What AI literacy is at work once the login exists

The licence arrives at the start of the week. A few days later a client note is in a chat window, and the reply goes out because it sounds finished. Nobody in the chain has written down what a person is supposed to check before that reply leaves.

The OECD describes AI literacy as a set of competences. A person can critically evaluate AI technologies, communicate and collaborate with them and use AI as a tool online, at home and in the workplace. The report says most workers will only need that literacy to use a system and work with it. It treats programmes that build this literacy as mattering as much as programmes aimed at advanced AI skills.

What the competence covers

Literacy here is a general understanding, used while the work is happening. It takes in ethics, an awareness of risks and harms and the habit of judging a technology as you use it. A person who can produce a fluent paragraph and can't say why it should be thrown out has practised the use and skipped the evaluation. The OECD puts evaluation, collaboration and use in the one definition. A team that drills only the use has taught a thinner skill than the one the report describes.

So once the login exists, I'd judge literacy by one thing: can a colleague explain the check they ran before an answer left the team? That includes knowing where a person still decides and what material never goes into the tool.

Where public programmes have pointed

The OECD also surveyed of 21 OECD member countries. Fourteen had invested in publicly funded training that's specific to AI. Of those, nine focus on developing AI professionals, while seven aim to build AI literacy among the general public. The pattern is a useful mirror inside a company. A specialist track for a small expert group is easy to point at. The larger group is the people who already have the tool on the desk, and the definition was written with them in mind.

The report also notes that training supply may fall short of demand, especially for this general literacy, and that many countries lean on funding when other tools would reach a wider set of people. Career guidance, public and private collaboration and train-the-trainer programmes are the examples it names. A firm can copy the lesson without copying a national scheme. Put a practitioner who already does the task in front of colleagues, on that task, and have them show the check.

What to practise on a live task

1. Take one task the team already owns and mark where the model helped, where a person checked and where the output should never have left the building.
2. Write that check in the same place as the task, so evaluation sits in the work and doesn't live only in an induction deck.
3. Let a junior person reject an output in front of a senior person, and treat the rejection as the skill the literacy was meant to build.

If people can already reach a model and still can't tell a colleague why an answer should be discarded, start there. An AI training day is one place to practise the check on live work. The skill still has to show up in the week that follows, whether or not that day is booked.