Out-of-Tolerance Impact & Recall Assessment: Decide Fast When an Instrument Fails Cal

The problem this kills

A micrometer comes back from calibration reading 0.04 mm high. A pressure gauge is found out of tolerance at its annual cal. A scale drifted. The instant that happens, a clock starts and a hard question lands on the quality desk: what did we accept based on that instrument since the last time we knew it was good?

Right now that question gets answered with a frantic, manual reconstruction. Someone digs out the last calibration certificate to find the "last known good" date, then trawls inspection records, travelers, and usage logs to figure out which lots and products were measured with that gauge in the affected window. The list is assembled in a spreadsheet under pressure, the impact is argued in a hallway, and the disposition — leave it, re-inspect, hold, or recall — gets made fast and documented poorly. Auditors hate that. So does the customer who gets a recall notice three weeks late. You don't need to be a developer to do this properly.

What you'll build

A simple internal web tool. When an instrument is found out of tolerance, you log the OOT event: the instrument, its last good calibration date, the discovery date, and how far out it was. The tool scopes the affected window (last-good-cal → discovery), pulls every inspection record or usage entry for that instrument inside the window, and rolls it up into the products, lots, and measurements at risk. The AI drafts an impact assessment — what was checked, what the out-of-tolerance amount could mean for those acceptance decisions, and a recommended disposition. Your quality manager opens the assessment, reviews the scope, and chooses the disposition: no impact, re-inspect, hold, or recall. Only after approval does the tool trigger the chosen action, email the right people via Resend, and export a clean impact assessment record for the quality file. When product action is needed, it links straight to your NCR or recall process.

What's inside the Implementation Plan

The downloadable plan is a step-by-step file you paste into an AI coding agent. It opens by interviewing you about your business — how you track last-good-cal dates today, where your usage and inspection records live, exactly how instruments and lots are identified in your data (gauge IDs, lot/serial conventions, work-order numbers), your typical and peak measurement volumes, the rules you use to judge impact, and the messy edge cases like an instrument shared across lines or one with no usage log at all — and then it tailors the data model, the impact logic, and every later step to your answers. This is not a generic template; the agent reflects a short spec back to you and waits for your thumbs-up before it builds anything. From there it walks the agent through logging the OOT event, scoping the window, gathering the affected lots, drafting the assessment, the manager's review-and-approve gate, triggering the disposition, the Resend notices, and the export — each step with a ready-to-copy prompt. There's also a fallback so you can build and run the whole thing today from a CSV or Google Sheet, even with no connection to your calibration system or MES.

The governance it includes (this is the point)

This is regulated quality work, so it ships with the controls an auditor expects to see: login so only your quality team can use it, row-level security so you only ever see your own site's instruments and lots, a complete audit trail of who logged the event, who assessed it, who approved the disposition and when, and the rationale they recorded. The disposition is a hard human-in-the-loop gate — the AI drafts the impact assessment, but no product is held, re-inspected, or recalled until your quality manager reviews the scope and approves. Duplicate guards keyed on instrument + OOT event mean the same failure can't be opened twice, and the assessment rationale is kept permanently so the decision is defensible long after the dust settles.

Who it's for

Quality and metrology teams — QA managers, calibration coordinators, and quality engineers — who have to move the moment a gauge fails calibration and must show an auditor a clean, scoped, approved impact assessment. If you can describe how you find "last known good" and which records tell you what a gauge measured, you can build this.

You've got this — start with the plan, paste the first prompt, answer the interview, and you'll see your first affected-lot list take shape the same afternoon.