That hasn’t stopped the economist and management professor from writing two books on the subject. In 2018, Gans and fellow economists Ajay Agrawal and Avi Goldfarb published Prediction Machines: The Simple Economics of Artificial Intelligence, which made waves for illustrating how companies can use predictive A.I. to make better decisions and ultimately transform how they do business.
One provocative thought experiment they sketched out is the idea that online retailers might use predictive A.I. to proactively ship customers items before they’ve even purchased them (a “ship-then-shop” model).
In the years that followed, the authors realized that while ideas like “ship-then-shop” demonstrate the enormous potential of prediction, they didn’t go far enough in explaining how entire systems would need to adapt A.I. to provide a return on the investment, or R.O.I. To get an R.O.I on A.I., business leaders need to take steps smartly and understand we’re in “The Between Times,” Gans says.
Meanwhile, many companies successfully adopted the tips they’d laid out but faced unforeseen challenges. In other words, they argue, we’re in The Between Times—“after witnessing the power of this technology and before its widespread adoption.”
To help leaders roll out A.I. strategies, they published a follow-up last fall: Power and Prediction: The Disruptive Economics of Artificial Intelligence.
If their first book sparked imaginations, Power and Prediction feels like a dose of realism: The book walks readers through the organizational steps they will take when building their own A.I.
The firms who take those steps smartly, write the authors, will remake our economy on the level of industrialists such as Henry Ford during the early 20th century.
How, with the work of the predictions being done by A.I., can people make decisions in a better and faster way?
A lot of people get very hooked on the term artificial intelligence, making it sound like you’ve got a new thinking machine available. When people talk about the goals of artificial intelligence, it is broadly to create that. However, the important thing to note is that A.I. is developing just one aspect of intelligence.
We have to do a lot of things to think, but one of the things that we do is use our senses to look at the world around us and convert all that information into what we might need for any particular thing we’re doing. The term for that is prediction, and that’s a statistical term. But it also encompasses things that you’d typically associate with the common usage of the term prediction, such as being able to forecast the weather or predict demand conditions in your industry.
All these advances in A.I. are an advance in statistics, not creating a new intelligence. Advancing statistics does not sound exciting, but there is an awful lot that can be better at predicting, even if it is just advanced statistics. There’s value in understanding that this is just making a prediction, not over-hyping it. It’s a mistake to overhype things.
What does that mean for business?
Often, we’re making decisions in which we’ve got a lot of uncertainty. Therefore, you can make better decisions by developing predictions and resolving some uncertainty. It enables you to consider many actions that might have been unpalatable if you needed a sharper forecast of what could happen.
[Without a forecast], you might think, “I don’t know what to do, but this choice will minimize the cost of mistakes, so I’ll play it safe.” That involves doing fewer things. There’s something freeing about uncertainty: you can say, “Well, I can’t tell what’s going to happen here, so I might as well do the same thing all the time.” But there’s an additional richness in your decision-making that comes from having access to more information. That’s essentially what artificial intelligence is doing.
What would you say to a business leader who is generally aware of the importance of A.I. but is also aware that they’re behind the eight ball?
Our first book looked at simple applications of A.I. that we regard as “point solutions.” In other words, someone’s making a decision, and there’s uncertainty. You then imagine, “If I can solve or reduce this uncertainty, will this be of value to me?” and choose to develop that point solution. A lot of good businesses are doing that.
We’ve learned that very few decisions are independent of other decisions. When making predictions based on predictions, think about how that looks to somebody else in the organization. They don’t see the prediction you’re using. They just see you’re doing a lot of different stuff, and if they try to work with you, it’s like you’ve become some sort of unruly child.
So not only do you have to adopt this technology, you have to work out whether it will be too disruptive elsewhere. This is more than just a problem that occurs with artificial intelligence. The initial impacts have been limited when we’ve had a big technological change, such as electricity, computers, or the Internet. Initial take-up is very, very slow.
With the benefit of hindsight, we took advantage of electricity when Henry Ford worked out he could have a production line. And that was forty years after Edison. The characteristics of this feel a lot like electricity. It feels a lot like the computer revolution.
Where do you see A.I. being used in a predictive way most effectively now?
In a supply chain business, one of the things that people are very interested in is the quality of the product that’s being produced. In the case of mining oil and gas, for instance, if you have a prediction, you can assess the oil quality beforehand and take some remedial action. That, in turn, reduces the cost of refining the oil, then enhances the efficiency of the places using the oil.
In a service industry, you may be dealing with a particular customer. You know some general things about the customer. Still, with additional data gathered, you could learn more and refine or personalize the product you’re selling to that customer.
Are these opportunities readily available? Not necessarily. Data has to be gathered. But in some fields, such as healthcare, much of this data is generated during healthcare practice. It is a matter of using that data along the way to make better diagnoses and predictions of individual reactions to certain treatments.
Do you think business leaders are intimidated by A.I.?
I suspect that most business leaders are aware of A.I., and their boards of directors are saying they should look at it.
It will require asking, “What is the biggest thing that I don’t know in my industry—-something that, if we knew it, would change everything?”
Imagine you’re starting with a clean slate. How would you build the organization? Then you see how different it looks from what you’re currently doing. It’s a pure opportunity for strategic thinking.
Replies to This Discussion