Why 20% of companies capture 74% of AI Value

Most organisations are not short of AI activity.

They have pilots. They have tools. They have enthusiastic teams experimenting with copilots, chatbots, workflow automation and data use cases.

But activity is not the same as value.

PwC’s 2026 AI Performance Study shows a sharp and widening divide: just 20% of companies capture 74% of AI’s economic value. The same study found that the most AI-fit companies generate AI-driven revenues and efficiencies 7.2 times higher than the rest. The message is uncomfortable but clear: AI value is not evenly distributed. It goes to organisations that know how to turn AI into operating capability, not just experimentation.

That distinction matters.

The winners are not simply “doing more AI”. They are using AI for growth, business model reinvention and process redesign. They invest in the foundations that make AI scalable: strategy, data, technology, governance, workforce capability and innovation discipline.

The laggards are often busy too. They just remain busy in the wrong way.

The AI gap is becoming an execution gap

The most important insight from PwC’s study is not that AI creates value. We already know that.

The real insight is that most organisations are structurally underprepared to capture that value.

The difference is not the number of tools installed. It is the quality of the management system around AI.

AI leaders make different choices. They are more likely to connect AI to growth opportunities, sector convergence and business model change. They are more deliberate about where AI is applied. They redesign work around AI rather than placing AI on top of existing processes.

That is exactly where many organisations struggle.

They start with the tool.

Leaders ask: “Which AI platform should we use?”
Teams ask: “Can we automate this task?”
Vendors ask: “Can we run a pilot?”

Those are not bad questions, but they are not the first questions.

The better questions are:

• Which business outcomes are we trying to improve?
• Which decisions, workflows or customer journeys create the most value?
• Which processes are currently too slow, expensive, manual or inconsistent?
• Where can AI improve quality, speed, scalability or growth?
• What governance is needed to make this safe and repeatable?
• Who owns the process after the pilot?

Without those answers, AI remains an experiment.

AI should not be treated as a productivity layer

Many companies still treat AI mainly as a productivity tool.

That is understandable. Productivity gains are visible, easy to explain and politically attractive. If AI helps people write faster, summarise documents or automate repetitive tasks, the benefit feels immediate.

But PwC’s research suggests that the strongest performers go further. They use AI as a driver of growth and reinvention, not only as a cost-reduction mechanism.

That requires a different mindset.

AI should not only make existing work faster. It should force organisations to ask whether the work itself is still designed correctly.

If an approval process has twelve steps, adding AI to summarise the request may help. But the bigger question is whether the approval model itself makes sense.

If customer service relies on fragmented knowledge, AI search may help. But the bigger question is whether the knowledge model, ownership and feedback loop are properly designed.

If portfolio reporting takes weeks, AI can generate summaries. But the bigger question is whether the portfolio has clean data, clear ownership and a useful decision rhythm.

The most valuable AI implementations are rarely just tool implementations. They are process, governance and operating-model changes.

Trust is not a soft factor

One of the most striking findings in the PwC study is the role of trust.

Employees in leading companies are more likely to trust AI-generated insights and act on them. That trust is not created by telling people to “embrace AI”. It is built through capability development, clear boundaries, responsible AI frameworks and visible leadership.

This is where many AI programmes become too technical.

Trust requires answers to practical questions:

• When can AI output be used directly?
• When must a human review it?
• Who is accountable for the decision?
• What data is the model allowed to use?
• How are errors detected?
• How do we prevent bias, leakage or uncontrolled automation?
• How do we train people for their actual role, not with generic AI awareness training?

Governance does not slow AI down when it is designed well. It speeds AI up because people know what is allowed, what is expected and where the guardrails are.

That is the paradox many organisations miss.

Unclear governance creates hesitation. Practical governance creates confidence.

From pilots to portfolio

A major reason AI value gets stuck is that organisations manage AI as a collection of experiments instead of a portfolio of business initiatives.

That creates familiar problems:

• too many pilots
• unclear ownership
• duplicated effort
• weak business cases
• little reuse
• no scaling mechanism
• no link to strategic priorities
• no consistent measurement of value

The answer is not to stop experimenting. Experimentation is necessary.

But experimentation needs a portfolio structure.

An AI portfolio should make visible:

• which use cases are being explored
• which business outcomes they support
• who owns them
• what data and technology they depend on
• what risks and controls apply
• what value is expected
• which pilots should be stopped
• which pilots should be scaled
• which reusable components can be shared

This is where AI implementation and portfolio management meet.

Without portfolio control, AI becomes noise. With portfolio control, AI becomes a managed investment in business capability.

The real work: redesigning processes

PwC’s study also highlights that leading companies redesign work processes more often than others.

That is critical.

AI does not deliver sustainable value when it is bolted onto broken processes. It delivers value when processes are redesigned around what AI can now do, what humans should still own and where decisions can safely be accelerated.

A practical AI process redesign should define:

• the current process and pain points
• the target process
• where AI supports, recommends or acts
• where human judgement remains essential
• what data is required
• what controls are needed
• how exceptions are handled
• how performance is measured
• how the process will improve over time

This is not abstract transformation language. It is the operational work required to make AI useful.

For many organisations, this is also the missing capability.

They can buy the tool. They can run the pilot. But they struggle to redesign the work.

Better AI, not more AI

The lesson is simple: the next phase of AI will not be won by the organisations with the longest list of tools.

It will be won by organisations that can make better choices, redesign work, govern responsibly and scale what works.

That means AI leaders need to focus on five practical disciplines:

1. Connect AI to business outcomes — Start with growth, margin, risk, customer experience, speed or quality. Do not start with technology.
2. Manage AI as a portfolio — Create visibility across use cases, ownership, value, risks, dependencies and scaling decisions.
3. Redesign the process — Do not automate fragments of work without questioning the process itself.
4. Build trust through governance — Responsible AI, role-based training and clear decision rights are not compliance overhead. They are scaling conditions.
5. Reuse what works — Reusable components, shared data patterns, common controls and standard implementation playbooks reduce cost and increase speed.

What this means for leadership

The AI gap is not mainly a technology gap.

It is a leadership, governance and execution gap.

The companies capturing most of the value are not waiting for AI to mature. They are maturing their organisations around AI.

That is the real lesson.

AI value does not come from scattered pilots. It comes from controlled execution: choosing the right opportunities, redesigning the work, creating trust, governing the risks and scaling what works.

For many organisations, the next question should not be:

“How do we do more with AI?”

It should be:

“How do we become fit enough to capture value from AI?”

That is where the real competitive gap is opening.

Need help turning AI experiments into controlled execution?

Oosterwal Consultancy helps organisations move from AI pilots to practical, governed workflows with clear ownership, process redesign and measurable business value.

Through AI Process Implementation and ProcesAIsering, we help identify the right AI use cases, redesign the work around them and build the governance needed to scale responsibly.

Book a 30-minute discovery call to explore where AI can create real value in your organisation.

Sources

• PwC, 2026 AI Performance Study — just 20% of companies capture 74% of AI’s economic value, and the most AI-fit companies deliver AI-driven revenues and efficiencies 7.2 times higher than the rest.
• Consultancy.nl, Kleine groep bedrijven vangt driekwart van alle AI-opbrengsten — Dutch-language summary of the PwC study, published 13 May 2026.