IDC forecasts that the use of artificial intelligence will grow at a good pace both in Spain and worldwide. However, many projects fail. Among the factors that must be taken into account to minimize the risk of failure are the selection of the right team, the right choice of initiatives and the search for promoters and supporters for the initiative.
IT User
April 04, 2022
Spending on artificial intelligence solutions will grow worldwide by almost 20% by 2022, according to IDC estimates, and in Spain it will grow at an annual rate of 27% until 2025. However, 85% of AI projects fail, according to psychologist and Harvard University professor Howard Gartner.
Francisco Díaz, business analyst at Compensa Capital Humano, of the Howden Group, recommends finding an internal promoter for the project, collaborating with data managers, making an optimal selection of Machine Learning initiatives, drawing up a project charter, selecting a team with the right profiles, involving stakeholders and maintaining constant monitoring. Let's examine these recommendations one by one:
Search for an internal promoter for the project
One of the main causes of failure in IA projects is the lack of support and leadership. Initiatives in this field are very attractive, but the probability of failure is high. Therefore, it is desirable to create a prototype that illustrates the concept, without the need to use all resources, and helps to glimpse its results.
Data collaboration
Artificial Intelligence is based on data and, to a greater or lesser extent, the company will have people or groups that handle information needed for the project. So there must be someone in a position to ask them for this information. Lack of collaboration is another of the most frequent causes of failure and will also manifest itself in the reluctance to allocate resources to the project for a wide variety of tasks to be executed outside the development itself.
Optimal selection of Machine Learning initiatives
A project of these characteristics requires an investment in resources, which will need to be well planned to justify its cost. In the proposal it is preferable to focus on the business problem to be solved rather than on the technological features. It should also include an approximate ROI (return on investment), the time to market the idea, the estimated effort and the pitfalls that will have to be overcome. Not to mention a technical feasibility analysis.
Drawing up a project charter (Project charter)
The definition of the project and its requirements is transcendental to start its development. This project charter must know the scope of the project, what we want to build and the business objectives.
Team composition
To avoid the lack of experience and the disconnect between software development and data science, we need to define the necessary profiles. We will need a data science specialist, but also a data engineer with knowledge of IT and more traditional programming. It is essential that business experts are involved in the team so that they can monitor the results. They do not necessarily have to be brought in externally; often there are already resources in-house or more appropriate training opportunities.
Involve stakeholders
During the life of the project, there will be interactions with a wide variety of professionals and suppliers that must be properly managed. We must also be aware of the reluctance that IA may cause as a substitute for tasks that are currently performed.
Constant monitoring
Problems cannot arise only in the implementation of the project, but it is necessary to pay attention to how to execute what we have drawn. The possibilities of artificial intelligence are endless, so it is advisable to keep a conservative scope and set up development phases. Also, keep in mind that AI projects have a software development component, but it is also important to choose the right management method.
Finally, the specialist explains that the set of technologies and algorithms that we can choose to implement our solutions is very broad. "It is important to choose simple and transparent solutions, and, above all, that it is easy to explain their inner workings," he concludes.