ChatGPT and Codex training for engineering firms and suppliers in Stuttgart

Documentation takes engineering time

At an engineering firm or a supplier, engineers spend part of their week on documents: operating manuals, test reports, change notes and answers to customer questionnaires. Much of that text follows a structure your company has used for years. ChatGPT can draft the structure and the wording. The values, tolerances and safety notes must come from the engineer.

The developers at such a company work on data and product code: scripts that evaluate measurement data, tools that move product data between systems, a product configurator. Codex can take on much of this work. A wrong unit or a changed rounding rule can go unnoticed for a long time, so the training puts tests with known results in front of each change.

The ChatGPT track: technical documentation

• A project per product family that holds the terminology, the document templates and the released texts.
• Chapters of an operating manual drafted from a component list and the engineer’s notes.
• Test report summaries from a table of results.
• Change notes worded from the difference between two parts lists.
• Translations checked against your term base.

The Codex track: data and product code

• An AGENTS.md that states the units, the rounding rules and which data stays read-only.
• Sample data with known results, so each change can be checked.
• Python and SQL: Codex writes the evaluation code, a developer checks it against the sample.
• Configurator rules: Codex explains a rule before it changes it, and a test lists the product variants the change affects.
• No change to an interface between systems without a person’s approval.
• The rules for agents on AI agents under the same rules as humans.

Who drafts and who decides in technical documentation

The table shows the split the ChatGPT track teaches. Your engineers adapt it to their own document types.

Document	ChatGPT drafts	The engineer decides
Operating manual chapter	Structure, order of steps, wording	Values, warnings, safety notes
Test report	Summary of the results table	Pass or fail, deviations
Change note	Wording of the difference between two parts lists	Whether the change needs approval
Answer to a customer questionnaire	Draft answers from released data sheets	Each technical value
Translation of a manual	First draft in the target language	Terms against the term base
Maintenance plan	Layout from the maintenance data	Intervals and methods

Who I am

I am Michael Wutzke, an IT architect for data platforms in Frankfurt. I hold a technical diploma in mechanical engineering, and early in my career I worked for an engineering office in network technology. Later I programmed the databases of a global real estate database and put a near-real-time search in front of it. Today I build software with Codex and Claude for a client and teach at Claude Hacker House. I am also interested in open-source AI models that a company runs on its own servers. Details: Teaching and certifications.