AI in Business: 74% of the Value Goes to 20% of Companies

74% of the Value, 20% of Companies

74% of AI's economic value goes to 20% of companies. The other 80%? They spend, they launch pilots, they build PowerPoint decks with "AI" on every slide. And they get nothing back.

The number comes from PwC, in a study published April 13, 2026, surveying 1,217 executives across 25 industries. Not a projection. Not an opinion poll. A measurement of actual gains (revenue and efficiency) attributed to AI, benchmarked against industry medians.

The result looks like the Pareto principle on steroids: leaders generate 7.2x more gains than average. And 56% of CEOs report no significant financial return on their AI investments.

Everyone "Does AI." Almost Nobody Extracts Value.

You might think this is a PwC quirk. It isn't. Four major consultancies published converging findings in 2025-2026:

• Roland Berger (203 executives): 90% of companies report returns below their investments. Only 10% capture meaningful value.
• BCG (1,803 C-suite): 75% of leaders rank AI in their top 3 priorities, but only 25% see real value.
• McKinsey: 88% of organizations use AI, but only 6% qualify as "high performers" with measurable bottom-line impact.

It's like an all-you-can-eat buffet where 80% of the guests load their plates but never sit down to eat. The activity is there. The digestion isn't.

The Eternal Pilot Trap

The most common pattern is the company that launches AI pilots. Lots of pilots. Dozens of pilots. But operationalizes none of them.

PwC identifies a figure that explains a lot: technology represents only 20% of an AI project's value. The remaining 80% depends on workflow redesign, governance, and reskilling. In other words, the AI works. It's the rest of the organization that doesn't keep up.

BCG adds a counter-intuitive insight: the worst performers spread their efforts across 6.1 use cases on average. Leaders focus on 3.5 and generate 2.1x more ROI. Do less, but in production rather than in demo mode.

The problem isn't access to technology. ChatGPT, Claude, Gemini: everyone can subscribe at the same price. The bottleneck is the ability to turn a tool into an operational process. No subscription covers that.

What the Top 20% Do Differently

Companies that capture value don't do "more AI." They point it somewhere else.

According to PwC, the number-one differentiator is industry convergence: using AI to enter adjacent markets, not just to automate existing operations. Leaders are 2.6x more likely to reinvent their business model through AI than to simply cut costs.

In practice, that means an insurance company using its data and AI to launch preventive health services. A retailer leveraging its logistics insights to sell supply chain consulting. AI, in this case, isn't a productivity tool. It's an expansion lever.

The other factor: leaders restructure entire workflows around AI (2x more likely according to PwC), instead of bolting an AI assistant onto an existing process. It's what Elon Musk said about Tesla: "We built a car around software; everyone else puts software in their car." Applied to AI, it's the exact same divide.

And companies that implemented responsible AI governance are 3x more likely to capture significant financial returns. Internal trust in the tool accelerates adoption. Without it, every project stays a "let's see" experiment.

When the Consultants Agree, Who Challenges Them?

Worth noting a piece of context. Four consultancies publishing studies that say "most companies are failing at AI" is also four consultancies selling AI advisory services. The diagnosis is probably right. The implied solution ("hire us") deserves a critical look.

That said, the convergence of numbers across different methodologies and samples is hard to dismiss. PwC, BCG, McKinsey, and Roland Berger don't arrive at the same conclusion by accident. The gap is real, documented, and widening.

PwC is explicit: without a change in approach, the gap between leaders and laggards will keep growing. Leaders learn faster, ship to production faster, automate decisions faster. The rest stay stuck in the pilot-PowerPoint-budget-pilot loop.

Three Questions Before Your Next AI Project

If you work at a company investing in AI, here's a quick test inspired by the PwC study:

1. Production or pilot? Are your AI projects running in production with measured KPIs, or stuck at the demo stage?
2. Costs or growth? Is AI aimed at cutting existing costs, or creating something new (new service, new market, new model)?
3. Tool or transformation? Is AI a tool added on top of existing processes, or have workflows been redesigned around it?

If the answers are "pilot," "costs," and "tool," the PwC study suggests you're in the 80%.

The good news: this isn't a budget or technology problem. It's a strategy problem. And that can be changed. What can't be changed is the time wasted on pilots that lead nowhere.