The Future of Management

Being competitive and successful in business today follows the same basic rule that it always has; offer something that consumers want to pay you more for than what it cost you to make. Thanks to technology and globalization the competition has grown in scope and in sophistication, meaning that obeying that rule has gotten increasingly complicated over the years. At first managers just told employees what to do and expected they would do it. They were getting paid after all. Then businesses got a little more sophisticated, realizing that happy employees are more productive. Those businesses that wanted to retain talent and be more productive saw the value of managers being less military generals and more coach and counselor. Later World War II introduced a whole new way to manage competitive challenge. The logistical challenges presented by global war forced military managers to become more sophisticated in their planning and analysis, these analytical methods were soon adopted by business managers for decision making. Soon after, as competition, supply chains, and market opportunities expanded across the globe managers realized that there was no one best way to tackle each challenge and that a manager must be able to assess the situation and adapt their management and decision making approach to that particular situation.

This approach was known as contingency theory, and its basic premise is that a manager must assess a situation and make a decision on behalf of the organization. For decades this approach seemingly worked fine. As the world became increasingly complex and less comprehensible managers had to become smarter and more capable. The demand for MBA educated managers exploded, and the proliferation of computers to organize, share and analyze data helped managers keep up. These computers of course became both the solution to a problem, but the cause of another (even more complexity). In recent years DDDM, Social Media, Smart Phones, Search Engines, and E-mail, among other technologies serve to augment human capability thus expanding management’s ability to coordinate and control.

Even so, as consumers and competition became mobile and social, and they began coordinating production and commerce in real time they created massive amounts of valuable data, so much so that humans can no longer keep up and make sense of it all. In a world with more information than we can make sense of, and being competitive in business means making sense of it, what kind of manager do we need?

A smart one? A brilliant one? No. At this point in the story we need a mythical one. That manager doesn’t exist.

Barring some global cataclysm this amplification of complexity was unavoidable. Like compounding interest, as companies and nations compete with each other they build upon their own complexity as they try to solve bigger problems and overcome each other. We can apply Joseph Tainter's theory of the Collapse of Complex Societies to business. When companies confront a problem or competitor they create new layers of capability, bureaucracy, infrastructure, technology, and specialization. Each layer builds upon the previous one to address each new challenge. Either the company’s increased complexity solves the challenge and the complexity is sustained or the company collapses. The imperative for businesses and governments is to get more complex or die. The challenge of course becomes how to manage all this complexity?

At this point in the story it would be a disservice not to mention Herbert Simon’s Bounded Rationality, the idea being that the rationality of individuals is limited by the information they have, the cognitive limitations of their minds, and the finite amount of time they have to process it.

Bounded Rationality dooms Contingency Theory in an incomprehensibly complex world like today because the reality is that identifying and understanding the relevant factors, how they interact and then making good decisions is becoming increasingly difficult, even impossible. How can an approach to management requiring accurate perception of an increasingly complex and unpredictable reality continue to survive? It cannot. The cognitive prowess of managers has been tested and found wanting in today’s competitive environment.

If bounded rationality is the coffin of contingency theory the Law of requisite complexity is its final nail. This law states that in order to control a system the complexity of the controller has to be at least as great as the complexity of the system that’s being controlled. If there is a complexity gap you are going to have to get more complex or you will lose control.

The response to this complexity and “chaos” has been two-fold. The first is a “chaos theory” of management, which has meant management has seen this complexity and tried to simplify it by (1) shortening the time frame to be considered by abandoning long-term planning and instead focusing on being adaptive in the present and to (2) decentralize decision making, because we just don’t have smart enough managers to be able to centralize such complex decision making.

The second reaction has been to make more capable managers by augmenting limited human intelligence with machine intelligence, and in recent years handing more and more decision making over to the machines themselves, in other words, AI (Artificial Intelligence). I would predict, and not be alone in that prediction, that long-term AI will become the dominate form of management in most every complex industry.

While I don’t expect to walk into the office one day soon to have a sit-down with Agent Smith or receive a Pink Slip from a T-800, it seems inevitable that we’ll be expecting our “intelligent machines” to be making an increasing number of decisions for us.

There’s a lot of talk these days about “data driven decision making”. DDDM rests on the idea that decisions should be backed by data, which means these decisions are only as good as your data and your ability to analyze and interpret that data. That seems like a good place for our AI managers to start, since AI will inevitably be better at collecting and analyzing data than any human, after all that’s what computers were built for.

So what types of decisions are we talking about? Decisions can be placed upon a spectrum, beginning with routine structured decisions, often related to operations such as inventory, logistics, and automated marketing responses. This is where AI will replace human managers first, because it’s the easiest. In fact, routine operational decisions are already made by AI in a significant number of industries.

Routine decisions are easy enough to program AI for. The real challenge is unstructured non-routine decisions, the sort that often arise from unexpected events. If AI can figure those out then management jobs are really in trouble. So, what’s the future look like for that? Can it be done?

Making adaptive artificial intelligence, the sort that can manage in complex uncertain environments is a real challenge. Are we years away? Decades? Is it even possible?

The answer probably lies in something called “probability theory”. The basic idea is to understand that individual events influence downstream events, resulting in patterns, which can then be analyzed and then predicted in future situations. The law of large numbers and the central limit theorem are examples.

The current paradigm in AI development is data driven, and largely circulates around advanced pattern recognition, looking for traits in the data. The problem with pattern recognition based AI is that it works great for predictable routine tasks like e-mail filtering and voice recognition, but uncertain, volatile, and ambiguous environments are much more challenging. One promising approach to this challenge uses mathematical algorithms and probability theory to quantify uncertain situations, thus allowing AI to evaluate and make decisions in uncertain environments. This is the dominate paradigm in AI presently, in which AI is really a data processing machine, only as powerful as it’s algorithms’, the quantity and quality of the available data, and the speed at which it can process that data. Give such an AI more data, or make it faster, or program it with more advanced algorithms and it appears smarter, but in reality it’s akin to a really really advanced calculator. It’s not necessarily intelligent in the way most people think of intelligence. This paradigm is different than the previous paradigm of AI, which focused on trying to replicate human reasoning (via neural networks etc.), logic, and knowledge rather than data, mathematics, and statistics. The current data and algorithm paradigm is dominate because the previous paradigm of trying to replicate human reasoning was mostly unfruitful. The question before us is whether mathematics is powerful enough to eventually replicate human reasoning, the kind we see in ambiguous environments such as the fog of war on the battlefield, or replicating human emotional intelligence, something employee’s value from their bosses. While the most challenging to develop AI for, these ambiguous and chaotic environments are precisely the environments where an AI superintelligence would be the most valuable because they are precisely the sorts of environments that are too complex for human reasoning and decision making. Given enough time and processing power could we solve these qualitative problems through quantitative means? Many people seem to think that we can.

AI will also be extremely valuable in strategic decision making situations, such as those found in business and geo-political environments where the environment is global in scale and complicated by hundreds if not thousands of interacting factors. Business and Geo-politics are complicated, and it’s also where the consequences of making bad decisions are the highest. There’s a strong incentive then for big businesses and governments to develop AI. They also happen to have the most money to do it with.

Big businesses like Google and Apple have already committed themselves to AI. In 2012 Google founder Larry Page personally recruited world renowned AI expert Ray Kurzweil to work for Google full time, and Google’s secretive X Labs has focused on an AI project called the “Google Brain”. Apple’s integration of SIRI meant that AI is now in our pockets and purses. SIRI is the result of a DARPA U.S. Defense Department funded project called CALO which was managed by SRI’s Artificial Intelligence Center, and this past January Apple purchased Emotient inc, a technology that reads people’s emotions based upon their facial expressions.

From a business and government perspective managers already augment decision making with software. The DDDM (Data Driven Decision Making) movement is the clearest indication of this, where managers are expected to back up their recommendations and decisions with hard data. It’s not that hard to imagine that in the future most decision making will be at the very least informed by AI, and CEO’s will be expected by their boards to be able to justify deviation from an AI recommendation. Many industries already rely on AI software for decision making. This is most evident in the Finance Sector, where high-frequency trading and derivatives trading is almost fully automated and run by AI software without a great deal of human input. Social networking and insurance are also largely dominated by AI based technologies. Hitachi recently introduced an “AI boss” where it was used to rework employee workflows in a warehouse, effecting an 8 percent increase in “human worker efficiency”.

AI management is seemingly inevitable. As a management scholar and professor I am forced to consider what the manager of 2030 or 2050 looks like. If AI continues to encroach on the role of the human decision maker the question isn’t how many decisions will AI make for us, but how we’ll make sense of whatever decisions it does not. At what point is the computer not something that augments our decision making but instead dominates it? This creates a real moral, spiritual, and human problem because it inevitably leads to another question. In a world that is increasingly automated, where robots do more of our work, and through AI make an increasing number of decisions what is the role of humans in business in 50 or 100 years? It’s not as if this question has been unexplored by some of our best thinkers, and the answer ranges from techno-optimism (it’ll be great!) to techno-apocalyptic (it’s the end of us!). I don’t have an answer, none do, but hopefully through this article you’ve now got a better understanding of how we got here, and maybe where we’re headed.