Everyone is arguing about AI proficiency, and almost everyone means prompting.
I keep a proficiency rubric for this. Five skills. Prompting is one of them, and it’s the only one that’s mechanical. The other four are judgment. Can you check a claim against a source and catch a confident, fabricated fact… and say what you actually verified? Can you tell when a task fits AI and when to keep it out, and explain why? Do you recognize sensitive data in context? Will you own the final product, or hand it off with a shrug?
Four of the five rows have nothing to do with writing a better prompt.
That’s the confusion. Prompting is teachable in an afternoon. Judgment takes years, and it’s the part that’s actually scarce.
Look at what’s happening in software. The demand for junior developers is thinning. The demand for senior architects isn’t. That’s not a story about who prompts better… juniors are often faster with the tools. It’s a story about who can tell whether the output is right, whether the problem was framed correctly, and whether the thing should have been built at all.
Critical thinking isn’t a nice complement to using AI. It’s the skill using AI is made of.
Monday step: pull your team’s AI training and count the minutes. How many teach prompting, how many teach evaluating output? If it’s mostly the first, you’re training the part that was already easy.
Source: RationalAI Method / AI Proficiency Rubric (own IP — no external source)
RationalAI helps organizations turn ideas like this into results: AI strategy and advisory services. Previous article: Agentic Development Rediscovers Classic Software Engineering
