Artificial Intelligence has been used in many fields of production with a certain degree of success. It has been used to provide supervisory control of automated facilities in the industry from many years ago. Supervisory control fits well artificial intelligence because it requires coping with unexpected situations far from the expected behavior of the system under automatic control. This can sound odd but intelligence is the ability to adapt our behavior to not common events from the environment. Large automated facilities can reach a working stated where the mathematical algorithms of controllers cannot preserve the system in a stable state. When the system is moving far from its working point, human supervisors can take actions in order to put it back to a better working state. The use of artificial intelligence instead of human one can be accepted depending on its error rate. Humans fail and humans fail a lot. Sometimes computers can produce better results than humans in many tasks because computers do not tire, for instance. They do not need to sleep and they are totally objective in their decision making process.
Management has a part of supervisory control. Any management activity can be divided into two different parts: one strategic and one operational. The strategic one is related to the definition of new tasks for the future, and the operational one is related to the control of the tasks of a certain process making decisions to preserve the organization near the planned evolution from the previous strategic activity. In other words, the operational part of management is supervisory control and it could be done by current artificial intelligence techniques.
The strategy requires more than analysis, the strategy requires imagination. The strategist imagines the future of the organization, and imagination is not control. This part of management is far from current artificial intelligence techniques. Our current AI is mostly reactive, with machine learning we can make machines that can learn from the past and provide an action for the future, but we cannot make machines that imagine or have its own dream.
This is not totally true. There are techniques that can be used to make that the system can provide unknown (not learnt) solutions. However, they are not so known as expert systems or common neural networks.
Artificial machines defining strategies exists from several years ago. All of us can have a chess game in our computer. This kind of software can create a path of movements trying to predict the movement of its rival; however, common management usually is much more complex. The number of rivals can be much higher or even undetermined. The number of factors that affect market relationships can be huge. Rules can change. The degree of adaptability that an artificial strategist for business management would require can be very large.
This is not science fiction. There are a lot of robots operating and making decisions about investments in Wall Street, however, their working is very far from our human intelligence. The use of predictive mathematical algorithms is very far our imagination. Creative process is, by definition, unpredictable.
A machine can make predictions about the volume of sales of a new mobile phone. But can a machine imagine what a mobile phone is before it was invented? This is a good question. People can create its own future and create the future of their environment, but can a machine create its own future too?
My answer to this question knowing a lot of AI techniques is: Of course, a machine can have some kind of creativity. I can program some software that can do that right now. But probably, nowadays, that ability could not be competitive in front of the natural intelligence that has evolved during thousands of years. While robots can have a lot of advantage to make analytics in front of humans, they have a lot of disadvantages to be creative. Creativity is not fast analytical processing, although sometimes fast analytical processing can seem to be creativity.
The robotic strategic manager is a great challenge still, but the supervisor or advisor robot is real yet.