Prompt reviewing

A consensus seems to be forming that the way we’re using AI now means that juniors won’t develop anymore because “AI can do the work”. Juniors are unable to develop and gain the experience they need to level up, do the work, make the mistakes, build the taste. Build judgment.

This isn’t just accumulated doing. It’s accumulated reflection on doing. And that is teachable, we’ve been doing that forever, and we can do that with AI as well.

What prompt reviewing is

A prompt review is a structured conversation between a senior and a junior (or between peers) where you go through the prompts used in a design task together. Not to evaluate whether the prompts were well-written. To evaluate whether the designer knew what to prompt, how to prompt, whether their assessment of the output was sound, and if the follow-up followed the right path. This isn’t different to a design review or critique.

A prompt is different in that it is unusually honest. It exposes reasoning and thought before the output buries it. Where a wireframe or design reveals a decision, a prompt reveals the thinking behind it: how the problem was framed, what was asked for, what was accepted, what was discarded, and why.

That’s the artifact. That’s where judgment lives. Or doesn’t.

Why it matters

The fear is that juniors using AI skip the steps and miss the feedback that builds judgment. Prompt reviewing doesn’t eliminate that risk but it addresses the reasoning that happens before the output, which is exactly where judgment can be taught. Judgment and experience develop through reflection, not just execution. A junior who reviews their own reasoning with someone experienced has a unique chance to reflect on that reasoning.

How to do it

Keep it small. This isn’t a formal review. It’s a conversation.

1. Ask the designer to walk you through a prompt chain from a recent task. Not the output but the prompts, thoughts, reasoning and the decisions along the way.
2. Follow the chain of prompts until an output was reached the designer was happy with, and reflect on how they got there and what other paths could have been explored.
3. Ask: why did you frame it this way? What were you trying to get?
4. Look at what they accepted. Ask: why was this good enough? What did you check it against? Did it align with your goal or did it surprise you?
5. Look at what was rejected. Ask: why didn’t this work? Was it frustrating that it produced something you didn’t expect? Would it have been faster doing this without AI?
6. Look at what they iterated on. Ask: what felt off from your original prompt? Did it divert, how?
7. Share your own read as calibration. Say: this is what I would have looked for. Here’s why.

The goal isn’t to improve the prompt. It’s to make the designer understand and reflect on their own reasoning and thought. This is what seniors have always done: sit with the junior, look at the work, ask why. The tools changed. The practice didn’t.