AI suggests, the user decides. TimeCamp’s new application

Project context

We have worked with TimeCamp since 2024. One area is a desktop application that uses AI to propose time entries based on computer activity.

The starting point is an ordinary working day: moving between applications, documents and tasks. We build a way to organise these signals into suggestions a person can review and correct. Development covers AI logic, background operation and the interface.

The challenge

Using an application or document does not always identify the project. Separate activity signals need to become a suggestion the user can assess and correct. An inaccurate suggestion should not automatically enter a report.

The application also needs to keep running in the background. Activity capture, operating system access and the presentation of results must fit everyday work and preserve the user’s control over time records.

Scope of work and solution

We develop the application in Electron, with a React and TypeScript interface. Background activity capture is connected to the suggested entries and their approval flow. Users can review and correct each suggestion before saving it.

We build the AI logic in Python. It uses activity signals to categorise work and suggest how time should be assigned to projects. C++ and Objective-C modules handle features that need operating system access.

Our work covers the interaction between these parts, suggestion quality and successive releases. Changes to activity capture or native modules need to remain consistent with the interface and time entry workflow.

Results for the client

Preparing a timesheet starts with suggestions based on recorded activity. Users check how time is assigned to projects, make corrections and choose which entries to save. This reduces reliance on memory while keeping users in control of their reports.