In the spring of 2026, PwC published its AI Performance study, based on conversations with 1,217 executives across 25 sectors. One finding stood out. Roughly three-quarters of all the value created with AI lands with just a fifth of companies. The other eighty percent do "something with AI", but are not yet truly reaping the rewards. And note this: the leaders perform 7.2 times better financially with AI than the rest. And that lead will grow, not shrink, in the years ahead. So says PwC's report. The outcome echoes an earlier study by MIT. A year ago, MIT arrived at almost the same ratio. In The GenAI Divide (2025), 95 percent of organisations saw no measurable return from generative AI; a small five percent did. Two independent studies, a year apart, with the same dividing line. But what do those 5% do differently?
What the 5% do differently
That is where the PwC study gets interesting. Because the answer is not about better models or bigger budgets.
One. They aim for growth. Not for cost-cutting. The strongest predictor of return was not efficiency. It was the ability to unlock new value and new markets with AI. Those who deploy AI only to cut costs fall behind. Those who pursue growth with it pull ahead.
Two. They let AI genuinely act. Most stay stuck at suggestions and summaries. Useful, but non-committal. The leaders deploy AI in a mature way almost twice as often: coherent tasks, within predefined boundaries, in systems that self-correct. The number of decisions taken without human intervention is considerably higher among them.
Three. And this is the most counterintuitive. It is precisely the organisations that automate the furthest that have their governance most in order. More often a framework for responsible AI. More often a governing body that keeps watch. And employees who trust the outcomes considerably more often. For them, governance is not a brake on autonomy. It is the precondition that makes autonomy possible.
MIT adds an observation that is logical in this light. Solutions built together with an external partner succeed roughly twice as often as what organisations develop entirely in-house.
From report to practice
Add up those points and a profile emerges. The companies that get a return have anchored AI in their core processes. They let agents work independently, under human oversight. They built governance into the architecture from the start. And they rarely did it entirely alone. That profile is exactly what Fyrm.ai is built for. With an Agentic Enterprise System we anchor AI agents in an organisation's processes and connect them to the existing software, under the direction of a lead agent that steers the whole. Security and compliance are woven into the architecture itself, not tacked on at the end, based on the ISO 42001 standard. And control stays with people.
And one more telling detail from the studies: many organisations in the 95% hesitate too long, stay stuck in doubt, perhaps even in denial, heads in the sand, that AI can have enormous impact. For those organisations it helps not to talk and persuade too much, nor to run yet another workshop, but simply to show it. Precisely Fyrm.ai's "See it. Scale it." approach.
In short. The direction is hard to misread. The distance between the organisations that move forward with AI and those that fall behind will grow in the years ahead. That gap will not close again. The question that matters: which side will your organisation be on?
Fyrm.ai. Orchestrated Intelligence. From Rotterdam. For all of Europe.
Sources: PwC, AI Performance study (2026), based on 1,217 executives across 25 sectors. MIT NANDA, The GenAI Divide: State of AI in Business (2025).
