Bulk Field Standardization Tool

The problem this kills

You inherited a CRM where the same thing is spelled forty different ways. "USA", "US", "U.S.", "United States", "united states ", and "Usa" are all sitting in the Country field. Industry says "SaaS", "Saas", "Software as a Service", and "software". Lead Source is a free-text disaster. Job titles are pure chaos.

The damage shows up the moment anyone runs a report. Your pipeline-by-industry chart splits one industry into six slivers. A filter for "United States" misses half your US accounts. Dashboards lie, and nobody trusts the numbers.

You could fix it by hand in a spreadsheet - find, replace, pray you didn't miss a trailing space - but that's hours of error-prone work, there's no record of what you changed, and you can't undo a bad replace once it's saved back. What you actually need is a reviewed, auditable, bulk find-and-replace that shows you exactly what it's about to do before it does it.

What you'll build

A small private web app - just for your team - that turns field cleanup into a safe, reviewed workflow:

1. Import a records export (or pull from a Google Sheet).
2. Detect every distinct value in each messy field, with a count of how many records use it.
3. Build a mapping of raw value to standard value - either by uploading a mapping table you already have, or by letting the tool propose mappings for you to review.
4. Preview exactly which rows change and how many, per field and per value.
5. Approve the plan as a human - nothing is standardized until you sign off, and ambiguous values get confirmed one by one.
6. Export a clean, standardized CSV in the exact columns your system expects, plus a mapping log that records every raw-to-standard decision.

Your original values are always kept alongside the standardized ones, so nothing is ever silently lost.

What's inside the Implementation Plan

The plan is a single markdown file you paste into Claude Code (a free AI coding assistant that writes and runs the code for you). You don't write code - you answer questions and approve what it builds.

It opens by interviewing you about your business. Before writing a line of code, the plan has the AI ask about your actual CRM, the exact fields that are messy, the naming and code conventions in your data, your typical and peak record volumes, your approval rules, and your weird edge cases. It reflects a short tailored spec back to you and waits for your thumbs-up - so the tool fits your data, not some generic template.

From there it walks the AI through every step: the database, login, the import screen, distinct-value detection, the mapping builder, the preview-and-counts screen, the approval gate, the CSV export, and the mapping log - each step ending with a ready-to-paste prompt.

The governance it includes (this is the point)

This isn't a throwaway script - it's built like an internal tool that an ops team can trust:

• Login so only your team can open it.
• Row-level security so each organization only ever sees its own data.
• A complete audit trail - who imported what, who approved which mapping, and when.
• A hard human-in-the-loop approval gate - the AI proposes mappings, a person reviews counts and row previews, and only an approved plan ever produces a standardized file. Mappings are never auto-applied.
• Duplicate guards - a record-id + field dedupe key so the same field can't be standardized twice in one run, and a versioned mapping so every run is traceable.
• Blanks and unknowns handled explicitly - nothing gets quietly bucketed.

Who it's for

CRM admins, RevOps, and sales-ops people who inherited fields with forty spellings of the same thing and need to clean them up safely - with proof of what changed, no risk of clobbering good data, and a process they can repeat every quarter.

You've got this - paste the first prompt and let the interview tailor the rest.