see a positive ROI ([1]); another 18% say projects failed. In retail, 90% of leaders are experimenting with AI, but 96% admit they still see no returns ([2]). Analysts caution that this is often the result of unchecked hype: early pilots came with "vague KPIs" and “a noticeable lack of measurable ROI” ([3]). Fewer than half of adopters even define what “success” looks like ([4]), meaning many projects proceed without clear goals.
For C-suite leaders, the imperative is to demand proof of impact. Don’t be swayed by buzz or experiments alone. Insist that every AI initiative tie to concrete metrics (cost savings, revenue gains, efficiency) and that ROI is measured end-to-end. Otherwise these projects risk delivering nothing more than well-publicized pilots.
Spending versus returns.
Companies are pouring money into AI, and investors are starting to worry about returns. WindowsCentral reports that Microsoft plans to spend about $146 billion on AI infrastructure in 2026 (nearly double its 2025 budget) ([1]). Yet this spending spooked Wall Street – Microsoft’s stock fell roughly 25% in the quarter amid concerns its AI outlays weren’t paying off. Other data show enterprise budgets shifting accordingly: one analysis of $18B in software spend found midmarket and enterprise organizations boosted their AI-related spend by ~58% year-over-year ([2]) (even as smaller companies cut back on legacy SaaS). At the same time, vendors treating AI as mission-critical are raising prices substantially; many now charge 20–37% more for AI-enabled features ([3]).
These trends put pressure on the P&L. Even feature adoption can fail to translate into revenue: Microsoft notes Copilot’s user base tripled year-over-year, but only ~3.3% of Microsoft 365 users who interact with Copilot actually become paying customers ([4]). In boardrooms, executives are now asking bluntly, “What does each AI-driven decision actually cost (including compute and carbon)?” ([5]). The lesson for CFOs is clear: scrutinize every AI dollar. Plan for rising costs — not just for cloud cycles but for training, data engineers and overhead — and require projected ROI before expanding budgets.
Infrastructure and data realities.
Often, AI stumbles on foundations. As one expert notes, many companies \u201cjump[ed] headfirst into\u201d AI without the bigger picture, only to find “data isn’t even close to being AI-ready” ([1]). In practice, this means pilots that looked successful in isolation hit immediate roadblocks in production. Conversational AI chatbots, for example, may generate plausible text, but the data behind them often "doesn’t always flow cleanly" – it can be duplicated, incomplete or out of sync ([2]). The result is that teams spend precious time merging and cleaning data instead of automating.
More broadly, legacy IT is now one of the biggest obstacles to value. Many companies never architected systems for AI’s data intensity and scale; what they have instead is often called “a patchwork of disconnected systems” ([3]). These old architectures slow down or even negate AI efforts. For CFOs, this translates to a hidden tax: adding new AI tools on top of brittle infrastructure will build more manual work and costs, not less.
Instead of pouring more into flashy tools, leaders should shore up the plumbing. Invest in data governance, cloud pipelines and integrated analytics first. When the data foundation is solid and systems communicate, AI tools can start delivering on their promise. Otherwise, even expensive AI licenses will simply layer more complexity onto broken processes.
Governance and strategy.
With hype giving way to scrutiny, executives must impose discipline. Boards are now pressing managers with pointed questions: What does each AI-driven decision cost? How are models governed? ([1]) In one forecast, committee members are said to demand full cost assessments for inference (compute and energy) as part of evaluating AI projects.
Despite this, many organizations lack oversight. According to new research, 43% of large firms have no formal AI risk framework ([2]). Even as 86% report AI-driven productivity gains for employees ([3]), fewer than half have defined success metrics or conducted impact assessments. In other words, companies let AI run forward but have few controls or scorecards to reflect on performance.
CFOs and CEOs should demand an enterprise-scale AI strategy: every project needs a clear business case, a governance trail, and defined ROI checkpoints. This means setting structured KPIs, assigning accountability for outcomes, and training teams on how AI changes workflows. (It also means cutting projects that don’t deliver.) The organizations that get ROI are those treating AI like any major transformation: building cross-functional change programs, continuous upskilling, and stage gates for investment. If CFOs insist on these guardrails from day one, AI spending will become an investment in-value, not just a line-item fueled by excitement.