Inventory is work that has absorbed effort but has not yet delivered value. In software, it can be an unmerged branch, an open pull request, a feature waiting for review, or code complete but not deployed. These items are work in progress: they have been started, but users cannot yet validate them. section 6.0
Inventory also lives inside a codebase. Logic added before its purpose is understood becomes something reviewers must reason about and maintain. Every line carries costs in attention, build time, updates, and eventual removal. Faster code generation can increase those costs even when writing the code took little time. section 3.3.2 section 4.2
A pull request that looks complete
An AI-assisted order endpoint includes ID validation and rules for canceled or empty orders, although those behaviors were not requested. Producing the extra code is overproduction; the unanswered questions and logic that must now survive review, testing, and support are inventory. A plausible implementation does not establish that its invented rules are correct. section 3.3.2 section 3.5
To spot inventory, ask:
- What is in progress but not yet in production? Check branches, pull requests, reviews, and deployment queues; aging work can drift from requirements or need rework. section 6.0
- What code must be understood or maintained without providing value now? Make it easy to remove when its job ends. section 4.2
- Which decisions are being deferred? Do not assume waiting is always cheaper: delaying an authentication pattern, for example, can create retrofit work across every new endpoint. section 11.2.2
Use inventory as a diagnostic lens, not a verdict on whether a team is busy enough. The practical aim is less work stranded between steps and fewer commitments the system does not yet need. section 2.4 section 6.0

