The typical sequence of AI-supported projects serves as a guide for determining the required roles within the project team and the skillset of the individual employees. Four phases are distinguished during implementation, which will be explained in more detail here:
Phase 1 – Problem identification by the data expert.
First of all, it is necessary to obtain an overview of all the requirements for the data material needed. This determines whether the project will be implemented using artificial intelligence or whether software development will suffice. The technical expert defines the nature of the required data as well as its trivial dependencies and the distinction between trivial and numerical data. The metrics for the AI application also need to be defined early on. The size of the data set also plays an important role.
Phase 2 – Data pre-processing and function development.
During data preparation and function development, the data expert sifts and cleans the existing data. It is necessary to check whether all input data are available and plausible. Now it has to be clarified which functions have to be embedded, whereby it is necessary to focus on the relevant functions. At the same time, it must be determined whether there are obvious dependencies between existing functions and whether new functions have been added.
Phase 3 – Selection, training and evaluation of the data model
The selected model must fit both the problem and the nature of the data. Appropriate models help to identify the extent to which the smart model correctly analyzes the data and whether it is suitable for the activities to be performed. At the same time, this phase determines whether the AI-powered application has learned the right data patterns. Important factors here are whether the analyses are performed expertly and whether the assessments performed by the AI-powered application produce the right results.
Phase 4 – Application in operation.
The AI model developed in the third phase is now integrated into a company’s operations. In this process, the technician in charge of AI tunes the model to the specific use case. Particular attention must be paid here to scalability and ease of use, as well as simple transfer to subsequent models. For an efficient application and subsequent use of machine learning models, a data platform for Big Data is absolutely necessary so that the data is and remains secure.
In order to make strategic decisions for the successful implementation of AI projects, a team of experts must be assembled. Important roles here are played by data experts, data analysts, innovation managers and software developers.