As of 13 August 2026, AI can detect when your production process is out of control.

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, power-user skill, and roughly 30 minutes until you can act on the result.

   How to actually do it

   1. Open the production measurement records and export them as a CSV or table with timestamps, batch or subgroup identifiers, measurement values and units.
   2. Gather the applicable specification limits, sampling method, subgroup size, normal operating conditions and a list of stoppages, maintenance, material changes and operator changes.
   3. Remove nothing silently: mark missing, duplicated, corrected or estimated readings, then paste the data and the process details into the prompt.
   4. Ask the chatbot to perform the data-quality checks and identify possible control signals using only the supplied data and explicitly stated rules.
   5. Compare each flagged record and calculation with the original production record, calibration or inspection source, and your documented control procedure.
   6. Ask a quality or process colleague to check the assumptions, investigate the flagged time periods on the line and decide whether production should continue, be held or be reviewed.
   7. Record the colleague's decision and evidence in your quality system, and do not use the AI output as a product-release or process-change approval by itself.

   Prompt

   Analyse the production-process data below for signs that the process may be out of statistical control.
   
   Process: [process name]
   Measured characteristic and units: [characteristic and units]
   Specification limits, if applicable: lower [value or none], upper [value or none]
   Sampling method and subgroup size: [method]
   Expected operating conditions: [brief description]
   Known changes, stoppages or unusual events: [details or none]
   Data, in time order with timestamps and batch or subgroup identifiers:
   [paste CSV or table]
   
   Use only the data and information supplied. First check for missing values, duplicate records, inconsistent units, changing sampling intervals and obvious data-entry errors. State the checks you could not perform.
   
   Then report:
   1. The exact records or subgroups that need attention.
   2. The evidence for any unusual point, sustained shift, trend, changing spread or other non-random pattern. Show the relevant calculations or comparisons in a reproducible table.
   3. Whether the evidence suggests a control issue, a specification issue, both, or neither. Do not treat a specification limit as a control limit and do not claim a root cause from this data alone.
   4. The assumptions and control rules used. If no rules or baseline are supplied, say so and give a provisional analysis rather than inventing them.
   5. A short list of checks for a quality or process engineer to carry out on the line, equipment, materials, operators and records.
   6. A clear conclusion using one of: possible out-of-control signal, no clear signal in the supplied data, or insufficient data.
   
   Do not approve product, release a batch, stop production or recommend a process change. Highlight every conclusion that requires a qualified human decision.

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