Four Management Lessons From Self-Driving Cars

It is hard to discuss artificial intelligence (AI) without mentioning self-driving cars. A typical perspective is to look forward to a driverless future, marveling at the possibilities and implications for how we work… or not.

However, there may be more to learn from self-driving cars by looking in the rearview mirror. What can we learn from the path toward autonomous cars about the road to business adoption of artificial intelligence?

Self-driving cars did not start out as completely autonomous; similarly, AI will affect business gradually.

Fantasizing about driverless cars entices us to think about complete replacement, instead of assistance. Even the phrase “self-driving car” encourages this.

But the word “automobile” derives from words meaning automatic motion — the original Benz Patent Motor Car produced in 1886 itself represented a huge step forward from prior locomotion, but it was just the beginning. Since that time, we’ve moved from hand crank starts to electric ignition, from hand throttles to adaptive cruise control, from double clutch to automatic transmission. And AI continues to add to the progress through lane assistance, smart parking, navigation systems: each of these a small step toward self-driving vehicles.

Similarly, modern jobs are complex mixtures of many different types of tasks, many of which have been automated — with more to come. Yet most jobs won’t be completely replaced; they will be progressively augmented in ways that allow refinement instead of outright substitution of machines in place of people.

AI will augment jobs, much as cruise control and back-up cameras incrementally improved the automobile. So the question is not “Which jobs will be replaced?” but rather, “How will jobs be increasingly assisted by AI?”

Self-driving is incredibly difficult; AI can be effective on simpler problems.

Some tasks will be easier for artificial intelligence than other tasks. For cars, early improvements made rapid gains quickly. It didn’t take long for windshields to eliminate unpleasantness from wayward bugs and for electric ignitions to avoid the potential for broken arms from hand crank starters.

But recent AI-enabled progress has required far more technology and effort to achieve. Part of the motivation for driverless cars is the number of fatalities due to motor vehicles; there were an estimated 40,200 motor vehicle fatalities in the United States last year — the vast majority of which involved human errors in judgment. These numbers speak to the enormous potential benefit from improvement.

The numbers also speak to the enormity of the task; those fatalities occurred with our (currently) most intelligent system (a human) in control. The trouble is that driving combines many already individually difficult components — sensing the environment, reacting to rapidly changing conditions, controlling precise physical action — all in the context of potentially dire consequences for errors. As AI takes over the tasks more amenable to AI, the remaining tasks (by definition) will be more difficult. Ferreting out the last, extremely difficult tasks for a self-driving future will take time and considerable effort.

Just as AI took over the relatively easy pieces first for driving, the same will happen in business. What’s more, many tasks of modern jobs are free from many of the difficulties of self-driving cars. Knowledge work, for example, requires little control of physical components. The typing that managers do is an avoidable artifact of humans in the loop between digital inputs and digital outputs. Watching a document being drafted automatically won’t likely appear in a sci-fi action sequence the way that a self-driving car might. But the task is easier, requires fewer technical advances to achieve, and entails less risk. Knowledge work isn’t easy, but it is augmentable. What simple part of your job could be better done by AI?

Self-driving will affect more than driving; the effects of AI will seep widely.

Transportation is a large part of the labor economy, with more than 4.6 million employees in the U.S. in 2014. But despite its size, most people don’t work in the transportation industry. Yet AI developments there can affect us all, considering how much we commute to enable our own profession or recreation. Britons, for example, spend two days a year just waiting at traffic lights. Americans spend one full work week a year commuting. And everyone seems to have stories of those “idiotic other drivers.” It is hard not to want that time back.

Cars are also — loosely speaking — the “crash test dummies” for advances in sensors, data processing, and robotics, not to mention vexing issues of ethics, privacy, technology adoption, security, and externalities.

Beyond these direct effects, the progress toward self-driving cars indicates how pervasive AI can be. When AI affects transportation, it reduces something that costs us. Driving is an overhead cost — an expense, not revenue. We would like for that cost to go away. As it does, we are learning how those changes in cost structures have far-reaching effects. Even if we never sit in a car, these transportation costs are part of every product and service we consume. How will the effects of AI in other industries affect your business?

Self-driving benefits spark attention; these pervasive benefits of AI to others may be risks to us.

Certainly, the future of self-driving cars is important. And fascinating. But a danger in our collective fascination with the benefits of self-driving cars is optimism bias, where we underestimate our own risk of a negative effect from AI. With self-driving, we may benefit from the upside without experiencing the downside. Because transportation is usually someone else’s job, the risk is thinking that AI is coming for “someone else’s job,” but not our own.

The pervasive focus on the benefits from self-driving distracts us from how AI will affect us. We’ve made a lot of progress with cars, but business use of AI is lagging. Don’t let fascination with self-driving cars distract you from the near future of AI in business. That future is about augmentation, not replacement. It’s crucial that business does not fall into a “but we’re not…” comfort zone that equates AI with robots (but we aren’t manufacturers), with radiology (but we aren’t doctors), or with chess (but we aren’t professional chessmasters). Advances in AI will likely affect everyone to some degree. How will AI blindside your organization?