Part two: Where judgment shows up, and how to build it in

AI can generate endless content, but that’s not the hard part anymore. The real challenge is knowing what’s worth keeping, refining or rejecting. That’s where judgment comes in—and where most teams fall short.

Where does judgment show up?

Bad AI marketing isn’t a tooling problem. It’s a judgment problem. And judgment—good or bad—shows up everywhere:

• Audience resonance

• Brand voice

• Credibility

• Risk detection

  
  

It’s what allows you to distinguish signal from noise. To recognize a strong idea before it’s obvious. To say, “This technically works, but it’s not right.”

Judgment isn’t a single step. It exists across the entire workflow—from the initial idea to the final output.

A better way to work with AI

One of the most useful frameworks I’ve seen comes from Anthropic’s Description-Discernment loop, which is part of the 4Ds from their broader AI Fluency framework.

Describe → Discern → Refine → Repeat

It treats AI as a thinking partner, not a one-shot tool (which is exactly where most teams go wrong).

1) Describe

Before you ask AI to generate anything, define what success actually looks like. Not just what you want, but:

• How the task should be approached

• What 'good' looks like

• How AI should collaborate with you

Yes, prompting skills* matter here. But they’re only part of the equation. This step still relies heavily on your judgment.

2) Discern

This is the step too many teams skip. You have an output. Now what?

Based on Anthropic’s model, you can ask:

• Is this actually good?

• Does the reasoning hold up?

• Is the AI helping—or getting in the way?

That’s judgment in action.

If you skip this step, you’ve effectively handed over creative and editorial control.

You let AI replace you.

3) Refine

This is where your expertise shines.

• Give clear feedback on what works (and what doesn’t)

• Adjust the direction

• Add your perspective, context and audience understanding

• Make final decisions about what stays and what goes

And please, for the sake of audiences and stakeholders everywhere, don’t stop at the first output and say, “Looks good, let’s go.” That’s how you end up with content no one remembers.

AI and judgment: What this looks like in practice

Imagine a team generating 10 campaign ideas with AI. Without judgment, they might quickly pick one output and move on.

With judgment built in, the process looks different:

• Reject most ideas immediately

• Combine and reshape the strongest concepts

• Re-prompt with more context and intent

• Iterate until something actually resonates

Same tool. Same team. Completely different outcome.

The point of the process

This framework reinforces a simple truth: AI can generate. It cannot replace human judgment.

Describing can be taught. Prompting can be learned quickly. But discernment requires experience. And refinement requires ownership. The system only works if humans stay actively engaged in the work.

What comes next

Judgment needs to be present across the entire workflow, from idea to publishing to distribution. In the best examples, AI is an execution layer, not the source of the idea, nor the final edit.

The campaign isn’t showcasing the technology. It’s showcasing the decisions behind it.

So, the solution isn’t just more prompting skills.* It’s better judgment.

But building better judgment isn’t merely a workflow challenge. And it's not a quick fix. It’s an identity shift—for marketers, teams and how we define performance in the first place. And that shift has bigger implications than just better content.

Read part three: 'Judgment isn't a skill, it's a practice'

---

*Prompting is a valuable skill. No question, better prompts = better outputs. If you want to get some great prompting tips, I’d recommend this free guide from Tina Huang and Hubspot or go to YouTube and search “learn to prompt”. Here’s one I did on AI tools for writing.