Machine Learning is (in our sphere) about solving problems or optimizing flows that have very large amounts of data to study. In those situations, a good first step is to identify and quantify the problem. Here, the quantitative part is of course important, but it is also important to gain such a deep insight into the qualitative aspects during data collection. Given that there is enough data to start a project, data can now be applied to the problem. If the data is of the quality that it is possible to get insights from, the next step is to use several algorithms to find patterns from the data.
You may need to supplement with more data based on the patterns you find or go deeper into the analysis. When insights are supported by data, an action plan can be formulated. It is important to present one in an educational way to ensure that everyone understands the value of the implementation. Once the implementation has been completed, it is important that the impact on results is ensured.