Data labelling and annotation
Data labelling is the annotation labour and tooling that turns raw data into training signal, often the largest single line in an AI budget. Labelling directed at a qualifying development-phase build can be capitalised as a directly attributable cost; labelling during research, exploration or ongoing operation is expensed.
Routing by framework
Labelling directed at a specific development-phase build can be capitalised once the six criteria are met; labelling during research or exploration is expensed.
Labelling that is a direct cost of developing internal-use software in the application-development phase may be capitalised; exploratory labelling is expensed.
Labelling is a cost of preparing the asset, so it inherits the phase or stage of the build it serves. The same annotation team can produce capitalisable cost on a qualifying build and expensed cost on exploratory work in the same monthIAS 38§66.
The labelling pipeline as an asset in its own right
Where the entity builds a durable in-house annotation pipeline, that software can itself be internal-use software under ASC 350-40 or a development-phase intangible under IAS 38, separate from the labels it producesASC 350-40350-40-25. The pipeline build and the labelling output are assessed distinctly.
An entity spends on outsourced annotation across a quarter. Labelling for a scoped, funded, feasible model in development phase is capitalised; labelling used to test whether a second use case is viable at all is expensed as research. All figures are illustrative.
- S1IAS 38 Intangible Assets, IFRS Foundation (IFRS). https://www.ifrs.org/issued-standards/list-of-standards/ias-38-intangible-assets/
- S2Handbook: Software and website costs (ASC 350-40 internal-use software), KPMG (US GAAP). https://kpmg.com/us/en/frv/reference-library/2026/handbook-software-website-costs.html