Why critical thinking beat AI literacy on the 2026 L&D priority list (and what that means for your curriculum)

For two years the calendar filled with tool tours. Where the button was. How to attach a file. Which model the company had paid for. The sessions were full. The work changed in pockets, usually where someone already had the judgment to throw a bad draft away.

Talk to learning leads who are rebuilding the catalogue this year and the request has moved. They still want people fluent with the tools. They worry more about people who cannot tell when a fluent answer is useless, risky or invented. Critical thinking sat on the list the whole time. They want it off the poster and into the course, beside the prompts.

Why the tour stopped being enough

A tour made sense when most of the company had never opened the product. A large share of knowledge workers have now followed a prompt recipe and produced something that looked finished. The scarce skill has moved. People can find the window. They struggle to decide what to do with what comes back, on a document the company will stand behind.

The tools also moved faster than the courses. A button tour from last year walks people through a menu the vendor has already redesigned. A habit of checking a claim against a source still works next year. The NIST AI Risk Management Framework is the reference I use for that check. Teams that built their identity on product features are tired of rewriting the same module every release. Teams that built the checking habit can keep the product clinic short.

New joiners still need the approved tools, the data line from AI governance, and a basic sense of where models fail. That belongs in onboarding. If you still announce it as the year's programme, the curriculum sits a year behind the problem the leads are describing.

What it looks like on a draft

In this curriculum the skill is a set of moves on live work. Frame the question before you ask a model, including what a useful answer would change. Separate what the source says from what the draft added. Name the assumption that, if wrong, would make the recommendation harmful. Decide, in writing, whether to use the output, change it or discard it.

You can look at a marked draft and see whether the person did them. A login tells you they opened the product. The marked draft tells you whether they checked the answer.

The year

Give the thinking work hours of its own. A short clinic for every new joiner, and for anyone who missed the first wave. Half a day. Approved tools, data rules, one workflow from their role, and one deliberate failure so they see a confident mistake. Refresh it when the tool changes. Leave the multi week academy for later.

Give the roles that send work outside the team a series on their documents. Across a month they mark a draft, they frame a problem before they generate options, and they defend a decision to discard an output in front of their manager. The critical thinking workshop is one of those reps. Keep the before and after samples, so the month has a page a manager can open.

Give managers a separate hour on how to coach the tool and the check, the work in the unpromptable stack. If they only know how to ask for more output, the thinking course becomes a pleasant day and the old standard wins on Monday, because Monday is when managers count the volume.

The pack

Completion of a literacy pathway makes a weak proxy once most people have finished it. Keep it as hygiene. Add the measures that match the new priority. How many work samples did you mark this month. How many outputs did you send back because a claim had no source. Can managers describe the check their team uses, without reading from a slide.

If the only number the executive team sees is seats and completions, they will ask for more literacy, because that is the number you taught them to want. Put a work sample in the pack. One page, a draft, the marks, the decision. A director can argue with that. A rank on a slide gives them nothing to argue with.

Literacy still belongs in the catalogue. Once people can already open the tool, I would stop leading with it. Pair every prompt exercise with a marked decision to use the output, change it or discard it. If the exercise ends when the draft looks finished, you have trained speed and called it judgment. A curriculum that only adds tools will keep losing to the next product release. The one that fits this year announces fewer tools and puts more marked paper in the room. The tools will keep arriving this quarter. The people who can hold an answer up to the job are the ones who stop the organisation getting faster while the work gets worse.