Applying Artificial Intelligence and Big Data Analytics in Forecasting the Effectiveness of Managerial Strategic Decision Making

In the modern business world, leaders face a constant challenge: making high-stakes decisions in a landscape that changes faster than any single person can track. To navigate this, many organizations are turning to artificial intelligence and big data analytics. These are not just tools for sorting files; they are systems designed to find patterns in massive amounts of information that would overwhelm a human mind. However, a critical question remains: can these digital systems make decisions that are not only smart but also legally sound and strategically effective? The answer is not simple. A decision might be efficient but break a rule, or it might follow the law but fail to meet a business goal. Researchers are now trying to build a bridge between the raw power of data analysis and the strict boundaries of legal compliance, asking whether a computer can learn to weigh the importance of a business goal against the risk of a legal penalty.

A researcher at Tay Do University in Vietnam set out to test this idea by building a digital system designed to simulate how a manager might make strategic choices. They created a virtual environment where an artificial intelligence agent had to navigate a complex set of rules based on information technology laws. The goal was to see if the system could predict how well a decision would perform, measured by a standard business score known as a Key Performance Indicator. To do this, the researcher did not just feed the computer raw numbers. Instead, they designed the system to mimic five specific ways a human mind processes information: the ability to manage resources, the capacity to think about one's own thinking, the skill to focus on what matters, the speed of learning new rules, and the ability to understand social and organizational norms. These five mental skills were treated as the building blocks for the computer's decision-making process.

The researcher constructed a massive digital pipeline to test their idea. First, the system gathered text from various documents, such as legal reports and policy papers, and cleaned them up so the computer could read them. It then translated these written rules into a format the machine could understand, assigning weights to different requirements like data privacy or executive alignment. Next, the system generated a simulated dataset of 1,500 different decision-making scenarios. In each scenario, the computer had to balance its five cognitive skills against the legal rules to produce a final score. This score represented how successful the decision would be, ranging from zero to one hundred. The researcher then trained a machine learning model on this data, teaching it to look at the five cognitive skills and guess what the final score would be. They used a method that builds many small decision trees to find patterns, hoping the computer could learn the hidden connections between how a system thinks and how well it performs.

When the researcher tested their model, the results revealed a difficult truth about the current state of this technology. The system was able to run the entire process smoothly, from reading the documents to generating a final list of predicted scores. However, when the researcher checked how accurate the predictions were, they found that the model struggled to capture the true complexity of the problem. The computer's guesses were often off by a significant margin, and in some statistical measures, the model performed worse than simply guessing the average score for every situation. The data showed that the actual outcomes fell into two distinct groups—decisions that scored very low and decisions that scored very high—while the computer's predictions tended to cluster in the middle, missing the extremes entirely. This suggests that while the system could process the information, it lacked the deeper context needed to distinguish between a safe, average choice and a high-risk, high-reward strategy.

The study concludes that while the framework for combining artificial intelligence with legal governance is feasible, the technology is not yet ready to replace human judgment in these complex areas. The researcher found that the current models could not fully account for the hidden variables that push a decision into the "high risk" or "high reward" categories. They noted that the system worked best when it stayed in the middle ground, avoiding the most extreme outcomes, which is a safe but limited approach for strategic planning. The researcher suggests that future work must focus on adding more detailed information to the system, such as better ways to measure the specific legal context of each decision, and developing models that can explain their reasoning more clearly. For now, this research serves as a proof of concept, demonstrating that it is possible to translate legal rules into digital data, but also highlighting that the path to truly autonomous, compliant strategic decision-making requires much more than just processing power.