5. Implementing and Evaluating AI systems

AI systems often have to be tailored for specific situations and uses. In the healthcare environment it is less likely that processes will be entirely automated. This often involves humans working with AI systems (human in the loop) where AI systems automate some (usually routine or repetitive) tasks to free up time for specialists to focus on other areas. This sees a closer working relationship between human experts and AI systems.

Once AI systems have been implemented they need to be evaluated to ensure they are safe and fit for purpose. Algorithmic drift can occur prompting the need for continual readjustment. The system also has to be evaluated in the local context to ensure it is configured correctly and working optimally for the chosen goals.