As of 13 August 2026, AI can only partly classify defects in your products.

What to actually do

1. Hand it to a person

   The route this page recommends

   A person who owns the outcome does this end to end, worth it when the failure is dear.

2. Use a tool built for this

   Second choice

3. Do it yourself

   The distant third

   A chat interface, chat-fluent skill, and roughly 1 hour until you can act on the result.

   How to actually do it

   1. Open your current quality standard, defect taxonomy and product specification, then copy the approved defect categories, definitions, tolerances and pass or fail rules into one document.
   2. Gather representative photographs and inspection notes for each item, including the item ID, product variant, defect location, measurements and the inspection conditions.
   3. Add examples of accepted products and known defects to the document, keeping the item IDs and defect labels attached to each example.
   4. Paste the taxonomy, criteria and inspection evidence into the prompt, and ask the chatbot to mark missing evidence as uncertain rather than guessing.
   5. Export or copy the resulting table into your quality-control record, then compare every pass or fail result and defect label against the original specification.
   6. Ask a trained quality colleague to inspect every low-confidence, uncertain or disputed item and to approve the final release, rework or quarantine action.
   7. Use the colleague-approved classifications to correct the taxonomy or examples before processing the next batch.

   Prompt

   You are assisting with quality control for [product or product family]. Classify each product defect using only this approved defect taxonomy:
   
   [PASTE DEFECT TAXONOMY, DEFINITIONS AND ACCEPTANCE CRITERIA]
   
   Use these product requirements and inspection rules:
   
   [PASTE SPECIFICATION, TOLERANCES, PHOTOS OR INSPECTION STANDARD]
   
   I will provide inspection evidence below. For each item, return a table with: item ID, observed evidence, most likely defect category, pass or fail against the supplied criteria, confidence as high, medium or low, and the exact evidence supporting the classification. Do not invent a defect category, measurement, cause or missing observation. If the evidence is insufficient or the defect is not covered by the taxonomy, write "uncertain" and state exactly what additional photograph, measurement or inspection is needed. Do not make a final release, quarantine, safety or customer-notification decision. Separate visible observations from inferences, and flag any item that needs a trained quality inspector to decide.
   
   Inspection evidence:
   [PASTE ITEM IDs, PHOTOS, MEASUREMENTS AND NOTES HERE]

   Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.