Pattern 01 — Source before output

Thesis

Before generating an output, make the source visible and trustworthy.

Why it matters

AI produces outputs by recombining patterns from source material, training data, and prompts. When the source is hidden, review becomes a guessing game and mistakes become hard to fix.

What it protects

Accuracy, accountability, and the ability to correct errors without starting over.

Useful question

What is this output based on?

Practical example

A report draft includes a surprising statistic. Instead of trusting it, the workflow requires the source document to be attached and cited before the draft can move to review.

Reflection — workflow review

Where do we currently assume the source is known?
When was the last time we had to reverse-engineer where an output came from?
What would need to be true for source to travel with the output?

Next pattern → Made ≠ Ready