Related question
How do we avoid creating a two-tier AI workforce?
Short answer
Give people role-appropriate literacy rather than concentrating capability in a small technical group. Not everyone needs the same depth. But everyone whose work is touched by an AI-enabled process should understand what is changing, which tools are approved, and what skills their role now requires.
This is a different problem from deciding who gets trained. It is what happens afterward, when the training was uneven and nobody planned for the consequence. One group pulls ahead. Everyone else keeps working the old way, and within a few months the gap is visible to the whole company, including to the people on the wrong side of it. What follows is not simply lost productivity. It is anxiety, resentment, and a quiet decision by capable people that this organization has already sorted them.
The fix is not universal advanced training, which would waste money and time. Training should match the role. Someone in accounts payable and someone in business development need genuinely different things. What they both need is to understand what is changing in their work, why, which tools are sanctioned, and what skills their job now asks for.
Managers are the group most often overlooked here, and they are the ones who will be redesigning work, evaluating performance under new conditions, and helping people adapt. Sending them into that without preparation is not a small oversight.
Restricting training to the technical team also cuts you off from your best source of implementation ideas. The people closest to the messiest workflows know exactly where the time goes, because they are the ones losing it.
Inclusive training is not an employee benefit. It is part of the implementation itself.