The problem
Raw Workday exports carry real personal data: Social Security numbers, salaries, employee IDs, home addresses, dates of birth, direct manager names. That’s exactly the kind of data you can’t responsibly paste into Claude or any AI tool, which meant any AI-assisted workforce analysis started with someone manually stripping and re-checking every column by hand, a slow, error-prone step that discouraged people from using AI on real data at all. Waiting on a fully reviewed enterprise pipeline wasn’t realistic on the timeline the work actually needed.
What I built
A standalone HTML tool, People Data Scrubber, that runs entirely in the browser: no server, no upload, no external connection. You drop a Workday .xlsx export onto the page and it classifies every column into one of three buckets, Keep (safe for AI), Remove (PII or otherwise sensitive), or Flag (needs a human decision), covering fields like SSNs, salary, employee ID, email, manager name, and date of birth.
Every row also gets a one-way anonymized trace ID that carries no personal data on its own but lets HR map a row in the clean file back to the original employee using the reference file, so nothing gets permanently disconnected from its source. The tool outputs two files: an AI-Ready file with only the safe columns plus the trace ID, meant to be pasted directly into Claude or any other AI tool with zero PII, and a Primary reference file with three tabs (original data, clean data, and usage instructions) that stays with HR for traceability.
What happened
I use it every time I need to bring workforce data into an AI tool. During heavy reporting periods, it saves me 5+ hours a week that used to go into manually scrubbing and re-checking export columns by hand before I’d trust the data enough to work with it. The classification rules are still pending formal internal sign-off, so for now it’s a personal safeguard rather than an official process, but it’s already the tool I reach for by default before touching Workday data with AI.
What I’d do differently
The real limitation isn’t the scrubbing logic, it’s making sure the data controls behind it stay airtight. A tool like this is only as trustworthy as its column rules, and those need to keep pace as Workday’s export format changes or new fields get added, not just work correctly on the day you built it.