Notes on product lifecycle management
Writing about the parts of PLM that are hard for structural reasons rather than technical ones — configuration, the ERP seam, and what data quality actually costs.
- AI in engineering
What an AI agent should not be allowed to do
The interesting design question for AI in a PLM is not what the agents can do. It is what they are structurally prevented from doing, and why each limit is worth its cost.
- Data quality
The duplicate item nobody created on purpose
Duplicate identity is the most expensive data condition a PLM can hold, and it is invisible by construction: both records are correct, nothing is malformed, and no validation fires.
- PLM and ERP
Who owns the part number?
Almost every failed PLM implementation fails at the same seam, and it is not a technical one. It is that nobody wrote down which system is the master before both started creating records.
- Configuration management
Approved is not effective
A change has three dates and most systems record one. The gap between approval and effectivity is where nine weeks of obsolete material gets ordered by people acting in good faith.