The Value Divide

THE BRIEF

Most organizations are investing in AI and not getting returns - and the data is specific enough now to explain why. PwC's 2026 Global CEO Survey, which covered 4,454 CEOs across 95 countries, found that 56% of CEOs reported no significant financial benefit from AI over the prior 12 months - neither increased revenue nor reduced costs. This is not a technology adoption problem. Adoption is near-universal. It is a value capture problem, and the mechanisms behind it are now well-documented enough that continuing to describe it as "early days" is a choice, not a diagnosis.

The value is going somewhere - just not to the majority. PwC's 2026 AI Performance Study found that the top 20% of AI-investing organizations captured approximately 74% of all AI-driven economic value generated. BCG's Widening AI Value Gap study, which surveyed approximately 1,250 senior executives across nine industries, found that only 5% of organizations qualify as "future-built" AI leaders - generating 1.7 times the revenue growth and 3.6 times the total shareholder return of their laggard counterparts. The gap is not statistical noise. It is compounding competitive advantage - and the leaders are reinvesting their gains into wider leads.

The most common ROI strategy isn't working. Gartner surveyed 350 global business executives at organizations with at least $1 billion in annual revenue that are actively piloting or deploying autonomous AI capabilities. Approximately 80% reported workforce reductions. Gartner found no correlation between those reductions and improved ROI. "Workforce reductions may create budget room, but they do not create return," said Helen Poitevin, Distinguished VP Analyst at Gartner. The organizations cutting people to demonstrate AI returns are getting neither. The organizations generating superior returns are, by the data, doing the opposite - investing in people who can guide and scale AI systems.

Most pilots are not becoming deployments. MIT's Project NANDA research found that approximately 95% of generative AI pilots deliver zero measurable P&L impact or fail to reach production. IDC's data adds structural context: for every 33 AI proofs-of-concept enterprise organizations initiate, only approximately 4 reach production. The failure rate is not a technology problem - model capabilities have not been the binding constraint for most enterprise applications since at least early 2025. The failure is organizational: pilots without production mandates, success metrics defined as adoption rather than outcomes, and AI programs owned by technology functions with no authority to redesign business processes.

The single biggest separator between AI leaders and everyone else is workflow redesign. McKinsey's State of AI 2025 survey found that AI high performers - organizations attributing more than 5% of enterprise EBIT to AI use - are 2.8 times more likely to fundamentally redesign end-to-end workflows around AI rather than deploying tools into existing processes. This distinction matters operationally: most AI implementations treat the process as fixed and the model as variable. Leaders invert that assumption. They treat the capability as fixed and rebuild the process around it. The productivity gains from the first approach are real but bounded. The gains from the second approach compound.

The competitive pressure is about to increase. BCG's data shows that AI agents currently account for approximately 17% of total AI value captured by leading organizations, and projects that share will reach 29% by 2028. Organizations currently in the leader category are allocating significantly more budget to agents and deploying them far more broadly than their laggard counterparts. Gartner's Q1 2026 research projected that 40% of enterprise applications will include task-specific AI agents by year-end 2026. The companies still working through first-generation pilots are not on a path to catching up to agent-deployed competitors. They are on a path to a more permanent structural disadvantage.

The CEO ownership variable is decisive. McKinsey's research shows that 72% of AI value leaders have CEO-level ownership of their AI agenda - nearly twice the rate of organizations at the prior measurement period. When AI is a technology program with a CTO sponsor, it generates technology results. When it is a business transformation with a CEO sponsor and a P&L mandate, it generates business results. The question for every board is not whether their organization is investing in AI. The question is whether the person accountable for the outcome has the authority to change anything that matters.

---

---

---

THE SIGNAL

The conventional narrative around enterprise AI is that everyone is progressing - adoption is up, capabilities are expanding, and the question is when the returns will materialize, not whether they will. The data tells a different story: a small group of organizations is winning significantly, a larger group is losing ground it does not yet realize it is losing, and the mechanisms separating the two groups are not primarily technological.

The tension is this: AI investment decisions are being made with an implicit assumption that the value curve is a matter of timing - that organizations investing now will capture value as the technology matures. BCG's research suggests the opposite. The characteristics that separate AI leaders from laggards in 2026 - CEO ownership, workflow redesign as the primary deployment model, sustained organizational investment in people alongside technology, measurement frameworks tied to business outcomes rather than AI activity - are not characteristics that emerge automatically as technology matures. They require deliberate organizational choices. Organizations that have not made those choices are not on a delayed version of the leaders' trajectory. They are on a different trajectory entirely.

Who is winning. The 5% of organizations BCG classifies as "future-built" leaders share a consistent profile. CEO-led AI strategy, not CTO-led technology adoption. End-to-end workflow redesign, not tool deployment into existing processes. AI investment treated as a business transformation budget line, not a technology capital expenditure. Measurement frameworks that connect AI activity to revenue, margin, and decision quality - not adoption dashboards counting users and interactions. The performance differential documented by BCG - in revenue growth, EBIT margin, and total shareholder return - is already visible to capital markets. Boards can see it.

Who is losing. The majority profile is also consistent. AI portfolios organized around pilot programs with undefined production pathways. AI strategy owned by a technology function or a "center of excellence" without budget authority or process redesign mandate. ROI pursued through headcount reduction - a strategy Gartner's May 2026 research found delivers budget room but no measurable return. Metrics defined as adoption rates, tool utilization, and efficiency gains rather than revenue attribution and margin improvement. These organizations have AI programs. They do not have AI-driven competitive advantages.

The gap is widening, not closing. BCG found that AI leaders plan to spend more than twice as much on AI as laggards in the coming year - and they are allocating a higher share of that investment to people, operating model change, and AI agents rather than to model access and infrastructure. The leaders are not satisfied with their position. They are actively investing to extend it. Meanwhile, the laggard group faces growing board pressure to either demonstrate returns from existing investments or reduce them - a dynamic that will compress the time available to close the gap.

The bottom-line connection is direct. BCG's TSR data for AI leaders is observed performance data, not a forward projection - it reflects organizations that already made the structural choices this report describes. At the capital markets level, the board already has a hypothesis about which side of the divide their company is on. The CEO needs to have the same hypothesis, and the evidence to confirm or challenge it.

The stakes: Gartner projects 40% of enterprise applications will include task-specific AI agents by end of 2026. Organizations currently running pilot portfolios without production mandates will be competing against agent-augmented operations in their core markets before year-end. The window to change organizational structure, ownership, and measurement discipline before that competition intensifies is measured in quarters, not years.

---

---

THE DEEP DIVE

Thesis: The AI value divide is structural, not cyclical - and closing it requires organizational decisions that most AI programs are not structured to make.

Most discussions of enterprise AI underperformance diagnose the problem as technological - insufficient data quality, inadequate model capability, immature tooling. McKinsey's research identifies a different cause. High performers are not succeeding because they have better access to models. They are succeeding because they are doing fundamentally different things with equivalent access. Specifically: they are redesigning workflows, assigning business-level accountability, and measuring outcomes rather than activities. These are organizational characteristics. They are not purchased with a model subscription.

The evidence layer

The convergence across independent research programs is notable. PwC surveying 4,454 CEOs globally, BCG surveying 1,250 senior executives across nine industries, and McKinsey running its annual global State of AI survey all arrive at the same structural finding: the distribution of AI value is highly concentrated, the concentration is driven by organizational behavior rather than technology access, and the gap between the top performers and the rest is growing.

PwC's April 2026 AI Performance Study introduces the concept of "AI-fit" organizations - those that have built the combination of governance, data infrastructure, enterprise-wide integration, and CEO-level commitment that translates AI investment into measurable financial results. AI-fit organizations were 2 to 3 times more likely than other organizations to report AI-driven gains in both revenue and costs. AI fitness is not primarily a function of investment level. It is a function of how that investment is organized and governed.

BCG's compounding dynamic finding deserves attention: leaders plan to spend more than twice as much as laggards in the next 12 months, and a higher proportion of that spend goes to organizational transformation - talent, operating model redesign, agent deployment - rather than model access and tooling. The investment gap is compounding the capability gap. Organizations that are laggards today will face a larger absolute gap next year even if they increase their own investment.

Practitioner signals

Three practitioner signals from the past two weeks illustrate the divide in concrete terms.

The first is Uber's AI cost experience, which surfaced in CFO commentary and circulated widely on X in late May. Uber's engineering team deployed AI coding tools at 84% adoption, with the tools generating approximately 70% of code output. The result: the entire 2026 AI budget was exhausted in four months. The CFO's question, as reported in the commentary, was whether token costs translate into customer value - a question that did not have a ready answer. This is the activity metrics trap in production: maximum adoption of a capability with no measurement framework connecting that capability to the business outcomes the CFO needs to see.

The second signal is the Box CEO's public warning about what he termed "AI psychosis" - the condition of overhyped expectations significantly outpacing real-world results at the organizational level. The signal is notable not because it is new, but because it is coming from a leader of a company with significant AI deployment. This is not a skeptic's warning. It is a practitioner's warning from inside the adoption curve, which makes it more credible and more concerning.

The third signal points in the opposite direction. TD Bank's deployment of AI in mortgage processing - reducing processing time from 15 hours to under 3 minutes - provides a clear example of what the leaders' playbook looks like in practice. The deployment targeted a specific, measurable workflow with a specific, measurable outcome. The business case was not "AI will improve efficiency." It was "this specific process takes 15 hours and should take 3 minutes." That specificity is what separates a business deployment from a technology pilot.

The Three Gaps framework

For CEOs and CFOs trying to assess their organization's position on the value divide, three diagnostic questions are more predictive than any technology assessment.

The Ownership Gap. Who is accountable for AI ROI? If the answer is the CTO, the outcome will be measured in technical metrics. If it is a center of excellence with advisory authority, the outcome will be measured in adoption. If it is a business executive with P&L accountability and board-level reporting obligations, the outcome will be measured in revenue, margin, and competitive position. McKinsey's finding that 72% of AI value leaders have CEO-level ownership is not a prescription - it is a data point about the authority level required to make the organizational changes that drive value capture.

The Workflow Gap. Are AI tools being deployed into existing workflows or are workflows being redesigned around AI capability? McKinsey's high-performer analysis identifies workflow redesign as the single most predictive variable separating organizations that capture value from those that don't. Deploying a tool into an existing process optimizes for the constraints of that process. Redesigning the process around the capability removes them. Most organizations are doing the first. The leaders are doing the second.

The Measurement Gap. Is AI success defined in terms of AI activity or business outcomes? AI activity metrics - user adoption rates, interaction volumes, efficiency percentages - are not business results. They are leading indicators that may or may not connect to results, depending on whether the activity is producing outcomes the business needs. Organizations with sophisticated measurement frameworks tying AI activity to revenue attribution, margin improvement, and decision quality are 21 times more likely to capture significant value, according to research cited in Larridin's 2026 State of Enterprise AI Report. That multiplier is not a measurement improvement. It is a value capture improvement driven by knowing what you are actually trying to accomplish.

Failure modes

The Headcount Proxy. Gartner's May 2026 finding is definitive: 80% of organizations deploying autonomous AI have made workforce reductions; none of it correlates with higher ROI. Companies cutting people to demonstrate AI returns are executing a strategy that the data says does not work. Worse, the human oversight capability being eliminated is the same capability required to guide and scale autonomous AI systems. Organizations executing this strategy may be destroying the organizational foundation required for the AI-driven performance they are cutting people to achieve.

The Pilot Portfolio Trap. A portfolio of AI pilots without production mandates is a budget commitment with an organizational escape hatch. Every pilot that does not have a defined production pathway, a business owner (not an IT owner), and a success metric tied to a business outcome is a commitment to indefinite exploration with no accountability for results. MIT's 95% failure-to-scale statistic is a measurement of organizations that built portfolios of these commitments.

The Technology Owner Problem. AI strategy assigned to the technology function generates technology results. Technology functions are good at deploying technology. They are not typically structured to redesign business processes, redistribute organizational authority, or hold business units accountable for outcomes. Those capabilities require business leadership with board-level mandate. Organizations that have not created that leadership structure have not created the conditions for the value capture they are investing to achieve.

The Activity Metrics Trap. An organization reporting high AI adoption across its workforce while unable to articulate the financial impact of that adoption has built an impressive dashboard and a weak business case. The Uber example is instructive: maximum adoption generated a budget crisis, not a ROI story. Measurement frameworks that start with business outcomes and work backward to the AI activity required to produce them are fundamentally different instruments from adoption dashboards. Most organizations have the latter. The value leaders have the former.

The consequence

The organizations currently capturing AI value will capture more of it. The reinvestment dynamic BCG identifies - leaders spending twice as much, allocating more to organizational transformation and agent deployment - means the gap will not close on its own as technology matures. The organizations not capturing value today face a narrowing window to make the organizational changes required to change that outcome. Those changes - CEO-level ownership, workflow redesign, outcome-based measurement, production mandates with business accountability - are not technology changes. They are leadership changes. The window to make them before agent-augmented competition arrives in core markets is measured in quarters.

---

---

THE PLAYBOOK

C-Suite

• Assign one executive - not a committee, not a center of excellence - as the owner of AI ROI, with explicit P&L accountability and a quarterly reporting obligation to the board. If that person is not you or your direct report, the organizational signal is that AI returns are a technology responsibility. That signal predicts the outcome.
• Require that every AI initiative in the current portfolio state its production criteria before Q3 budget review. A pilot with no production pathway is a cost center, not an investment. Budget renewals should reflect that distinction.
• Ask your CFO to produce the current financial impact of your AI investment in the next 60 days. Revenue attribution, cost reduction, margin contribution - not adoption rates, not efficiency percentages. If that number is not readily available, your measurement infrastructure is not adequate for the investments you are making or planning to make.

CMO / VP Marketing

• Map every AI tool currently deployed in your function to a revenue outcome - pipeline velocity, conversion rates, or customer acquisition cost. Any tool not connected to one of those three metrics within 30 days is operating as a pilot regardless of its deployment status. Treat it accordingly.
• Assess where your top three competitors are publicly demonstrating AI-driven capability in customer acquisition, personalization, or content production. The gap between their demonstrated capability and yours is a proxy for where your competitive position is heading, not where it is today.

Department Leads / AI Initiative Owners

• Build the production case for your highest-value AI initiative before the next budget cycle, not during it. Include the specific workflow being redesigned - not enhanced, redesigned - the measurement baseline, and the criteria for production readiness. Get sign-off from a business owner with P&L accountability, not just an IT owner with deployment authority.
• Identify one workflow in your domain where AI could reduce cycle time by 80% or more. Build the business case around that specific outcome. That case - with a number, a baseline, and a timeline - is the kind of proof the CFO can act on in a portfolio rationalization conversation.

---

---

THE NUMBERS

PwC 2026 Global CEO Survey, January 2026 | 4,454 CEOs, 95 countries

PwC 2026 Global CEO Survey, January 2026

PwC 2026 AI Performance Study, April 2026

BCG, Widening AI Value Gap, September 2025

BCG, Widening AI Value Gap, September 2025

BCG, Widening AI Value Gap, September 2025

Gartner, May 2026 | 350 executives at $1B+ revenue organizations

MIT Project NANDA, 2025

McKinsey State of AI 2025

Gartner Q1 2026

The top 20% of organizations investing in AI are capturing 74% of the value being created. The board can already see this in TSR differentials. The question for every executive in the room is whether their AI program is structured to be in that 20% - and whether they have the evidence to answer that question with confidence.

---

WHAT'S NEXT + WHAT'S COMING

The signal gaining traction across the most reliable enterprise AI channels this week - CFO and CIO threads on X, CIO Dive coverage, and Gartner analyst commentary - is the shift from AI adoption measurement to AI portfolio rationalization. CFOs are beginning to require stack rankings of AI initiatives by demonstrated ROI, not anticipated return. The companies that built large AI portfolios during the adoption phase of 2024 and early 2025 are entering a culling cycle: initiatives without documented production pathways and outcome-based metrics are being defunded. The organizations that built their portfolios without those anchors are now building them under budget pressure rather than before it - a significantly worse position from which to argue for continued investment. One specific thing to watch before next Tuesday: Microsoft Build 2026 is complete and enterprise coverage of the event is now circulating - watch specifically for enterprise CIO and CFO reaction to Microsoft's AI agent announcements and whether the conversation is about capability or cost governance. That reaction is a leading indicator of enterprise AI budget dynamics for the second half of the year.

M&A and developments to watch:

• Salesforce + Informatica integration (acquisition closed November 2025, $8 billion) - the combined data management and CRM platform is being positioned as the enterprise data foundation for agentic AI. Watch for the first integrated ROI case studies to surface at Dreamforce 2026. The signal to track: whether enterprise buyers are purchasing the integrated platform for AI readiness or simply consolidating existing licenses.
• Microsoft Finance agents (2026 Wave 1 enhancements rolling through September 2026) - Microsoft Copilot for Finance reached GA in October 2025 and is receiving significant enterprise capability additions this wave. The CFO is the direct target buyer. The market question is whether Microsoft can produce a measurable ROI story that survives CFO scrutiny before AI budget skepticism hardens into procurement policy.
• AI portfolio rationalization - Multiple enterprise technology analysts are forecasting that Q3 2026 will see significant AI initiative consolidation at large enterprises as Q2 earnings calls require CFOs to characterize AI ROI with more precision than prior cycles allowed. Watch earnings call transcripts for the language shift from "investing in AI" to "scaling AI" versus "evaluating AI performance."
• On the horizon: Q2 2026 earnings season begins in July - AI ROI commentary from tech-forward CEOs and CFOs will be the most significant forward signal on enterprise AI budget dynamics for the rest of the year. Gartner Data & Analytics Summit (June 2026) - early agenda signals a substantial focus on AI measurement frameworks and portfolio governance.

---

---

Research confidence notes for QA:

• PwC 2026 Global CEO Survey: HIGH - named survey, January 2026, 4,454 CEOs, 95 countries, press release at pwc.com
• PwC 2026 AI Performance Study: HIGH - named study, April 2026, pwc.com
• BCG Widening AI Value Gap: HIGH - named publication, September 2025, press release and publication at bcg.com
• Gartner May 2026 autonomous business layoffs finding: HIGH - named press release, May 5, 2026, gartner.com, Helen Poitevin attributed quote
• McKinsey State of AI 2025: HIGH - named annual survey publication, mckinsey.com
• MIT Project NANDA 95% figure: HIGH - covered by Fortune August 2025, named research project
• IDC 33 POC / 4 production statistic: MEDIUM - cited via secondary source (wizr.ai), recommend direct IDC verification before final publication
• Gartner 40% enterprise applications with agents projection: MEDIUM - cited in aggregated research summaries, recommend direct Gartner source verification
• Uber AI budget / Box CEO / TD Bank practitioner examples: SIGNAL - sourced from X, labeled as observational, not presented as verified facts
• Salesforce/Informatica acquisition: HIGH - Salesforce press release, November 2025
• Microsoft Copilot for Finance GA: HIGH - Microsoft blog, October 2025

---

This report was produced with AI assistance and human editorial review.