Using Al involves a series of steps: collecting data, labelling it, developing and training the model, testing and evaluating it, and finally deploying the solution. At the beginning, data collection and labelling take 20-50% of the time (higher % for smaller or standard projects).
The rest of the time is spent on model development, testing, and deployment, which depends on how unique the problem is, the performance level required, and the complexity of deployment. Al projects can last anywhere from 3 months to several years, depending on the size and collaboration needed.