Observations Concerning Year-end AI Inventory
Looking back at a year of AI-assisted work surfaces a boundary between personal output and tool output that was not tracked at the time.
The year-end AI inventory is not an official task. Nobody assigns it. It happens when someone looks back at twelve months of documents, communications, and deliverables and tries to identify which portions involved AI assistance and which portions did not.
Year-end AI inventory adds one private manual check. The check is not complete. It is a sample review: a document from February, one from May, one from August. The quality is consistent across the sample. The method is less consistent. Some months the tool was used heavily. Some months it was not used at all. The boundary between those months is not marked anywhere in the documents themselves.
That check becomes part of the normal workflow.
Under the governing rule, year-end ai inventory becomes procedure once repetition feels safer than reconsidering it. By December 2023, a significant portion of knowledge workers had been using AI assistance for long enough to have a year's worth of output to review and not long enough to have developed a personal policy for what to disclose, what to check manually, and what to treat as straightforwardly their own work.
The help is useful enough to use and awkward enough to disguise.
Assisted work becomes routine while disclosure stays exceptional. This is the observation. The tools moved faster than the norms. Usage is common. Disclosure remains case-by-case, context-dependent, and often absent. What fills the gap is the private inventory: the year-end check that nobody requested and nobody will see, in which someone attempts to account for what they actually produced.
The observation concerns what the inventory reveals. Mostly it reveals that the boundary moved gradually enough that nobody noticed when it moved. The output looks the same. The method changed underneath it.