AI adoption is up, training is not

Sixty-four per cent of workers used AI at work in the past year, up ten points on the year before, and daily generative AI use jumped from 14% to 22%. Those are the headline numbers from PwC's latest workforce survey of 49,000 people across 48 countries. It is not the whole story.

PwC splits the workforce into cohorts. The one that matters is the "engine room": 56% of all workers, the people who keep daily operations running, with limited AI proficiency and limited access to training. Only 40% of that group say they get the learning resources they need.

That trust gap is not a culture problem. It is a training problem with a trust symptom. Overall trust in management dropped seven percentage points year on year, and the group getting the least support is the group losing faith fastest.

The front-runners, 14% of the workforce, use generative AI daily at five times the rate of the engine room: 51% versus 11%. They are also the most likely to leave: 29% plan to change employer within a year. The cohort with the training is the one walking out the door.

BCG's 2026 study tells a similar story: 67% of regular AI users say the technology improved their enjoyment of work, but only 36% feel adequately trained, unchanged since 2025. In the UK specifically, 70% of employees experiment with AI at work, yet only 19% have taken any formal training.

The pattern holds at firm level too. 20% of companies capture 74% of AI-driven returns, and the differentiator is not which tools they bought. It is that they redesigned workflows and provided role-based learning.

I have seen this at the scale of a five-person team. Usage was high, confidence was high, but the ability to diagnose a broken step in our pipeline sat with one person. When that person was unavailable, the work stalled.

If you are rolling out AI tools without a structured plan for the majority of your workforce, you are building on the cohort least equipped to make the tools work and least inclined to trust you. The bottleneck is not the technology. It is the decision to train.