MCCM Wiki
Redacting personal data before prompts
Redact names, IDs, and secrets before prompts so the model never sees what your policy forbids. If you cannot redact, do not run the workflow on live data.
Why this matters now
Teams stall on redacting personal data before prompts when nobody owns the next step. Name the owner before you pick a tool.
Practical steps
1) Write the job in one sentence. 2) Baseline the current time or error rate. 3) Design a reviewed workflow. 4) Run a 6–10 week pilot. 5) Keep or kill based on the KPI.
Risks to avoid
Do not expand scope mid-pilot, skip human review, or paste personal data into unapproved tools.
Related MCCM work
Pair this page with an AI audit, a clear pilot KPI, and GDPR-safe prompt rules.
Done looks like
The owner can show the KPI trend and the review gate without opening a slide deck.