Preparing for an AI Future

Published:

December 10, 2018

Updated:

April 10, 2025

Featured Image on Preparing for an AI future

Published on INFORMS OR/MS Today (Joseph Byrum)

How realistic is ‘O.R. in a box’ and what does it portend for O.R. professionals?

O.R. in a box: It could happen with AI using known algorithms, applying them to new situations and constantly adjusting and recalculating to maintain optimal efficiency.

The workplace as we know it may not be recognizable in a decade’s time. If the McKinsey Global Institute is to be believed, automation will render obsolete as many as 800 million jobs worldwide by 2030. Artificial intelligence (AI) is the key driver in this economic disruption, which is why it is critical to prepare oneself for a future dominated by AI.

In the report “Jobs Lost, Jobs Gained” [1], McKinsey warned that robots and AI will prove far more disruptive in the workplace than anything we’ve seen before. Neither mass production nor personal computing took hold as quickly as AI is expected to do, and that rapid pace means many are destined to be left behind. Paralegals and accountants, for example, may be surprised to find themselves in endangered professions as many of their routine functions are already being replaced by algorithms.

Surely such a fate could never befall an experienced operations research (O.R.) practitioner. Or could it?

Deep-learning AI is becoming ever more capable. It its simplest form, a system fed thousands of images of cats and dogs will recognize the distinct patterns of a cat image and the patterns of a dog image. Fed enough samples, the system can then be shown a photo of a pet greyhound and identify it with some level of accuracy as a dog. Such systems are ideal for classification tasks, but the person using AI algorithms of this sort is in the dark about how the system actually makes its decisions. Not much expertise is needed to use it.

O.R. in a Box?

Now imagine what might happen if a general, deep-learning AI routine could, some time from now, be “trained” with a company’s data, automatically run simulations of the feasible futures and create algorithms needed to deliver the desired business efficiencies. The AI might use already known O.R. algorithms, apply them to new situations and constantly adjust and recalculate to maintain optimal efficiency. It would be O.R. in a box.

Surely the O.R. professional’s first thought would be “that can’t happen to us.” But, then again, the last thing a card-carrying member of the Writers Guild of America or a Hollywood producer would ever think is that an AI algorithm could replace their creative work. Nonetheless, AI is in the primitive stages of doing exactly that – drafting screenplays [2] and editing movie trailers [3]. Of course, none of these undertakings is Oscar-worthy … yet.

O.R. in a box might start simply. It could develop algorithms to optimize the shipment of consumer products for a single widget manufacturer, identifying the most efficient way to deliver each item from the factory to the store shelf or the consumer’s doorstep. It would gradually expand capabilities to ensure an optimized delivery of the raw materials needed to produce the widgets and every other logistical task needed to cut costs and boost factory productivity.

This AI application might then optimize marketing efforts, ensuring the product design meets customer expectations, as well as ensuring that advertising accurately targets those most interested in the product. Soon enough, other widget makers would use the same system, and the general principles would be adapted from widget-making to other forms of manufacturing. From there, it might adapt to help retailers in optimizing product placement on store shelves, and then provide support for logistics, marketing, accounting and all the other common business tasks.

Under this hypothetical scenario, the days of laboriously customizing O.R. solutions for each industry, each business unit and each specific task would be over. The off-the-shelf O.R. product, if it’s possible, would massively lower the barrier to entry for data analytics.

But how realistic is this?

As we’ve seen with the prospect of fully self-driving cars, hype sometimes outruns reality. Premature deployment of autonomous vehicles on public roads has led to a few high profile accidents, ultimately setting back the cause the vehicles’ backers were hoping to advance. The roads are too chaotic and complex to be managed solely by an algorithm – for now.

So it is with AI and O.R. Even things that are clearly possible take time to become workable products. Self-adapting O.R. algorithms are not a near-term reality. The business environment is far too varied and complex for a one-equation-fits-all solution – at least not until general AI systems become reality. The important conclusion to draw from this is there’s still time for O.R. practitioners to develop the skills that we know will be essential in the AI future.

Key Skills in an AI-dominated future

In the near term, AI is most powerful when wielded as an efficiency tool. Like a robot in a factory, or a plow for a farmer or a hammer for the carpenter, AI takes human energy and ingenuity and magnifies their impact.

This is why the most effective, present-day AI systems are designed around the concept of intelligent augmentation. As such, they handle the math and basic analysis, relieving skilled operators of a tedious burden. They absorb data, classify and prioritize information, conduct simulations and ultimately leave it to the human operator to decide the course of action. Such systems don’t render humans obsolete, they rely on experienced operators.

The Defense Advanced Research Projects Agency came up with this approach in the 1980s with a system for easing the cognitive burden on combat pilots [4]. It worked, and the same approach can work for business today. O.R. professionals will not lack opportunity in developing and implementing these solutions for quite some time.

In fact, AI tools of this sort will provide a competitive advantage to the companies that use them, meaning O.R. will be in more demand. That’s good news for an industry already on an upswing. The Bureau of Labor Statistics predicted the demand for O.R. jobs would grow 27 percent over the next decade [5].

In the AI future, jobs requiring human creativity and management expertise will enjoy the greatest job security. This is our competitive advantage as humans. Maximum value will be created by those who understand how to collaborate with other humans while working with AI tools to enhance their organization’s effectiveness.

Over time, AI will continue to lower barriers to entry into the field. The power of AI makes optimization techniques more generalized and easier to use. That means technical roles that focus on the intricacies of software environments like R and MATLAB will be less essential. As AI picks up more of these details, the key to professional survival will be in understanding the big picture of business optimization.

No particular outcome is guaranteed, of course. If it were possible to predict decades into the future with perfect accuracy, then our driveways would have those flying cars that we’ve been promised since the 1950s. On the other hand, favorite science fiction props like handheld communicators became cell phones, and characters using their voice to tell a computer what to do became Alexa. So at least some technology trends are discernible.

Today’s students ought to pay close attention to AI trends. They would be best served by focusing on developing general skills that are highly adaptable, such as mathematical modeling, statistics and science. That way, they will be prepared to use increasingly powerful AI tools in the unexpected roles they are likely to play in the future.

Those waiting for O.R. to work on autopilot, on the other hand, might end up waiting a bit longer than expected.

O.R. in a box: It could happen with AI using known algorithms, applying them to new situations and constantly adjusting and recalculating to maintain optimal efficiency.

O.R. in a Box?

Key Skills in an AI-dominated future

References

Related Posts