The unpromptable stack: coaching, judgment, and collaboration skills L&D must protect in an AI-heavy workplace

You can watch someone get better at prompts in a single afternoon. The input gets tighter. The output gets less generic. They leave with a template they will reuse. Run that afternoon. Once the tools are ordinary, that afternoon covers the smallest part of what the team needs.

The rest refuses a better prompt. A manager has to help someone see their own work. A person has to decide that a fluent answer is still the wrong one. A group has to turn a pile of options into a decision they will own. Spend the year on prompt templates and let those thin out, and you get faster drafts and weaker decisions.

Coaching

A model can list feedback in a friendly tone. It cannot watch someone flinch, pick the one point that matters this week, and stay while they try again. That conversation is coaching, and you already have too little of it, because you promoted managers for being good at the work and you never taught them to teach it.

The tools make the gap expensive. A draft appears in seconds, so the temptation is to edit it yourself and move on. The person who brought the task learns that the job is to fetch a draft. Six months later the team can summon text, and the manager is still the only one who can tell a good draft from a bad one.

Protect a slot a tool cannot fill. The productivity workshop is where a manager defends that slot. One conversation a fortnight, on real work, with the model output in the document and the human decision circled. The manager's job is questions. What did you accept too quickly? What would the customer argue with? What will you do differently on the next one? The model can suggest the questions. The conversation still has to happen between the two people.

Judgment

Judgment means you refuse to confuse a complete looking answer with a right one. It sits on domain knowledge, on what the organisation has promised, and on the nerve to slow down. None of that lives in a prompt library. I point teams at the NIST AI Risk Management Framework when they want the human check written down.

Once a first draft is always available, people stop holding the problem. They hold the prompt. You still need people who can look at a recommendation and see which assumption is doing the work, especially where the work touches money, safety, customers or the law.

Assess it with a work sample, and skip the quiz on prompt structure. Here is a draft, here is the source, mark what you would hold back and say why. Do this where a miss is expensive. Promote people who are good at it. If you only recognise the people who produce the most output, you will train volume, and the checking will stop.

Collaboration

A shared chat with a bot in it can feel like collaboration. The work that commits a group is messier. People who see different parts of the problem have to line up, including the person who walked out unimpressed by the demo. A model can summarise the thread. It cannot carry the trust, the status or the tradeoff.

Hand that summary over too early and the group starts agreeing with a document none of them wrote. The meeting gets shorter. The commitment gets thinner. Keep a part of the work where people still think together before a tool tidies the language. Put the options on the wall in a design thinking workshop. Leave a disagreement visible. Name an owner. Write the tidy version after the choice.

The year

Write the year in two columns. One is tool fluency. Prompts, approved tools, data rules, a workflow or two. Keep it short and refresh it when the tools change. The other column is coaching, judgment and the decision a group still has to make. Make that column harder to cut.

Managers get a coaching rep on real assisted work, and you skip the slide about empathetic leadership. High stakes roles get a judgment sample every quarter, marked by someone who knows the domain. Cross functional work keeps a decision ritual that happens before anyone generates the summary. Prompt training is a clinic, on demand, and you leave it off the programme you announce at the town hall.

The critical thinking workshop belongs in that second column. People practise holding an answer up to the problem, which a template will never do for the person in the room.

One workflow with an owner

Let the tools take a first pass on summaries, sorting and empty drafts. Spend the hours you get back on the coaching conversation, on judging a live draft and on the argument a group still has to have. Hand those hours straight back to more volume and the old job gets faster while the part that needed a person gets thinner. Your job this year is to see which of those is happening, and to fund the work that still needs a person in the room.