Hi Reader,
​
“But my work is judgment work. You can’t measure that.”

I’ve heard it a hundred times, and I understand why people believe it — the only way you’ve ever experienced AI is without a standard, without a review, without a number. Just vibes. So measuring it sounds impossible, or like pretending.

Here’s the thing. I spent 35 years measuring things people swore couldn’t be cleanly measured — risk, exposure, the probability a number was wrong — at two of North America’s largest banks. Judgment-heavy, high-stakes, “you just have to feel it” work. It can be measured. Not perfectly. Usefully. Enough to see a trend.

A standard, a score, a cadence

Three moving parts, none of them hard. Full method here: curiochat.ai/blog/how-to-measure-ai-output-quality​

1. A standard. Two sentences for what good looks like on one task. “A good client email is warm, under 120 words, no bullet lists, ends with one clear next step.” You just made it measurable.
2. A score. Rate each output against it, 1–5. Same rater, same standard, every week. Consistent, not precise.
3. A cadence. Weekly. One score, one task. Now you have a line on a graph instead of a feeling.

A bank doesn’t measure risk to four decimals of truth — it measures usefully, consistently, enough to act on the trend. A rough-but-consistent 1–5 you actually track beats a perfect metric you never build.

You don’t have to take it from me — the point ran through this week’s AI news too: measuring AI by what it actually delivers is the boring discipline that turns “I use a lot of AI” into a number you can act on. Roundup here: curiochat.ai/blog/this-week-in-ai-2026-07-19​

Why measurement is the thing that builds trust

Here’s the calculus most people miss: you don’t trust what you can’t see. The reason you re-check every AI output by hand is that its quality is invisible to you — no number, no trend, no visibility. The moment quality is measured, two things happen at once: drift can’t hide (you catch the dip the week it starts), and trust stops being a hope and becomes something you can see. Visibility is what converts “I think this is fine” into “the score went 3.1 → 4.2 over six weeks.” One is a vibe. The other is evidence.

That’s not a nice-to-have. It’s the first move of an engineering-grade system, because if you can’t measure it, you can’t improve it — and you can’t trust it either.

Your Pattern Tweak — your first standard

Quick read: Architect — spec the structure. Surfer — score the vibe task. Keeper — write the bar. Pilot — score the fastest-shipped output. (Want to discover your Type? Reply “quiz”.)

Try-This-Now (≤5 minutes)

1. Pick the single task you hand to AI most often.
2. Write two sentences: what does good look like for it?
3. Score this week’s output, 1–5. Put the number and date where you’ll see it next week.

Next week, score again. That’s the first measurement of AI quality you’ve ever taken — the first brick of an engineering-grade system, built in five minutes for free. Stop — this counts.

Excelsior,

Pierre/
​Founder, Curio Chat Academy

P.S.: A growing slice of the Pack Library is measurement skills — the scoring, the weekly trend, the standard-keeping — built so the number gets taken consistently without depending on your discipline. The two-sentence standard above is the free, by-hand version of the same idea.