Your AI budget keeps growing but the returns still fall short

Increased AI budgets are not delivering predicted returns

AI investment is growing at an exceptional pace. Boardrooms continue to approve larger budgets because the technology is improving quickly and the competitive pressure is real. The problem is not that companies are investing in AI. The problem is that many organizations expect financial returns that their operating model cannot produce.

This gap exists because technology alone does not change how a business works. Many AI programs are built on assumptions that processes, data, and governance are already in place. In reality, those foundations are often incomplete. Companies automate existing workflows without questioning whether those workflows should exist in their current form. As a result, AI executes inefficient processes faster instead of creating meaningful business value.

Organizations that consistently achieve stronger returns approach AI differently. They do not begin with the latest model or the newest automation platform. They begin by asking which business outcomes matter most, how work should be redesigned, and what data is required to support reliable decisions. Technology becomes the final step.

For executives, this changes how AI investments should be evaluated. Budget size is not a reliable indicator of success. Leadership attention is. Process redesign, data governance, accountability, and cross-functional execution have a much greater impact on return on investment than simply purchasing more advanced AI tools.

Bain & Company’s survey of 951 global companies illustrates the challenge. While 37% of companies targeted cost reductions between 11% and 20%, nearly 40% of those measuring outcomes achieved only 0% to 10% savings. Despite these results, 90% of companies planned to increase their AI spending. The data suggests that many organizations are expanding investment before resolving the reasons earlier investments underperformed.

AI agents currently rely heavily on human oversight

AI agents are advancing rapidly, but most businesses are still operating far from fully autonomous systems. Much of the public discussion focuses on AI making complex decisions independently. Production environments tell a different story. Most organizations continue to rely on humans to review, approve, or intervene in important decisions.

For many business processes, especially those involving financial, legal, regulatory, or customer impacts, human oversight is a sensible control. The challenge appears when financial projections assume near-complete automation while operational reality still depends on people. That difference directly affects labor costs, productivity gains, and overall return on investment.

Executives should build AI business cases around current capabilities rather than future expectations. If an AI agent requires frequent human approval today, the financial model should reflect that reality. As the technology matures and trust increases, organizations can gradually increase autonomy. This creates more predictable investment outcomes and reduces the risk of disappointing stakeholders.

Companies that achieve stronger results are not necessarily deploying more advanced AI. They are aligning expectations with operational reality. They understand exactly where human judgment adds value, where AI can operate safely, and where additional automation is likely to generate measurable returns. That discipline produces more reliable economics than assuming autonomy that does not yet exist.

The Bain & Company survey highlights the current state of adoption. Only 7% of companies reported running fully autonomous AI agents in production. The largest group, representing 38% of respondents, still requires human approval before agents complete important actions, while another 32% use AI agents with defined guardrails and human intervention for exceptions. The survey also found that companies meeting their savings targets were more likely to have agents operating with higher levels of autonomy than organizations that missed their financial goals. The lesson is not to remove people as quickly as possible, but to match investment assumptions with what AI can reliably deliver today.

New AI investments are often funded by unrealized prior returns

Many organizations say they will fund the next generation of AI from the savings created by earlier automation programs. On paper, this is a disciplined approach. In practice, it only works if those savings were actually achieved.

The problem is that many companies continue to rely on projected returns instead of verified financial results. Earlier automation initiatives may have improved productivity or reduced manual work, but the expected cost reductions often did not fully reach the bottom line. When new AI budgets are based on those original projections, rather than audited outcomes, financial risk increases with every investment cycle.

This creates an important leadership responsibility. Before approving another AI initiative, executives should ask a simple question: What did the previous investment actually deliver? That answer should come from measurable business outcomes.

This requires close coordination between finance, operations, and technology leaders. The chief financial officer should validate realized savings. Business leaders should confirm whether operational improvements were sustained over time. Technology leaders should explain where AI delivered value and where organizational barriers limited adoption. When these perspectives are combined, investment decisions become significantly more reliable.

There is also an important distinction between productivity gains and financial savings. AI may allow employees to complete work faster, but unless that additional capacity is redirected toward higher-value work or reduces operating costs, the organization may not capture measurable financial returns. Boards should expect evidence that productivity improvements have translated into business value before treating them as funding sources for future AI programs.

According to Bain & Company’s survey, 44% of companies plan to fund generative AI and agentic AI investments using savings from previous automation programs. That makes it the most common funding approach among respondents. The strategy can be effective, but only when those savings have been independently verified rather than assumed.

Data access and integration remain the primary barrier to AI success

AI systems depend on reliable access to business data. If information is fragmented across departments, stored in incompatible systems, or governed inconsistently, even the most advanced AI models will produce limited business value. The technology is often ready before the organization is.

Many companies have invested heavily in modernizing their data infrastructure over the past decade. Even so, connecting data across the enterprise remains difficult. Different business units often use different systems, maintain separate standards, and define the same information in different ways. These organizational issues reduce AI performance far more than model quality alone.

The highest-performing organizations do not wait for perfect data before deploying AI. Instead, they identify business processes where data is already available, establish clear governance, and begin creating value immediately. As these projects mature, they improve data quality and integration through practical implementation rather than delaying progress until every data issue is resolved.

This reflects an important shift in executive thinking. Data management is a business capability that affects strategy, operations, risk management, and customer experience. Executive teams that treat data governance as a leadership priority are generally better positioned to scale AI across the enterprise because they reduce organizational barriers before they become operational constraints.

Another important observation is that companies achieving stronger AI results often report data challenges more frequently than weaker performers. This does not mean they have poorer data. It often means they are deploying AI more broadly and therefore encounter integration issues that smaller or less mature programs have not yet faced. Their advantage comes from actively addressing these challenges rather than allowing them to delay progress.

Bain & Company’s survey found that 41% of respondents identified data access and integration as the single biggest barrier to AI adoption, ahead of compliance concerns, budget limitations, skills shortages, and executive buy-in. Among companies that achieved their business targets, 44% cited data as their primary challenge, compared with 40% of companies that underperformed. The findings suggest that successful organizations recognize data complexity early and manage it directly instead of treating it as a reason to postpone AI initiatives.

Redesigning workflows is critical before automating processes

Many AI initiatives begin with a simple question: where can we apply AI? That is often the wrong starting point. The better question is whether the process itself still makes sense. If a workflow contains unnecessary approvals, duplicate work, manual transfers between teams, or outdated policies, AI will simply execute those activities more efficiently without improving the underlying business outcome.

Every organization accumulates process complexity over time. New regulations, acquisitions, legacy systems, and evolving business priorities all add layers of work. Employees develop manual workarounds to keep operations moving, and those workarounds eventually become part of the standard process. If AI is introduced without addressing these issues, the organization may automate activities that no longer create value.

Executives should encourage teams to redesign important workflows before selecting AI technologies. This means identifying which steps genuinely contribute to customer value, regulatory compliance, or business performance, and removing activities that exist only because of historical decisions. AI should then be applied to the streamlined process rather than the original one.

This approach also improves long-term flexibility. Processes designed around clear business objectives are easier to update as AI capabilities evolve. Organizations avoid becoming dependent on automated versions of inefficient workflows that later require expensive redesign.

Business leaders should also recognize that process redesign is not solely an operational exercise. It requires collaboration across functions because decisions made in one department often affect performance elsewhere. Finance, operations, technology, legal, and customer-facing teams all need a shared understanding of how work should flow before automation begins. This alignment reduces implementation risk and increases the likelihood that AI investments produce measurable business outcomes.

Organizations should ask, “If we were designing this process from scratch today, what would it look like?” Only after answering that question should the technology discussion begin.

Verified financial assumptions and clear governance are essential for AI program success

Strong AI programs are built on disciplined financial planning and clear accountability. Both are often underestimated. Organizations frequently spend significant time evaluating technology vendors while giving much less attention to validating business assumptions or defining who is responsible when AI systems make important decisions.

Every AI investment should begin with an honest review of previous results. If earlier automation projects delivered only part of their expected savings, future business cases should reflect those actual outcomes. Building investment plans on verified performance creates more realistic expectations and improves capital allocation. It also strengthens confidence among boards, investors, and executive teams because decisions are supported by measurable evidence instead of optimistic projections.

Governance is equally important. As AI systems become more capable, they will influence decisions involving customers, employees, financial transactions, and regulatory compliance. When something goes wrong, organizations cannot afford uncertainty about ownership. Accountability must be established before deployment.

This responsibility extends beyond the technology organization. While IT manages infrastructure and technical implementation, decisions about acceptable risk, customer impact, legal obligations, and business priorities belong to executive leadership. CEOs and senior leadership teams should determine who owns AI governance, how decisions are reviewed, and what escalation procedures exist for high-impact situations.

Effective governance should also evolve alongside AI capabilities. As organizations gradually increase the autonomy of AI systems, oversight mechanisms, performance monitoring, audit processes, and risk controls should be updated accordingly. Governance is not a one-time approval process. It is an ongoing management discipline that supports responsible scaling.

Deploy AI in targeted use cases to address data challenges incrementally

Many organizations delay AI adoption because they believe their data infrastructure must be fully modernized first. This assumption often slows progress unnecessarily. While strong data foundations remain important, waiting for enterprise-wide perfection can postpone meaningful business value for years.

A more effective approach is to identify focused, high-value workflows where the necessary data already exists and is reasonably reliable. These projects allow organizations to demonstrate measurable results, strengthen internal capabilities, and improve confidence in AI without requiring large-scale transformation before deployment.

Executives should prioritize use cases where employees spend significant time collecting, organizing, validating, and summarizing information. These repetitive activities are often well suited to AI because the process is clearly defined and the benefits can be measured. Success in these targeted areas creates operational improvements while generating insights that support broader AI adoption across the organization.

Early projects also provide valuable operational experience. Teams learn how to manage AI outputs, improve data quality, establish governance practices, and build trust among employees. These lessons become increasingly valuable as AI expands into more complex business functions. Rather than treating early deployments as isolated experiments, organizations should use them to establish repeatable practices that support future growth.

Business leaders should also recognize that AI can improve data management itself. AI systems can help classify information, identify inconsistencies, automate data extraction, and reduce manual reporting work. This means AI is not only dependent on data quality but can also contribute to improving it over time.

Successful AI deployment demands a redesign of the operating model

AI changes how work is performed across an organization. Automating individual tasks is only part of the transformation. As AI systems assume more routine activities, employees increasingly focus on supervising AI, making complex decisions, handling exceptions, and improving business outcomes. This requires organizations to rethink how work is structured, how teams collaborate, and how performance is measured.

Many companies underestimate this organizational shift. They invest heavily in technology while expecting existing roles, reporting structures, and management practices to remain largely unchanged. This creates friction because employees continue working within operating models designed for manual processes rather than AI-supported ones.

Executives should view operating model redesign as a core component of AI strategy. This includes redefining job responsibilities, updating decision-making authority, investing in workforce training, and creating new collaboration models between business teams and technology functions. These changes help employees move toward higher-value activities while allowing AI systems to handle repeatable operational work.

Change management is equally important. Employees need clear guidance on how AI will affect their responsibilities, what new skills will be expected, and how success will be evaluated. Organizations that communicate these changes early are generally better positioned to build trust, encourage adoption, and reduce resistance during implementation.

Leadership commitment also plays a decisive role. When operating model changes are actively sponsored by CEOs and executive teams, organizational priorities become clearer and transformation efforts receive the authority needed to succeed. AI initiatives become business transformation programs rather than isolated technology projects.

Enterprise-level outcome measurement is crucial for AI investment success

Many organizations evaluate AI using project-specific metrics such as hours saved, tasks automated, or short-term cost reductions. These measurements are useful, but they do not provide a complete picture of whether AI is improving business performance. An AI initiative can meet every operational target and still fail to create meaningful value for the organization as a whole.

Executive teams should evaluate AI based on enterprise outcomes that directly support strategic objectives. These include faster decision-making, higher-quality decisions, improved customer experience, stronger revenue growth, greater operational resilience, and more effective risk management. When AI is measured against business outcomes instead of isolated project metrics, leadership gains a clearer understanding of whether investments are strengthening the company’s competitive position.

This also changes how success is managed across the organization. Individual business units naturally optimize for their own objectives, whether that is reducing costs, increasing productivity, or improving service levels. Without enterprise-wide performance measures, these local improvements may not translate into better organizational performance. Leaders should therefore establish a common set of executive metrics that connect AI initiatives to broader business priorities.

Measurement should also extend beyond the initial implementation phase. AI systems continue to evolve as business conditions change, new data becomes available, and employees develop new ways of working. Regular performance reviews help organizations determine whether AI continues to deliver value, identify where adjustments are needed, and ensure that investments remain aligned with strategic goals.

Another important consideration is balancing financial and non-financial outcomes. Some AI initiatives produce immediate cost savings, while others create value by improving customer retention, accelerating product development, strengthening regulatory compliance, or enhancing decision quality. Executive dashboards should reflect both categories because long-term competitive advantage often depends on improvements that cannot be measured through cost reduction alone.

Concluding thoughts

The next phase of AI adoption will not be defined by who spends the most. It will be defined by who executes the best.

The technology is improving at an extraordinary pace, but technology is no longer the primary constraint. The real differentiators are leadership decisions, disciplined execution, and organizational readiness. Companies that continue treating AI as a technology initiative will likely see incremental improvements. Those that treat it as a business transformation effort are far more likely to achieve lasting competitive advantage.

For executive teams, this means asking tougher questions before approving the next investment. Are your workflows designed for AI, or are you automating outdated processes? Are your projected returns based on verified business outcomes? Is accountability for AI decisions clearly defined? Are you measuring enterprise performance instead of isolated project metrics?

The organizations creating the greatest value are not waiting for perfect conditions. They are making deliberate decisions, solving practical business problems, and building the capabilities needed to scale responsibly. They recognize that success depends as much on governance, operating models, and leadership as it does on the AI itself.

AI will continue to reshape every industry. The opportunity is significant, but so is the execution challenge. The companies that consistently generate strong returns will not necessarily have the largest AI budgets. They will be the ones that combine disciplined investment, organizational alignment, and a relentless focus on measurable business outcomes. That is what turns AI from a promising technology into a durable business advantage.