The rise of Artificial Intelligence (AI) methods in the past decade, along with the realisation that automated data-based AI tools can outperform even top-level human experts at some tasks, has given rise to the hope that AI may soon be able to take over many of the functions of a manager. This is hardly a new dream.
Frederick Taylor, in the late nineteenth century, already imagined the transformation of management to an engineering subject. The discipline of operations research was invented in the last century, mostly by experts working for the US Department of Defence, to quantify and automate managerial tasks such as scheduling of work, planning of inventories, distribution and logistics, and production planning and management. A lot of tools that are used in AI were invented by operations research engineers. The advance of computer and communication technology in the latter part of the last century brought the dream of automated management and decision making even closer to reality.
Automated decision making is indeed, one of the central topics studied in AI. The general case is that of an intelligent agent that needs to make a sequence of decisions in the presence of uncertainty. A number of AI tools are already available that can take a description of the business environment, its laws of evolution, and the set of actions that the manager can take and find a sequence of actions that minimises costs or maximises rewards. However, most of these tools are of recent vintage, still have a lot of theoretical shortcomings, and have not been tested extensively under real life conditions. Also, they work well only when the business problem they address is very well-defined and can be expressed numerically.
As such, these tools can be of great use when the decisions to be made are in a well-specified and precisely measured domain such as production planning, or advertisement targeting, or financial asset-liability management. When the problem is less clearly specified, and less clearly measured, say as in customer satisfaction improvement, or human resource management, or strategy formulation, the available tools are still at a very early stage of their life cycle and they are of an elementary nature. As such human intuition and expert judgment are more successful that any automated AI based method.
What will it take to change this? The answer turns out to lie in the relentless expansion in the capabilities of computer hardware and software, as embodied in Moore’s Law and its corollaries. There is little doubt that in the 2030s we will look back at the early 2020s as an era of primitive computers that had only terabyte storage, gigabyte memories and gigahertz clock speeds and 5G communications.
As we push on to exabyte storage, petabyte memories and 7G communications, we will be able to pose messier real-life problems using conversational natural language interfaces, and solve them using much more advanced algorithms, maybe even quantum methods. This will bring business management automation within our grasp and ever more realistic decision problems can be solved to a degree of efficiency that will be beyond the abilities of even the best human managers.
The writer is Dr Debashis Guha, Director, Master of AI in Business, SP Jain School of Global Management
