Why 56% of CEOs See Zero ROI from AI — What the 12% Who Profit Do Differently (2026)
PwC asked 4,454 CEOs one question that should keep every executive in technology up at night: what has AI actually delivered for your business?
The results, published in January 2026 after surveying 95 countries, are not what the AI vendors promised. Fifty-six percent of those CEOs reported neither increased revenue nor decreased costs from their AI investments. Forty-two percent got one or the other — revenue growth without cost reduction, or cost savings without revenue growth. Only 12% reported both simultaneously.
The technology is the same for all of them. The results are not.
The explanation, according to PwC's analysis, is not that some AI is better than other AI. It is that 56% of CEOs are running AI investments as tactics, while the 12% who profit are running them as an operating system.
The Numbers — What the Full Distribution of AI ROI Actually Shows
PwC CEO Survey, January 2026 (4,454 CEOs, 95 countries):
- 56% — neither increased revenue nor decreased costs from AI
- 42% — one or the other (revenue growth only, OR cost savings only)
- 12% — both cost savings AND revenue growth simultaneously
This is not a bell curve with a disappointing mean. It is a bimodal distribution. There are two distinct groups: those who get compounding value from AI, and those who get isolated, non-compounding deployments that consume budget and produce nothing.
McKinsey corroborates from a different angle: 88% of organisations report using AI in at least one business function, but only 39% report any EBIT impact. Most of that 39% attribute less than 5% of their EBIT to AI.
S&P Global tracked AI initiative abandonment: it jumped from 17% in 2024 to 42% in 2025. Nearly half of all AI proof-of-concepts in 2025 were scrapped before ever reaching production.
Forrester puts the proportion of AI projects that reach sustained production use at 10–15%. The remaining 85–90% stall somewhere between pilot and production.
The data is consistent across sources. The interpretation from PwC is the most useful: the problem is not that AI does not work. The problem is how most organisations are running it.
The Root Cause — Why 56% Get Nothing
PwC's diagnosis is precise: most CEOs are treating AI as a set of tactics — individual projects, point solutions, technology procurement.
What "tactics" AI looks like in practice:
You bought 500 ChatGPT Plus subscriptions and handed them to the team. Productivity went up slightly for two weeks, then reverted. Nobody redesigned the workflows. The team is still doing the same work the same way, just with a slightly faster autocomplete.
You ran a pilot AI system for sales forecasting. It produces decent predictions. The sales team ignores it because the CRM integration was never finished and nobody is accountable for acting on the forecast.
You deployed an AI tool for customer service deflection. Resolution rate improved by 12%. The team that was supposed to handle the escalation queue was not resized. Customer satisfaction went up. Costs stayed flat. The AI created value that the organisation failed to capture.
This last example is the most important and the least discussed. The hidden zero ROI problem is not that AI fails to create value. It is that organisations fail to redesign the work around the AI — so the value created never shows up on the financial statements.
What the 12% Who Profit Do Differently
Practice 1: Enterprise-scale deployment, not pilot proliferation
The 12% run AI at scale across the enterprise — not in isolated pockets. One integrated AI deployment across customer service, operations, and finance creates compounding value. Ten isolated pilots each produce modest, non-compounding results. The question they ask is not "how many AI projects do we have?" It is "how integrated is our AI deployment?"
Practice 2: AI roadmaps owned by executive strategy, not IT
The 12% have AI roadmaps owned by the CEO and board. The roadmap is a business strategy document that happens to describe AI capabilities, not a technology deployment plan written in technical language. The CFO and COO are co-owners of the AI roadmap — not reviewers.
Practice 3: Data as a strategic asset, not an afterthought
The 12% have invested in accessible, high-quality, structured data before deploying AI at scale. The failure mode of the 56% is predictable: they deployed AI on top of messy, siloed, inaccessible data and got unreliable outputs that could not be trusted for business decisions.
Practice 4: Governance built before deployment, not after
The 12% have formalised governance and risk management processes before AI goes live — not in response to a problem. AI governance is not a compliance exercise. It is an operational necessity that, if done properly before deployment, becomes a competitive advantage.
Practice 5: Use cases tied to P&L outcomes, not technology metrics
The 12% start every AI initiative with one question: what P&L line will this move, and by how much? Every initiative has a named P&L owner, an explicit outcome commitment, and a measurement plan.
The Mid-Market AI Survival Problem
The enterprise versus mid-market gap in AI is not a tools gap. It is a governance gap.
Large enterprises can absorb a stalled AI initiative. A mid-market company with lean IT resources, constrained budgets, and limited leadership bandwidth is in a different position. When an AI project fails to make it from pilot to production, the cost is felt directly. Getting it right the first time is not a preference for mid-market companies. It is a business necessity.
The 22% median budget increase among enterprises raising AI spend in 2026 means capital for AI investment is available across the market. The constraint is not money. It is governance discipline.
The Framework — How to Move from 56% to 12%
Step 1: Audit your current AI portfolio
Before you add anything new, count what you have. How many isolated pilots versus integrated deployments? Which AI projects have explicit P&L outcome targets? If the answer to the P&L question is "we have not defined one," you already know which group you are in.
Step 2: Name the operating system
What does AI-driven operations look like for your specific business? Not "more efficient" — specific. Where does AI touch every major workflow? You cannot build something you cannot describe.
Step 3: Reboot the roadmap with CFO co-ownership
Every AI initiative in the roadmap needs a named P&L owner who has committed to a specific outcome. Governance is built before deployment, not after.
Step 4: Integrate before expanding
Stop adding pilots. Integrate the ones you have. One integrated deployment with compounding data flows is worth more than ten isolated ones.
Conclusion
Fifty-six percent of CEOs see zero ROI from AI because they are running AI as tactics — isolated projects, no integration, no governance, no P&L accountability. The 12% who get both cost savings and revenue growth are running AI as an operating system. The practices are not secret. They require executive discipline and structural patience.
The hidden zero ROI problem is that measurement is broken. AI creates efficiency that never shows up on the spreadsheet because nobody redesigned the work around it. The efficiency is real. The financial value is not being captured.
S&P Global's 17%→42% abandonment jump is happening in real time, across industries. If your organisation has run three or more failed pilots, you are in the pilot fatigue zone. The next pilot needs to be different structurally — not better tactically.
The question is not whether AI works. It does. The question is whether your organisation is running it in a way that lets you capture the value it creates.
Book a free 15-min call: https://calendly.com/agentcorps
Written by Vishal Singh. Builder of AI agent systems that replace repetitive workflows at scale. 10+ years building automation systems; founder of AgentCorps.