CASE LOG · 02

Case study · AI content evaluation

SalesScore

A custom AI tool that scores every content submission against a fixed rubric and returns feedback automatically — hands-off, with none of the cost or drift of a human reviewer.

The brief

Content was being judged by gut. Feedback was inconsistent from one piece to the next, and reviewing it ate a real chunk of every month. The need wasn't more content — it was a reliable, repeatable way to evaluate what was already being made.

SalesScore answers that with structure. Every submission is measured against the same fixed rubric, so the system tells you exactly what's missing instead of offering a vague "make it better." The hard part wasn't generating an opinion — it was getting the AI to apply the identical rubric, with the same depth and specificity, on every single run.

That reliability was solved through deliberate prompt engineering rather than basic prompting, and it has held consistent ever since.

Type Custom AI content-evaluation tool

The job Score submissions against a fixed rubric and return feedback automatically

What it does

Rubric scoring Automatic feedback Prompt engineering

THE PROBLEM

The challenge

Making content feedback consistent, not subjective.

"Most feedback on content is vague. 'Make it better' isn't a system."

The core challenge was never generating content — it was evaluating it reliably. Feedback had to be structured instead of subjective, and consistent across very different inputs. Outputs had to be actionable, not just descriptive.

Early versions struggled with the natural variability in AI responses, which made the evaluation hard to trust. Getting the model to apply the same rubric — with the same structure, depth and specificity — on every single run is what moved this from a prompt into a real system.

What it does

A scoring engine, not an opinion.

checklist

Fixed-rubric scoring

Every submission is measured against the same defined rubric, so evaluation moves from gut-feel to a repeatable, structured judgement — and points to exactly what's missing.

tune

Consistent by design

Deliberate prompt engineering — criteria, output format and scoring logic — keeps the system producing stable, trustworthy results on every run. The headline outcome: 100% rubric consistency.

bolt

Automatic feedback

Instead of generic suggestions, each evaluation returns specific feedback tied directly to the rubric — returned automatically, so it's usable rather than just informative.

person_off

No reviewer to hire

A human reviewer would have added cost and reintroduced the very inconsistency the tool removes. SalesScore runs hands-off, with none of that drift.

The outcome

What it changed.

100%Rubric consistency, every run

5–10 hrSaved every month

0Human reviewers needed — no cost, no drift

System insight

SalesScore isn't just a content tool — it's a decision framework.

subjective content review→structured evaluation

inconsistent feedback→repeatable scoring

guesswork→clear next steps

By systemising how content gets judged, it lets people spend their time improving the work rather than second-guessing it.

Next case study

DuaComposer

An AI app shipped fast and validated faster — idea to revenue in three weeks.

View case study →

Diagnostic note · 03

"The goal was to validate quickly, not build a cathedral. I shipped a focused MVP with AI at the core and cut everything that wasn't the one thing users needed."

1.6K+Downloads

29Paying subscribers

Judging
by gut?

If a repetitive judgement call is eating hours every month, it might be a custom build. Take the 2-minute diagnostic — I'll give you the honest move, even if it's "not yet."

Start the diagnostic → Just email me