AI
How does TimeCampus decide which project an hour belongs to, and can I tell when a model guessed?
TimeCampus uses a two-stage attribution workflow to assign captured activity to projects: 1. Deterministic Keyword Classification: When an event is captured, TimeCampus…
About this answer — Produced by our own assistant on 23 September 2026 from our published documentation. The pages it drew on are linked under the answer. It is generated text, not a statement from a person — where it and a linked page disagree, the page is right.
Answer
TimeCampus[1] uses a two-stage attribution workflow to assign captured activity to projects:
- Deterministic Keyword Classification: When an event is captured, TimeCampus first scores it against your project names and defined terms using a deterministic keyword classifier.
- Optional LLM Refinement: For more complex or ambiguous activity, the system can optionally refine the match using a large language model.
Spotting Model Guesses and Suggestions
You can clearly identify when a model guessed:
- Explicit Method and Model Labels: Every project suggestion is stored with metadata showing the exact method and model that generated it, so you always know whether it came from a keyword rule or an AI model.
- No Automatic Timesheet Changes: Neither the keyword classifier nor the LLM is permitted to alter your timesheet on its own. Every match is presented purely as a suggestion and only becomes an actual project record when you explicitly accept it via timeline controls.
- Auditable Records: All tracked and suggested hours can be reviewed, edited, or recategorized through the Timer and Time Entries[2] interface before reaching reports or invoices.
To learn more about how suggestions and attribution guardrails work, visit the TimeCampus AI project attribution use case[3].
Sources
The pages this answer was drawn from.