The execution gap: why most enterprise AI strategies fail to deliver and what leaders can do about it.

Executive summary

Enterprise AI adoption has reached an inflection point. Nearly nine in ten organizations now use AI in at least one business function, yet fewer than one in ten report meaningful financial returns. The gap between strategy and execution is not a technology problem. It is a leadership, governance, and organizational design problem. This paper examines five converging forces that explain why most AI initiatives stall after an initial proof of concept: the failure to treat AI as an operating model transformation, a widening leadership readiness deficit, the trap of waiting for perfect data, the inability to scale beyond early wins, and emerging questions about the long-term cost sustainability of AI infrastructure. Drawing on current research from McKinsey, BCG, Deloitte, Gartner, EY, PwC, KPMG, MIT Sloan Management Review, and Harvard Business Review, we argue that the enterprises most likely to capture AI's value in the next 12 to 18 months are those that redesign how they operate, not those that simply add AI to what already exists. For PE-backed and high-growth companies, where speed and capital efficiency are existential, the cost of getting this wrong is compounding daily.

The convergence

Something unusual is happening in the enterprise AI conversation. After two years of breathless investment, pilot programs, and vendor promises, a more sober set of questions is emerging. Not whether to adopt AI (that debate is settled) but how to make it work at a scale that justifies the cost, the organizational disruption, and the leadership attention it demands.

The numbers tell a stark story. BCG surveyed 1,803 C-suite executives in early 2025 and found that 75% rank AI among their top three priorities, but only 25% report realizing significant value. Sixty percent generate no material value despite meaningful investment (Source: BCG, "From Potential to Profit: Closing the AI Impact Gap," 2025). McKinsey's parallel finding is equally striking: 88% of organizations use AI in at least one function, but only 39% report any positive impact on earnings, and for most, that impact is less than 5% of EBIT (Source: McKinsey, "The State of AI in 2025," 2025).

These are not early-stage adoption numbers. These are the returns from organizations that have already invested, already hired, already built. The problem is not awareness or enthusiasm. It is execution. And execution, as practitioners consistently report, breaks down not in the technology layer but in the human and organizational layers that surround it.

AI is an operating model problem, not a technology problem

The most persistent misconception in enterprise AI is that it is primarily a technology initiative. Buy the platform, integrate the API, train the model. The reality that research increasingly confirms is that AI demands a redesign of how the business operates: its workflows, decision rights, team structures, and incentive systems.

McKinsey's 2025 research found that only about 6% of organizations are truly "rewiring" their operating models around AI, even though 78% use it in at least one function. High-performing organizations are nearly three times more likely to have redesigned workflows end-to-end rather than bolting AI onto existing processes (Source: McKinsey, "The State of AI in 2025," 2025). McKinsey's subsequent work on "The Agentic Organization" positions this as a paradigm shift: organizations must redesign roles, workflows, and decision rights around AI capabilities (Source: McKinsey, "The Agentic Organization," 2025).

Deloitte's "State of AI in the Enterprise 2026" quantifies the readiness gap more precisely. While AI strategy preparedness sits at roughly 40%, talent readiness is only 20% and governance readiness is 30%. Deloitte names this an "execution gap": enterprises are strategically confident but operationally unprepared. Most have focused on training employees rather than restructuring how work gets done (Source: Deloitte, "State of AI in the Enterprise," 2026). Among organizations with advanced agentic AI adoption, MIT Sloan reports that 66% expect to change their operating model and redefine roles, including flattening hierarchies and reducing middle management (Source: MIT Sloan, "Looking Ahead at AI and Work in 2026," 2026).

For companies operating under PE ownership or on aggressive growth timelines, the operating model question is especially acute. These organizations cannot afford multi-year transformation programs that consume capital without returning value. But they also cannot afford the alternative: layering AI tools onto broken processes and hoping for efficiency gains. BCG's data supports a focused approach: leading firms concentrate on 3.5 use cases on average and generate 2.1 times more ROI than lagging firms that spread resources across 6.1 (Source: BCG, "From Potential to Profit," 2025).

Leadership readiness: the multiplier nobody budgeted for

If AI amplifies organizational capability, then the quality of leadership determines whether that amplification creates value or accelerates dysfunction. This is not a theoretical concern. Multiple research streams now confirm that the single largest barrier to AI value creation is not technology, not data, and not budget. It is leadership.

McKinsey's 2025 findings reveal a telling blind spot: C-suite leaders are more than twice as likely to blame employee readiness for AI failures than to examine their own role. Meanwhile, employees report they are ready and willing to adopt AI, with the primary barrier being a lack of clear direction from above (Source: McKinsey, "Superagency in the Workplace," 2025). BCG's parallel survey found that only 25% of companies create significant value from AI, and the differentiator is not technology spend but leadership discipline: leaders who change core processes, upskill teams, and measure returns rigorously (Source: BCG, "From Potential to Profit," 2025).

EY's Work Reimagined survey of 15,000 employees across 29 countries found that companies are missing up to 40% of AI productivity gains due to gaps in talent strategy. The gap, EY concludes, is not technology but human readiness and organizational design (Source: EY, "Work Reimagined Survey," 2025). PwC's survey of 767 operations leaders reveals a striking self-deception: 85% say they are ahead of competitors in digital transformation, yet 89% admit their technology investments have not fully delivered expected results (Source: PwC, "Digital Trends in Operations Survey," 2026). Only 27% have fully embedded an AI strategy across business units.

Harvard Business Review identifies AI fluency, cross-industry network building, organizational redesign ability, and behavioral reskilling (critical thinking, evidence evaluation) as non-negotiable leadership competencies for the AI era (Source: HBR, "5 Critical Skills Leaders Need in the Age of AI," 2025). Gartner adds a cautionary dimension: by 2026, the firm predicts 50% of organizations will require "AI-free" skills assessments that test critical thinking and judgment without AI assistance, recognizing that over-reliance on AI may erode the very capabilities leaders need most (Source: Gartner, "Change Management Trends for CHROs," 2026).

Leadership readiness is not a line item on most AI business cases, but it may be the single factor with the highest return. KPMG reports that 87% of organizations now prioritize upskilling the current workforce, and 62% cite workforce skills gaps as the primary barrier to ROI (Source: KPMG, "AI Quarterly Pulse Survey," 2025). The organizations that close this gap first will compound their advantage; the ones that delay will find the gap widens as the technology accelerates.

The data readiness trap

"We need clean data before we can do AI" is one of the most common statements in enterprise technology conversations. It is also one of the most dangerous, because it is simultaneously true and paralyzing.

On one side: data quality is a real and measurable barrier. Gartner predicts that through 2026, organizations will abandon 60% of AI projects unsupported by AI-ready data. A Q3 2024 survey of 248 data management leaders found that 63% of organizations either lack or are unsure they have the right data management practices for AI (Source: Gartner, "Lack of AI-Ready Data Puts AI Projects at Risk," 2025). Organizations with successful AI initiatives invest up to four times more in data quality, governance, and change management, and those with the highest maturity of AI-ready capabilities achieve up to 65% greater business outcomes (Source: Gartner, "Successful AI Initiatives Invest 4x More in Data Foundations," 2026).

On the other side: waiting for perfect data is its own form of failure. BCG's data shows that leading firms focus on fewer use cases, suggesting that the productive approach is not cleaning all data everywhere but targeting governance efforts at specific, high-value applications (Source: BCG, "From Potential to Profit," 2025). McKinsey reports that 70% of high-performing organizations still face difficulties developing data governance processes and integrating data into AI models, indicating that even the best organizations do not wait for perfection (Source: McKinsey, "The State of AI in 2025," 2025).

MIT Sloan and HBR frame the problem differently, arguing that 92% of respondents say cultural and change management challenges, not data quality per se, are the primary barrier to becoming data-driven and AI-ready (Source: MIT Sloan Management Review, "Five Trends in AI and Data Science for 2025," 2024). The resolution is practical: start from a specific business problem that would meaningfully move revenue, margin, or operational efficiency. Identify what data that problem requires. Invest in cleaning and governing that data. Use the results to build the case for broader governance investment.

From proof of concept to production: crossing the valley

The pattern is now well documented: a team builds an AI prototype, leadership gets excited, and then the initiative stalls. The reasons are consistent across industries, geographies, and company sizes. Governance, change management, organizational readiness, and sustained executive sponsorship were never addressed.

BCG reports that the average organization scrapped 46% of proof-of-concepts before reaching production. Only 5% of companies in their survey create substantial value at scale (Source: BCG, "The Widening AI Value Gap," 2025). Gartner predicted that 30% of generative AI projects would be abandoned after proof-of-concept by the end of 2025, with average prototype-to-production timelines running eight months for projects that survive. Organizations with AI governance platforms are 3.4 times more likely to achieve high governance effectiveness (Source: Gartner, "Top Predictions for IT Organizations," 2025). Bain confirms the scale problem: only 22% of companies have moved beyond proof-of-concept, with the gap being organizational rather than technological (Source: Bain, "Technology Report 2025," 2025).

MIT Sloan's field research identifies executive sponsorship as a critical, and frequently missing, success factor. Each MVP sprint should be backed by a C-level sponsor who can remove organizational obstacles in real time. When end users do not understand why they should use or trust AI, the initiative fails regardless of technical merit (Source: MIT Sloan Management Review, "The Human Side of AI Adoption," 2025).

The POC-to-production gap is particularly expensive for PE-backed companies, where every stalled initiative represents capital deployed without return. The practical moves: time-box prototypes with clear success criteria and executive sponsors. Use storytelling and peer comparisons to break analysis paralysis. Build the governance scaffolding (roles, guardrails, measurement frameworks) alongside the prototype, not after it succeeds. And recognize that the first use case is not the end goal; it is the training ground for the organizational muscle that will power the second, third, and fourth.

The cost question nobody wants to ask

Beneath the execution challenges lies a question that most enterprise AI conversations avoid: are the current economics of AI sustainable? The answer, increasingly, is that current pricing may not reflect true costs.

Axios reports that OpenAI is projected to burn $14 billion in 2026, up from $8 to $9 billion in 2025. Current API pricing may need to increase three to ten times for sustainable economics (Source: Axios, "AI Models Costs IPO Pricing," 2026). DRAM prices rose 30% in Q4 2025, and GPU cloud costs increased between 40% and 300% during 2025. Hyperscalers plan $670 billion in AI infrastructure spending in 2026, but demand may be inflated by venture capital subsidies that keep prices below actual cost.

The parallel to ride-sharing economics is instructive. Early Uber and Lyft pricing was subsidized by investor capital to build market share and behavioral adoption. When prices normalized, users were locked in. Enterprise AI may follow a similar trajectory. The cost question is not a reason to delay AI adoption. It is a reason to be disciplined about where and how AI is deployed. Prioritize use cases where the return is so significant that it survives a substantial price increase. Avoid building critical dependencies on a single vendor's subsidized pricing. And include cost sensitivity analysis in every AI business case.

Where the enterprise AI conversation is heading

Taken together, these five themes describe a market that is moving through a predictable but painful maturation. The hype cycle has peaked. The organizations that invested early are now confronting the gap between what they expected and what they received. And the research consensus is converging on a clear message: the bottleneck is not the technology. It is the human system that surrounds it.

In the next 12 to 18 months, we expect three shifts. First, operating model redesign will move from the periphery to the center of AI strategy. Organizations that treated AI as a technology overlay will either restructure or fall behind. Second, leadership readiness will become a measurable, investable category. The firms that develop AI fluency programs, translator roles, and governance competencies will have a compounding advantage over those still debating which platform to buy. Third, cost discipline will separate the winners from the capital destroyers. The organizations that survive the pricing correction will be those that chose their use cases wisely, built governance that scales, and never confused subsidized adoption with sustainable value creation.

The J.Caresse point of view

At J.Caresse & Company, we work with PE-backed and high-growth enterprises that are building scalable, AI-ready infrastructure. Our perspective, informed by direct work with executive teams and the research presented here, is that the organizations most likely to win the next phase of enterprise AI are not the ones with the biggest technology budgets. They are the ones that invest in the human systems that make technology productive: clear leadership, disciplined governance, organizational designs that adapt rather than resist, and a commitment to building capability rather than buying tools. The technology will keep improving. The question is whether your organization is built to use it.

Sources

• Axios

  "AI Models Costs IPO Pricing." March 2026.

• Bain & Company

  "Technology Report 2025: Why AI Gains Are Stalling." 2025.

• BCG

  "From Potential to Profit: Closing the AI Impact Gap." January 2025.

• BCG

  "The Widening AI Value Gap." September 2025.

• Deloitte

  "State of AI in the Enterprise." 2026 Edition.

• Deloitte

  "The Great Rebuild: Architecting an AI-Native Tech Organization." 2026.

• EY

  "Work Reimagined Survey." November 2025.

• Gartner

  "Change Management Trends for CHROs in Age of AI." March 2026.

• Gartner

  "Lack of AI-Ready Data Puts AI Projects at Risk." February 2025.

• Gartner

  "Successful AI Initiatives Invest Up to 4x More in Data Foundations." April 2026.

• Gartner

  "Top Predictions for IT Organizations and Users in 2026 and Beyond." October 2025.

• HBR

  "5 Critical Skills Leaders Need in the Age of AI." October 2025.

• HBR

  "Most AI Initiatives Fail: This 5-Part Framework Can Help." November 2025.

• KPMG

  "AI Quarterly Pulse Survey, Q3 2025." 2025.

• McKinsey

  "Superagency in the Workplace." 2025.

• McKinsey

  "The Agentic Organization." 2025.

• McKinsey

  "The State of AI in 2025." March 2025.

• MIT Sloan Management Review

  "Five Trends in AI and Data Science for 2025." December 2024.

• MIT Sloan Management Review

  "Looking Ahead at AI and Work in 2026." 2026.

• MIT Sloan Management Review

  "The Human Side of AI Adoption." 2025.

• PwC

  "Digital Trends in Operations Survey." 2026.