Prediction Machines: The Simple Economics of Artificial Intelligence

by Ajay Agrawal

AI is a prediction technology, predictions are inputs to decision making, and economics provides a perfect framework for understanding the trade-offs underlying any decision.

Prediction Machines is not a recipe for success in the AI economy. Instead, we emphasize trade-offs. More data means less privacy. More speed means less accuracy. More autonomy means less control.

From the economist perspective, Google made search cheap. When search became cheap, companies that made money selling search through other means (e.g., the Yellow Pages, travel agents, classifieds) found themselves in a competitive crisis. At the same time, companies that relied on people finding them (for example, self-publishing authors, sellers of obscure collectibles, homegrown moviemakers) prospered.

Tim Bresnahan, a Stanford economist and one of our mentors, pointed out that computers do arithmetic and nothing more. The advent and commercialization of computers made arithmetic cheap.5 When arithmetic became cheap, not only did we use more of it for traditional applications of arithmetic, but we also used the newly cheap arithmetic for applications that were not traditionally associated with arithmetic, like music.

What will new AI technologies make so cheap? Prediction. Therefore, as economics tells us, not only are we going to start using a lot more prediction, but we are going to see it emerge in surprising new places.

Prediction is the process of filling in missing information. Prediction takes information you have, often called “data,” and uses it to generate information you don’t have.

When prediction is cheap, there will be more prediction and more complements to prediction.

Prediction facilitates decisions by reducing uncertainty, while judgment assigns value. In economists’ parlance, judgment is the skill used to determine a payoff, utility, reward, or profit. The most significant implication of prediction machines is that they increase the value of judgment.

From religion to fairy tales, knowledge of the future is consequential. Predictions affect behavior. They influence decisions.

From a purely statistical point of view, data has decreasing returns to scale. You get more useful information from the third observation than the hundredth, and you learn much more from the hundredth observation than the millionth.

One major benefit of prediction machines is that they can scale in a way that humans cannot. One downside is that they struggle to make predictions in unusual cases for which there isn’t much historical data. Combined, this means that many human-machine collaborations will take the form of “prediction by exception.”

Prediction machines will have their most immediate impact at the decision level. But decisions have six other key elements (see figure 7-1). When someone (or something) makes a decision, they take input data from the world that enables a prediction. That prediction is possible because training occurred about relationships between different types of data and which data is most closely associated with a situation. Combining the prediction with judgment on what matters, the decision maker can then choose an action. The action leads to an outcome (which has an associated reward or payoff). The outcome is a consequence of the decision. It is needed to provide a complete picture. The outcome may also provide feedback to help improve the next prediction.

Figuring out the rewards from these various actions—the key piece of judgment—is reward function engineering, a fundamental part of what humans do in the decision-making process. Prediction machines are a tool for humans. So long as humans are needed to weigh outcomes and impose judgment, they have a key role to play as prediction machines improve.

Automation occurs when the return to machines handling all functions is greater than the returns to including humans in the process.

Automation can also arise when the costs of communication are high.

The researchers found that machines do a better job of recommending jokes, but people prefer to believe the recommendations came from humans. The people reading the jokes were most satisfied if told the recommendations came from a human, but when the recommendations were actually determined by a machine.

Like classical computing, AI is a general-purpose technology. It has the potential to affect every decision, because prediction is a key input to decision making. Hence, no manager is going to achieve large gains in productivity by just “throwing some AI” at a problem or into an existing process. Instead, AI is the type of technology that requires rethinking processes in the same way that Hammer and Champy did.

To make the most of prediction machines, you need to rethink the reward functions throughout your organization to better align with your true goals. This task is not easy.

Knowing when yogurt is going to sell helps you know when you should stock it and minimizes the amount of unsold yogurt to discard. An AI innovator who offers prediction machines for yogurt demand could do well, but would have to deal with a supermarket chain in order to create any value. Only the supermarket chain can take the action that stocks yogurt or not. And without that action, the prediction machine for yogurt demand has no value.

First, as in Amazon’s shipping-then-shopping model, prediction machines reduce uncertainty. As AI advances, we’ll use prediction machines to reduce uncertainty more broadly. Hence, strategic dilemmas driven by uncertainty will evolve with AI. As the cost of AI falls, prediction machines will resolve a wider variety of strategic dilemmas.

Second, AI will increase the value of the complements to prediction. A baseball analyst’s judgment, a grocery retailer’s actions, and—as we will show in chapter 17—a prediction machine’s data become so important that you may need to change your strategy to take advantage of what it has to offer.

C-suite leadership must not fully delegate AI strategy to their IT department because powerful AI tools may go beyond enhancing the productivity of tasks performed in the service of executing against the organization’s strategy and instead lead to changing the strategy itself.

Uncertainty has an impact on a business’s boundaries.s1 Economists Silke Forbes and Mara Lederman looked at the organization of the US airline industry around the turn of the millennium.2 Major airlines like United and American handled some routes, while regional partners like American Eagle and SkyWest dealt with others. The partners were independent businesses that had contractual arrangements with the majors. Absent other considerations, the regional airlines typically operated at a lower cost than the majors, saving money on salaries and less beneficial work rules. For instance, some studies showed that senior pilots at the majors received 80 percent higher pay than those at their regional partners. The puzzle is why majors rather than regional partners handle so many routes, given that partners can deliver the service at lower cost. Forbes and Lederman identified a driving factor—the weather—or, more specifically, uncertainty about the weather. When a weather event is out of the ordinary, it delays flights, which, in the tightly networked and capacity-managed airline industry, can have ripple effects throughout the entire system. When the weather goes sour, major airlines do not want to be hamstrung by partners checking their contracts when they have to make fast changes with uncertain costs. So, for routes where weather-related delays are likely, the majors retain control and operation. The three ingredients we highlighted in the previous chapter suggest that AI migh...

It could well be that up-front prediction gives airlines and automakers the confidence to allow for more complex arrangements and products. It is not clear what the impact on outsourcing would be since better prediction drives more outsourcing, while more complexity tends to reduce

When performance measures change from objective (are you keeping the bank queues short?) to subjective (are you selling the right products?), human resource (HR) management becomes more complex. Economists will tell you that job responsibilities have to become less explicit and more relational.

The forces affecting capital equipment also affect labor. If the key outputs of human labor are data, predictions, or actions, then using AI means more outsourced contract labor, just as it means more outsourced equipment and supplies. As with capital, better prediction gives more “ifs” that we can use to clearly specify the “thens” in an outsourcing contract.

If the prediction machine is an input that you can take off the shelf, then you can treat it like most companies treat energy and purchase it from the market, as long as AI is not core to your strategy. In contrast, if prediction machines are to be the center of your company’s strategy, then you need to control the data to improve the machine, so both the data and the prediction machine must be in house.

But what does the notion of AI-first mean? For both Google and Microsoft, the first part of their change—no longer mobile-first—gives us a clue. To be mobile-first is to drive traffic to your mobile experience and optimize consumers’ interfaces for mobile even at the expense of your full website and other platforms. The last part is what makes it strategic. “Do well on mobile” is something to aim for. But saying you will do so even if it harms other channels is a real commitment.

Two economists—Lesley Chiou and Catherine Tucker—studied search engines that took advantage of differences in data-retention practices.8 In response to the EU’s recommendations in 2008, Yahoo and Bing reduced the amount of data they kept. Google did not change its policies. These changes were enough for Chiou and Tucker to measure the effects of data scale on search accuracy. Interestingly, they found scale didn’t matter much. Relative to the overall volume of data that all the major competitors used, less data did not have a negative impact on search results.