Will AI Replace Managers

In June 2017 the global consulting firm PwC issued a report on the economic impact of artificial intelligence (AI) on the UK economy, concluding that 30 percent of the jobs in the UK and other major economies in the EU were at high risk of automation by 2030. In the United States, PwC estimated that 38 percent of jobs were at risk; in Germany, 30 percent; and in Japan, 21 percent. PwC defined high risk as meaning a substantial likelihood that the job would be replaced by AI or substantially changed, resulting in lower compensation. According to PwC, the sectors of highest risk are transportation and storage, with 56 percent of jobs at risk through deployment of autonomous vehicles and automated warehouses; manufacturing, with 46 percent at risk due to advances in robotics; and wholesale and retail, with 44 percent at risk due to growth in e-commerce. In the UK, workers with a high school or less education face a 46 percent risk, but for those with a college degree, the risk is substantially less (only 12 percent). On the more positive side, these same studies report that only 10 percent of these jobs will be eliminated entirely, although a larger percentage will decline in growth and compensation over time. Moreover, in the past, new technologies have created new jobs, enhanced productivity in the economy as a whole. The net effect on aggregate employment is there-fore not clear, or perhaps not as worrisome, as is the impact on the distribution of the benefits, which is likely to be highly skewed toward educated and professional workers, leading to even higher levels of inequality than already exist in industrial societies. The results of the PwC report echoed the results of similar studies by consulting firms.

As concerning as forecasts for the impact of AI on employment and compensation may be, the rapid advances of AI in the last five years also raise questions about the fate of occupations that currently require relatively high levels of education and professional degrees, including managers in business and government sectors. AI systems today are capable of speech recognition (Siri and Alexa), natural language recognition (Google Search), language translation (Google Translate), vision, pattern discovery, and facial recognition (Facebook photo tagging). Because these functionalities are central to the performance of various tasks involved in management and professional work, there is the potential that many professional and managerial jobs will also be at high risk of automation.

What can managers do that a machine cannotat least not yet? First-generation AI tools captured the expertise of human decisions in computer software, and could repeat these decisions faithfully at large scale and minimal cost. This approach worked for fairly simple routine decisions where the decision could be codified into a number of steps and human experts were able to specify exactly how they per-formed the task. However, humans often cannot accurately describe how they perform complex decision tasks. For instance, driving a car or any complex physical or mental activity is something most of us know how to do but cannot write down as a set of concrete steps.

Second-generation AI systems rely largely on ma-chine learning techniques, which are not traditional computer programs but instead rely on pattern recognition in large data collections (big data) of prior human decisions, images, text, and speech. These systems can take input from millions of digital im-ages, documents, emails, and social media posts, and produce outputs. For example, AI programs are currently being used in transport to drive autonomous cars and trucks based on data describing how human drivers actually behave when facing typical driving situations; in medicine to identify tumors, make diagnoses, and perform surgery; in law to analyze legal documents; and in business for tasks such as hiring and promotion decisions, personalizing customer service, improving customer loyalty and retention, smoothing supply chains, detecting fraudulent trans-actions, predicting machine failures, monitoring and managing physical facilities, and career planningto name just a few current applications. In com-plex games like chess, Go, and the television game Jeopardy, computers have often beaten world-class players.

But management is a complex task that involves both routine codifiable tasks and decisions (such as writing and filing monthly sales reports, scheduling appointments, and responding to email), as a well as more complex tasks that cannot be codified or taught in any simple sense to machine learning systems. Generally, these two types of tasks are interrelated and complimentary: the performance of com-plex tasks is supported by the successful performance of routine tasks. Real-world managers define and solve problems, are able to quickly realize changed circumstances, create new products and services (often where none existed before), and play both interpersonal roles (as leaders and colleagues) and informational roles (as disseminators, spokespersons, and translators of corporate policy to workplace set-tings and business processes). These tasks require emotional intelligence, general intelligence, and com-mon sense: in short, what we call human judgment. The message to managers is clear: if your expertise relates to performing routine tasks that can be codified or behavior that can be patterned, your opportunities and compensation are likely to decline in the future. But if your expertise is in problem solving and in creating and understanding people and situations, then AI may be able to help you do your job even better, increasing both your opportunities and compensation.

CASE STUDY QUESTIONS

1. What sectors of the economy will be most affected by AI? Which the least?

2. Why are those with a college degree or above at the least risk of being replaced by AI?

3. What new service sector jobs have been created or left unchanged by automation in the past? Why?

4. Why is it unlikely that AI will replace managers?