After billions in AI spending with little to show for it, organizations are making a surprising choice: they’re investing even more. In our work with data leaders across industries, we’ve seen this pattern emerge repeatedly. The question isn’t whether to continue investing in AI — it’s how to ensure those investments finally deliver the returns they promise.
The numbers tell a story that should give any executive pause. According to a recent survey of 350 global CEOs by Teneo, fewer than half report tangible returns from their AI projects. Separate research from MIT's NANDA initiative report "The GenAI Divide: State of AI in Business 2025" paints an even starker picture, with 95% of AI initiatives failing to generate positive ROI. Yet despite these sobering statistics, 68% of CEOs plan to increase AI spending in the coming year, potentially pushing global AI investment past the $2 trillion mark. This isn’t irrational exuberance. It’s a calculated response to a complex reality that many organizations now face.
The Uncomfortable Truth About AI Returns
The gap between AI promise and AI performance has become impossible to ignore. Organizations have spent the past two years deploying AI tools, licensing platforms, and launching initiatives intended to revolutionize operations, unlock productivity gains, and create competitive advantage. For most, these investments haven’t delivered.
The pattern we see repeatedly includes:
- Rushed deployments that prioritized getting AI tools in front of users over ensuring they could use them effectively
- Disconnected initiatives where different departments pursued AI projects without coordination or shared infrastructure
- Technology-first thinking that assumed the AI itself would solve business problems without addressing underlying data and process issues
- Unclear success metrics that made it nearly impossible to measure real impact or justify continued investment
These aren’t just implementation challenges. They represent a fundamental misunderstanding of what makes AI successful in enterprise environments. Organizations discovered that simply licensing AI tools doesn’t translate to business value — even when adoption rates look promising on paper.
The reality many organizations have discovered is sobering. Even when enterprises successfully deploy AI tools across their workforce, actual usage often falls far short of expectations. More critically, the employees who do adopt these tools frequently use them for convenient but low-impact tasks rather than in ways that create meaningful business value or advance strategic objectives. High deployment numbers mask low actual value creation.
This gap between deployment and value realization has become the defining challenge of enterprise AI adoption. It’s also why simply spending more money on the same approach is unlikely to produce different results.
Why CEOs Are Choosing to Double Down
Given these disappointing returns, why are leaders planning to increase AI investment rather than cutting their losses? The answer is more nuanced than simple optimism about future potential. The commitment to continued AI investment reflects several converging pressures:
First, many organizations have already invested so heavily in AI transformation that reversing course would mean writing off not just financial investments, but also the organizational changes, training programs, and strategic pivots that came with them. The cost of unwinding these commitments — both tangible and intangible — often exceeds the cost of pushing forward.
Second, competitive dynamics have created a situation where standing still feels riskier than continuing to invest. Leaders worry that reducing AI spending will signal to boards, investors, and markets that they’re falling behind. In an environment where AI capabilities are increasingly viewed as table stakes for competitive relevance, being perceived as behind the curve carries real consequences.
Third, and perhaps most importantly, the promise of AI hasn’t been disproven — only the initial approach to realizing that promise. Leaders recognize that the technology itself is sound, but the way organizations have attempted to implement it has been flawed. The solution isn’t to abandon AI, but to fundamentally change how they approach AI investment and deployment.
The Real Cost of Stopping Now
The decision to continue investing isn’t just about avoiding sunk costs or maintaining appearances. There are legitimate strategic reasons why reducing AI investment now could be more costly than seeing the transformation through.
Organizations that significantly cut AI spending face several risks:
- Competitive disadvantage as competitors who persist with smarter approaches begin to realize genuine advantages
- Talent challenges as top performers seek opportunities at organizations committed to modern, AI-enabled ways of working
- Missed efficiency gains that could help navigate economic uncertainty or market changes
- Strategic inflexibility as business processes and systems become increasingly difficult to adapt without AI capabilities
The question organizations face isn’t whether to invest in AI, but how to ensure that continued investment produces better results than the first wave of spending. This requires a fundamentally different approach — one focused less on deploying AI tools and more on building the foundation that makes AI work.
A strong data foundation leads to clear AI models
Where Past AI Spending Went Wrong
To invest more effectively going forward, it’s worth examining where the first wave of AI spending fell short. In our experience working with organizations to implement AI successfully, we’ve identified several common pitfalls that undermined initial investments. The most significant issues include:
Tool-First Thinking
Organizations rushed to deploy the latest AI platforms and capabilities without first establishing whether their data, processes, and people were ready to use them effectively. They assumed that powerful AI tools would somehow overcome deficiencies in data quality, unclear business processes, or inadequate change management. This is roughly equivalent to buying Formula 1 race cars for a team that hasn’t learned to drive stick shift — the technology’s potential is irrelevant if the fundamentals aren’t in place. Many organizations deployed enterprise AI tools only to discover their data wasn’t ready to support meaningful use cases.
Scattered Efforts
Instead of building centralized capabilities that could serve multiple use cases, organizations allowed different departments to pursue independent AI initiatives. This created redundant investments, inconsistent approaches to data and governance, and made it nearly impossible to learn from successes or failures across the organization. Each team was essentially starting from scratch, rebuilding infrastructure and capabilities that other teams had already developed. Siloed AI projects create redundant costs and prevent organizations from building reusable capabilities that scale.
Adoption Over Enablement
Organizations measured success by how many employees had access to AI tools rather than how effectively those employees used them to create business value. They focused on deployment timelines and license utilization rates while ignoring whether users understood how to integrate AI into their workflows in ways that improved outcomes. High adoption numbers became a vanity metric that masked low actual value creation. Giving people access to AI tools doesn’t automatically translate to business value — enablement requires understanding how work needs to change.
Infrastructure Neglect
Perhaps most critically, organizations underestimated the importance of data infrastructure, governance, and semantic understanding in making AI work. They assumed that AI models were sophisticated enough to work around messy data, unclear definitions, or inconsistent taxonomies. In reality, AI amplifies data quality issues and requires even higher standards for metadata, lineage, and semantic clarity than traditional analytics. AI doesn’t fix bad data — it makes the consequences of poor data quality more visible and more costly. These mistakes share a common thread: they prioritize the technology itself over the organizational capabilities, data foundations, and process changes that determine whether AI creates value. The good news is that these are fixable problems. The challenge is that fixing them requires a different kind of investment than simply licensing more AI tools.
A Smarter Approach to AI Investment in 2026
The organizations that will succeed with AI in 2026 won’t necessarily be those that spend the most. They’ll be the ones that invest strategically in the capabilities that make AI deployments successful. Based on our work helping organizations build these capabilities, we recommend focusing AI investment in several key areas.
Identify High-Impact Use Cases
Stop spreading AI investment across numerous experimental projects and instead concentrate resources on initiatives with clear business impact and realistic paths to implementation. This means developing rigorous processes for evaluating potential AI use cases based on expected value, implementation complexity, and organizational readiness. Organizations should be able to articulate specific business outcomes for each AI investment and define clear metrics for measuring success. Move from “throwing things at the wall” to strategic deployment of AI where it can deliver measurable business impact.
Redesign Work, Not Just Workflows
Meaningful AI impact requires rethinking how work gets done, not just automating existing processes. This often means redesigning job roles, decision rights, and organizational structures to take advantage of AI capabilities. Organizations need to invest in understanding how AI should change the nature of work in different roles and functions, then provide the change management support to help people adapt to these new ways of working. AI’s potential isn’t realized by speeding up old processes — it comes from enabling entirely new approaches to creating value.
Build Reusable Capabilities
Instead of custom-building infrastructure for each AI use case, invest in creating shared data platforms, governance frameworks, and integration capabilities that can support multiple applications. This includes establishing common approaches to data quality, metadata management, security, and compliance that become organizational capabilities rather than project-specific solutions. Organizations that build once and reuse often will see dramatically better returns on AI investment than those rebuilding infrastructure for each initiative.
Prioritize Enablement Over Access
Shift investment from licensing more AI tools to ensuring people can use existing tools effectively. This includes training programs, workflow redesign, success metrics that focus on outcomes rather than adoption, and ongoing support that helps users discover new ways to leverage AI in their work. Organizations should measure success by business impact, not by how many employees have logged into an AI platform. The limiting factor in AI value creation is rarely access to technology — it’s the ability to use that technology effectively.
AI tools, when set up correctly, can lead to controlled outcomes
The Foundation That Makes AI Spending Worthwhile
All of these strategic investments share a common requirement: they depend on having the right data foundation in place. This is where many AI initiatives have failed, and it’s where organizations need to focus if they want different results from their next wave of AI spending. The data capabilities that determine AI success include:
Semantic Intelligence
AI systems need to understand the contextual meaning of data, not just process it syntactically. This requires business glossaries, taxonomies, and ontologies that create a shared understanding of business concepts across the organization. Without semantic intelligence, AI models make mistakes that undermine user trust and limit their practical utility. Organizations need to invest in metadata management and semantic modeling as foundational capabilities for AI success. AI can only be as smart as the semantic foundation it’s built on — garbage in, garbage out still applies.
Data Quality at Scale
While data quality has always been important for analytics, AI demands higher standards and broader coverage. AI models consume far more data than traditional analytics, often pulling from sources that weren’t originally designed for analytical use. Organizations need to invest in data quality capabilities that can operate at the scale and speed AI requires, with automated monitoring, issue detection, and remediation. AI amplifies data quality problems — what was a minor issue in a report becomes a major failure in an AI-driven process.
Governance that Enables Rather than Restricts
Effective AI governance isn’t about creating barriers to innovation; it’s about establishing guardrails that allow teams to move quickly while managing risk. Organizations need governance frameworks that address AI-specific concerns around bias, explainability, and ethical use while still enabling rapid experimentation and deployment. This requires updating traditional data governance practices to account for AI’s unique characteristics and risks.
Good governance accelerates AI adoption by providing clarity and confidence — poor governance creates bottlenecks that slow everything down.
Integrated Data Architecture
AI initiatives often need to pull data from across the organization, combining structured and unstructured sources in ways that traditional analytics rarely required. Organizations need data architectures that can support this kind of integration without creating fragile point-to-point connections or data silos. This typically requires investment in modern data platforms, API strategies, and integration patterns designed for AI use cases. Siloed data creates siloed AI — breaking down those silos is often the highest-leverage investment an organization can make. In our work with organizations implementing these capabilities, we’ve seen how they transform AI from an experimental technology into a practical business tool. The organizations that invest in these foundations first, then deploy AI on top of them, achieve dramatically better results than those that try to build the foundation while simultaneously rolling out AI applications.
From Experimentation to Transformation
The shift in AI spending we’re seeing represents a maturation of how organizations think about AI. The experimental phase, where success was measured by how many AI projects were launched or how many employees had access to AI tools, is ending. We’re moving into a phase where success is measured by business outcomes and where organizations demand real returns on their AI investments. This shift requires different kinds of investments. Instead of spending primarily on AI platforms and licenses, organizations need to invest in the capabilities that make AI work: data infrastructure, semantic intelligence, governance frameworks, and organizational change. These investments are less exciting than launching new AI applications, but they’re what separate organizations that realize AI’s promise from those that continue to struggle.
The organizations succeeding with AI aren’t necessarily the ones spending the most — they’re the ones investing in the right foundational capabilities.
The good news is that these foundational investments have value beyond AI. Better data quality, clearer semantic understanding, stronger governance, and more integrated architectures improve all forms of analytics and decision-making, not just AI applications. Organizations that make these investments are building capabilities that will serve them regardless of how AI technology evolves. The challenging news is that these investments require patience and discipline. Unlike deploying a new AI tool, which can happen in weeks or months, building robust data foundations takes time. Organizations need to resist the pressure to show immediate AI wins and instead focus on creating the conditions for sustainable AI success.
Ready to Make Your AI Investments Pay Off?
The decision to continue investing in AI isn’t just about having faith in the technology. It’s about committing to doing the foundational work that makes AI successful. At FSFP, we help organizations build these capabilities — from metadata management to data governance frameworks and quality programs that support AI at scale. If you’re planning to increase AI spending in 2026, we can help ensure those investments produce better results than your first wave of AI projects. Let’s talk about how to build the foundation your AI initiatives need to succeed.
Array
