We want to connect an AI tool to automate our customer quoting, but our legacy ERP has duplicate customer accounts and ten years of obsolete SKU pricing. How do we establish a lean data cleanup standard before running AI on garbage inputs?

Question

Tyler Smith · Accepted Answer

Do not launch a multi-year digital transformation. Treating data hygiene as an all-encompassing IT project is a reliable way to burn capital without getting results. Instead, apply the EOS® 80/20 rule to your data architecture before hooking up any machine learning tools.

Define the vital twenty percent of data that directly powers your customer quoting workflow. This usually consists of active accounts from the past twenty-four months and active inventory or service SKUs. Quarantine everything else. Archive obsolete pricing tables, inactive legacy accounts, and discontinued product lines into a cold-storage backup folder that your AI model is explicitly blocked from indexing.

Next, establish three strict data hygiene rules for your quoting seat on the Accountability Chart:

- Every customer record must contain a standardized corporate domain, primary billing address, and verified tax status.
- Every active SKU must have an unbundled unit cost, minimum margin threshold, and assigned vendor ID.
- Any duplicate record detected by your database rules must be merged or tagged as inactive before an estimate can be finalized.

Assign your Integrator or operations lead to run a weekly spot check on newly created records during your Level 10 Meeting prep. Once your active catalog and current customer list are clean, your AI tools can generate accurate, reliable draft estimates without hallucinating obsolete material pricing.