AI in manufacturing: root cause analysis, inspection instructions, safety training
Written with AI and checked against the sources by a second AI. As of November 16, 2026. More in the notes on AI.
In manufacturing, there is a lot of writing that hardly anyone enjoys: complaint reports, inspection instructions, safety training, fault reports. An AI can help organize, ask and phrase. Dimensions, tolerances and safety details, however, come from the drawing, the inspection plan and the safety data sheet, never from the AI.

Key points
- When troubleshooting, an AI supplies questions and hypotheses, not a diagnosis.
- For 8D reports and inspection instructions, the AI provides the form and you the facts.
- You write safety training and operating instructions only from the risk assessment and the safety data sheet.
- Drawings, formulations and customer data only belong in an AI your employer has approved.
- Analyses of individual employees need a clear rule and, if there is a works council, its consent.
Which tasks in manufacturing fit
| Task | What the AI needs from you | What you check |
|---|---|---|
| Collect possible causes of a defect | Defect pattern, data, what has already been checked | Whether the questions fit your machine |
| Prepare an 8D report | Your facts for every step | Cause, actions, dates |
| Write inspection instructions | Characteristics and values from the inspection plan | Every value against the drawing and the inspection plan |
| Prepare safety training | Risk assessment, operating instructions | Completeness, clarity |
| Phrase a fault report | Your bullet points, machine as a code | Whether maintenance has everything it needs |
| Write up an improvement suggestion | Idea, current state, expected benefit | Whether anything is exaggerated |
In the trial lesson you put your first instruction together and get feedback. No account, in a few minutes. Start the trial lesson
Looking for defects systematically: 5 Whys and Ishikawa
Konrad is a certified industrial foreman (Industriemeister) in the machining department of a supplier of hydraulic parts. For two weeks, a diameter on a series of turned parts has repeatedly been at the upper edge of the tolerance. He doesn’t use the AI to find the cause, but to make sure he doesn’t forget anything.
Two well-known tools are suited to this. With the 5 Whys, you ask “Why?” several times in a row until you reach a cause you can fix. The Ishikawa diagram sorts possible causes by area, often the “6 Ms”: man, machine, material, method, measurement, mother nature (the environment).
I’m looking for the cause of a dimensional deviation on a turned part. Defect pattern: [characteristic] has repeatedly been at the upper tolerance limit for [period], especially during the [shift/time of day].
Already checked: [list].
Create an Ishikawa outline by the 6 Ms with three to five questions each that I can check on site.
Ask questions, don’t make claims about the cause. Mark questions that are settled by what has already been checked.
At the end, suggest three measurements that will narrow down the areas fastest.The list points Konrad to two things he hadn’t thought of: the temperature in the shop in the afternoon and the change of coolant. Whether either of them is the cause is shown by his measurements, not by the AI.
Customer complaint: preparing the 8D report
Asha is a quality manager at a plastics plant. A customer has sent a complaint and asks for an 8D report, a report in eight steps from the team through cause and actions to recognizing the team’s work. Asha collects the facts with the team. The AI helps her present them clearly and completely.
Put my notes below into the form of an 8D report with the steps D1 to D8.
Only use my information. Where something is missing for a step, write “missing: …” and what exactly is missing.
In D4, separate clearly between a proven cause and a suspicion.
Customer, part number and names are replaced with codes; leave them that way.
Notes:
[notes]The “missing” list shows Asha where the report is still thin, for example the proof that the actions work. She doesn’t upload the customer’s drawings and inspection records; they belong to the customer and are often protected by contract.
Inspection instructions that work at the machine
Inspection instructions must be understood the same way by every person on every shift. An AI can turn the details in the inspection plan into a clear sequence of steps. It doesn’t fill in the values itself.
Write inspection instructions for operator self-inspection at [machine as a code].
Characteristics, measuring equipment, tolerances and frequency are in the table below. Copy them unchanged.
Per characteristic: what to check, with what, how often, what to do if it is “not OK”.
Short sentences, one step per line, understandable even for colleagues with little German.
Table:
[excerpt from the inspection plan]Before the instructions go to the machine, you compare every value with the drawing and the inspection plan. Then have a person from the shift read them and explain what they would do.
Safety training and operating instructions: only from the original sources
Under Section 12 of the German Occupational Safety and Health Act (Arbeitsschutzgesetz), employers must instruct employees on safety and health protection, tailored to the workplace and repeated regularly. For hazardous substances, Section 14 of the Hazardous Substances Ordinance (Gefahrstoffverordnung) requires written operating instructions (Betriebsanweisung) in a form and language that the employees understand.
Liv is a food technology specialist at a dairy. She is to prepare the safety training for a new cleaning agent. She gives the AI only the details from the safety data sheet and from the risk assessment that the company drew up with advice from the occupational safety specialist (Fachkraft für Arbeitssicherheit).
Turn the details below into slides for a ten-minute safety training session at the cleaning system.
Only use these details. Don’t add any protective measures, limit values or first-aid instructions.
If something is missing for an understandable training session, list it at the end instead of adding it.
At the end, five questions I can use to check whether everyone has understood.
Details from the safety data sheet and the risk assessment:
[text]The occupational safety specialist reviews the finished slides before Liv gives the training. She gives the training itself orally, based on the operating instructions, with questions and at the real workplace. Under Section 14(2) of the Hazardous Substances Ordinance, this happens before the work starts and at least once a year after that; content and date are recorded in writing and confirmed by the trainees with their signature.
Reporting faults so that maintenance can get started
Ivan is a machine and plant operator on a packaging line. His fault reports were often too brief, and maintenance had to ask follow-up questions. Now he has the AI turn his bullet points into a report and tell him what is missing.
Turn my bullet points into a fault report for maintenance.
Structure: system (code), time, what happened, error message word for word, what I have already tried, current state.
Don’t invent any causes.
At the end, ask me for the two details that would help maintenance most and that are still missing.
Bullet points:
[bullet points]Writing up an improvement suggestion convincingly
Mustafa is an industrial electronics technician and has an idea: an additional sensor on a conveyor belt would report jams earlier. The company has a form for improvement suggestions, but he finds it hard to write the idea up so that the committee understands it. The AI is supposed to ask him questions before it phrases anything.
I want to write an improvement suggestion for our company suggestion scheme.
Idea in bullet points: [idea].
First ask me at most five questions about the current state, effort and expected benefit.
Then write the suggestion in the sections current state, suggestion, effort, benefit, risks.
Only use my answers. Don’t estimate savings in euros if I don’t give any figures.The AI’s questions force Mustafa to get figures from the downtime statistics instead of estimating. That makes the suggestion more credible than any fine phrasing.
Checking before approval
- Compare every value, every tolerance and every inspection frequency against the drawing, inspection plan or data sheet.
- Safety details only from the safety data sheet and the risk assessment, reviewed by the responsible specialist.
- Search for “missing”, “?” and placeholders, and clear them all up.
- Have a person from the shift read the instructions and explain them.
- Replace codes with the real names only in the company system.
Drawings, formulations, performance data: what stays out
Under the German Trade Secrets Act (Gesetz zum Schutz von Geschäftsgeheimnissen, GeschGehG), a trade secret is only protected if the company takes appropriate confidentiality measures. That is why many companies specify exactly what may go into which AI. Without approval, these don’t belong in an AI chat:
- Drawings, CAD data and inspection records, especially those of customers
- Formulations, process parameters, purchase prices
- Complaints with customer names
- Output figures, defect rates or times of individual employees
The last point also has to do with co-determination: as a rule, a company with a works council (Betriebsrat), the elected employee representation, may only introduce programs that can analyze employees’ performance or behavior with its consent. Analyses per machine or per shift are something different from analyses per person, as long as no conclusions about individuals can be drawn from them. With small shifts, that may be exactly the case.
How to practice this
In production, what counts is that the AI only organizes and asks, and all values come from your documents. This division of labor is at the center of the learning app “KI können”, with examples from manufacturing, the trades and the office, carried out in your own AI. The trial lesson shows you in a few minutes how it works; you don’t need an account for it.
Sources
- German Occupational Safety and Health Act (Arbeitsschutzgesetz), Section 12 Instruction (in German) (opens in a new tab)
- German Hazardous Substances Ordinance (Gefahrstoffverordnung), Section 14 Information and instruction of employees (in German) (opens in a new tab)
- German Trade Secrets Act (Gesetz zum Schutz von Geschäftsgeheimnissen), Section 2 Definitions (in German) (opens in a new tab)
- German Works Constitution Act (Betriebsverfassungsgesetz), Section 87 Rights of co-determination (in German) (opens in a new tab)
Information as of November 16, 2026. Laws and offers can change; this article is not legal advice.
Read more
Try it now
The trial lesson with examples for Production manager. No account, a few minutes.
Start the trial lesson