Parfois

Guilherme Efe Macário

While AI is a concept that has been studied for over three decades, its practical application in non-tech businesses has remained elusive. Despite its long history, the question of when and how AI would become accessible to companies outside the tech realm has posed a challenge. In my view, we have reached a point where AI can be efficiently and effectively used by companies, but this requires genuine interest and commitment. This interest is pivotal for the success of AI-based projects, as they demand time, expertise, and real-world business experience to yield meaningful results. To succeed, it is crucial to view knowledge not as a passing fashion trend but as a ‘neo-classical’ approach to problem-solving for businesses dealing with vast amounts of information and data.

So, how can non-AI or even non-tech companies benefit from AI and predictive statistics to enhance problem-solving and the decision-making process, thereby increasing efficiency?

Based on my experience and perspective, we must ensure two key points and define a third one:

1. Presentation of AI Projects: It is essential that the first AI projects are presented to the business only when they are near-perfect versions. This means they should be well-explained, with the business impact accurately measured and without any major flaws. This approach might differ from using an agile methodology, but it is necessary to instill confidence in stakeholders by presenting results that leave no doubts. While not mandatory for all projects in the future, it is crucial when introducing data science projects to companies not accustomed to this approach. This establishes engagement between data scientists and business stakeholders.

2. Involvement of Team Leads and Managers: It is mandatory to involve team leads and managers, along with a well-structured data team possessing knowledge in fields such as predictive statistics – using historical data and statistical algorithms to forecast future trends. Without this knowledge, it's not even fair, to say the least, to engage in these discussions. Understanding how these models work at a higher level is crucial for comprehending their varying performance over time and the human effort required for monitoring. We must bear in mind that predictive statistics are not a crystal ball to predict the future. This means that the output can be much better than using descriptive statistics. However, it takes time, and it is a long and complex process with no endpoint and in constant motion.

"AI or predictive statistics should be seen as one among several available tools to help solve complex or simple problems with better insights, depending on the context."

3. Leveraging External Expertise: For non-tech or non-AI companies, there are numerous data science expert companies in the market that can help accelerate your business cases with greater reliability than attempting to develop everything in-house. These companies are more attuned to the constant innovations in this field, and partnering with them is advantageous. Working with these companies has significant benefits in terms of internal training, as companies can learn from the experience of working with these experts. I strongly believe that both parties benefit from synergies between people with mathematical and technological expertise merged with business knowledge.

Considering this, it is important to note that simple descriptive statistics (means, median, variance, etc.) are often sufficient for most companies' problems. It is a big mistake when data scientists or team leads attempt to implement complex and sophisticated big data projects, while simple descriptive metrics such as averages and variances are more than adequate for making quick and intelligent decisions. AI or predictive statistics should be seen as one among several available tools to help solve complex or simple problems with better insights, depending on the context. It is our responsibility to understand where we should use it, always measuring the return on investment and the time required for meaningful results.

In conclusion, as team leaders, our interest in making decisions based on predictive statistics should be grounded not in belief but in knowledge. Given the tools available today that have democratized the usage of these approaches, this is possible. However, it necessitates creating a work environment conducive to efficiency and effectiveness in daily business operations. To achieve this, we must have the right people, team leads, tools, knowledge, commitment, and sometimes the patience to change our company culture. While it's not easy, it is certainly achievable.