AI literacy isn’t the ability to write a clever prompt; it’s a more basic and more durable skill, knowing roughly how these systems fail, so you know when to double-check an answer and when it’s safe to move on. Someone with it treats a confident-sounding output with appropriate suspicion for the stakes involved; someone without it either trusts everything or trusts nothing, and both are expensive mistakes at scale.
It’s unevenly distributed across most organisations right now, and that gap, not model quality, is usually what determines whether a rollout succeeds. The same tool in the hands of someone who can spot a hallucination and someone who can’t produces very different outcomes, even though the underlying AI is identical.
Building it isn’t a one-off training session. It’s closer to an ongoing capability, refreshed as the tools change, with the goal of getting everyone able to use AI well, not just the specialists who already understood how it worked.