Discussions about the economic dangers and potentials of artificial intelligence (AI) tend to center around the ability of AI to do specific, clearly-defined tasks quickly and well. This is of great concern to employees whose jobs solely revolve around doing these tasks, as they run a risk of losing their job, or at least that part of their job, to a machine in the not-so-distant future.
But what about at the management level? As the technology that allows machines to be “intelligent” gets better, will AI begin to take over the complex decision-making processes that often fall under the responsibility of management? To answer these questions, let’s consider what AI is capable of now, what experts think AI will be capable of in the future, and how these capacities may impact the decision-making process.
The current use of AI in management
Managers currently use AI to help complete specific rote tasks, as is fitting for the limited AI we now have. Administrative work like scheduling and record keeping—which can take humans an excessive amount of time to accomplish—can easily be handed over to a machine. These activities are distinct from the higher-level decision-making and creative tasks that often concern management, and indeed AI grants humans more time to focus on these issues.
That said, some decision-making tasks are already being done or informed by AI. Artificial intelligence can be relied on to make decisions when dealing with highly-structured data and clear-cut choices based on that data. This includes such areas as determining which ads are most effective, making certain financial choices, and even determining when new inventory or supplies must be reordered.
AI can collect, process, and summarize data in ways that humans never could. Many also find that AI can effectively review data for patterns that humans may miss, providing data analysis that informs managers’ decision-making processes. This is the primary benefit of using AI in decisions, as humans can either miss important information or make cognitive errors that machines can be programmed not to.
There is a limit to this, though. As former Stitch Fix chief algorithms officer Eric Colson puts it in the Harvard Business Review:
“There are many business decisions that depend on more than just structured data. Vision statements, company strategies, corporate values, market dynamics all are examples of information that is only available in our minds and transmitted through culture and other forms of non-digital communication. This information is inaccessible to AI and extremely relevant to business decisions.”
This point seems unlikely to change in at least the near future; humans will remain the “Central Processing Unit” for many high-level decisions.
An increased role for AI in the future
While Colson advocates for combining human and machine skills in the decision-making process, might we soon reach the point where machines can go it alone at high levels? At the moment, despite the optimism of many computer scientists and the fears of many economists, a general AI capable of directly participating in decision-making processes at a human level seems to be far away. Abstract thinking remains difficult for even the most powerful machines and, perhaps most importantly, some information still cannot be digitized in a meaningful way.
Given this, a more modest estimate of where technology will be in a few years includes ever-improving but weak AI capabilities. But even if we don’t presume the eventual creation of a general AI, the ever-increasing capacities of weak AI still hint at changes to business management processes in the future. Colson mentioned several key areas where AI can currently make good decisions. All of them share the traits of using large amounts of data and an easily-defined goal. The sphere of activities this encompasses is likely to expand in the next few years while remaining tied to concrete actions.
Colson’s proposed model of decision-making in areas where machines can’t, or perhaps shouldn’t, work alone places machines first, giving them access to large amounts of raw data which they can use to model possible options for executives and decision-makers to choose between and act upon. Humans, with their ability to see beyond the cold rationality of AI and to include variables such as values in their thinking, would select from the presented choices or work from them when making a large decision.
In a sense, a more advanced version of this could be the “Goldilocks” outcome. One where machines do some of the work for us, but not all of it. Where human touches and abilities are still applied to executive decisions, but augmented by the increased accuracy of information offered by AI data processing. This outcome does require, however, that people are willing to work with technology in new ways, and to understand what management leaders and machines are good, and not good, at doing. A step that many will find difficult at first, but impossible to avoid taking in the long run.
Visions of machines making executive decisions are still far away. Leaders hoping to incorporate AI into their decision-making processes will have to be content with using AI to augment, rather than replace, humans. While this might not be the technological utopia some had wished for, it does offer opportunities for improved management, less wasted time, and the ability to combine what humans and machines both do best.