AI project indicators express what will change through the deployment of a solution, and their values become a commitment towards the provider. They should be chosen so that they are measurable from data the company actually has – typically the time saved processing a single case, the number of documents or requests processed, the share of tasks handled automatically, the reduction in error rate, or the number of employees using the solution. Indicators that depend on the market, such as revenue growth, should be avoided, because failing to meet them is hard to defend. Before submitting the application, it is worth measuring the baseline state, otherwise improvement cannot be demonstrated. Overstated promises earn points at evaluation but lead to a reduction of the contribution at inspection. It is therefore worth describing the method of measurement directly in the application, including which system the data will be obtained from.
See also: Output and result indicator, Return on AI investment, Measuring skills during training.
