I sit with people who are good at their jobs and watch a short routine. They open a chat tool, type the task in a handful of words, wait, and drop the first block of text into the document they owe. The document goes to a colleague. The tool is praised for speed, and the thinness of the page is treated as the person's fault.
The Nielsen Norman Group watched people work this way. Maria Rosala, writing on how AI literacy shapes use, separates two skills that do not grow on the same schedule. Prompt fluency is the ability to say what you are trying to do, what the constraints are and what context the tool needs. Output literacy is the habit of judging the answer: the gaps, the confident mistakes and the moment you should check a source. People with low prompt fluency wrote lines that looked like search queries. They left the context and the constraints out, because they were using the tool the way they use a search box.
The gap was visible in the tasks she recorded. One person asked which car to pick for a commute and named two models in a broken line. Another explained a family need, the constraints that mattered and asked for a table that compared them. A participant, while writing the longer version, said "I'm just trying to provide some context here. It seems like it often improves the results a little bit." That sentence is the skill. A licence does not install it.
What changes on a real task
The checking habit lags behind. The same study found people getting better at the ask and still accepting a polished answer without checking the parts that are easy to get wrong, such as a price or a set of instructions. A frequent user, only a few months in, could not say where the information had come from. Speed without that check is how a weak page leaves the building with a colleague's name on it.
AI skills training for employees has to put both skills on work the person already owes. They write the aim, the constraints and the context before they ask for a draft. They ask again, adding the fact the first answer missed, instead of wiping the thread and starting from an empty box. Then they mark what they would send and what they would verify. A tour of buttons leaves them where they started. A prompt sheet written for a generic email does the same, because the sheet is someone else's job.
The document you bring
Collective Campus builds high-performance teams for the AI world. Our AI training workshop runs that rep on the team's own work, with the reviewer in the room so the standard is visible. What a broader programme for staff should hold is in generative AI training for employees. What happens when the first week looks busy and the old pack returns is in your AI rollout stalls in week two.
Bring one document a person will send next week, and the name of the person who reviews it. If the session uses a made-up brief about a holiday, the skill stays in the room. Send me the task and the reviewer. We will shape the day so the next draft is one they can stand behind.
