Over the weekend, Vivek Wadhwa published a perspective in the Washington Post under the headline “Don’t believe the hype” about AI. I sat down, prepared to agree on everything, because that’s an angle I can usually rally behind. However, with every new sentence, I realized I had a difference of opinion. I had more than 500 words just responding to the title and the first sentence. So I’ll cover those two, and then call it a day.

The title “Don’t believe the hype: Artificial intelligence isn’t taking over business decision-making” is a broad sweeping statement designed to lure you in, but that doesn’t make it true. There are many places where computers without AI have already taken over business decisions, and others where AI is very much automating business decisions.

Just a few decades ago, large companies had to plan and manage inventory replenishment on ledgers. These inventory replenishment decisions were performed by humans without the assistance of computers. But today, there are mature commercially available software solutions that have automated this process. The rise of ERP systems in the 1990s, combined with algorithmic implementation of operations research, brought a huge wave of business process optimization and automation. Computing has become so pervasive that we sometimes forget how much automation has taken over human decisions before the era of modern business computing.

Now you may say the example above is not AI. Exactly. Computers have been slowly taking over human tasks and decisions for more than 30 years. It just happens that the progress on various forms of AI are enabling us to solve problems that previously were impossible or inadequately solved.

Let’s move on to the first sentence of Wadhwa’s article.

While AI systems can now learn a game and beat champions within hours, they are hard to apply to business applications.

Are you sure about that? AI has taken over extraordinarily complex decisions is in the world of finance. The bots now rule trading and only 10% of volume comes from human discretionary investors, according to JP Morgan. Goldman Sachs’s NY headquarters used to employ more than 600 human equity traders buying and selling stock. Now they employ just 2, according to a recent MIT Technology Review article.

There is a seemingly infinite permutation of things that can influence global market moves. Despite this enormous complexity, the bots seem to be doing just fine at outperforming their human counterparts. While inventory management is complex and has many external variables, is it more complex with more variables than the global equities market?

I would certainly challenge the assumption that most business problems can’t be turned into a game. The financial world is not limited to two players, and it’s clear who is winning and losing in that game.

The game of business

In fact, many business processes can be modeled in terms of a game. In business, ‘score’ is defined by KPIs such as profitability, revenue growth or cost efficiencies. The players’ moves are called operational levers or business process actions. In a videogame, if you haven’t made the right moves, you die. In the business world, if you haven’t made the right decisions, you’re probably out of business.

In business there are many ‘games’ simultaneously being played by many different people. Product demand is influenced by both the marketing game and the pricing game. In turn, this impacts the manufacturing schedule and planning games. The interdependencies between these games is extraordinarily complex and not expressed by a simple set of linear equations. The human brain is not capable of playing all these games at once, nor predicting the future based on the past, nor calculating complex sets of probabilities across various games. This is why AI is changing business decision making: intelligent systems powered by AI can play hundreds of ‘business games’ simultaneously, while also factoring in the complex interdependencies between the games. The result is the highest possible ‘score’ for the organization.

Without AI, one business unit might optimize for its own KPIs, but accidentally end up sub-optimizing the whole organization. This is sometimes referred to as the efficiency paradox, where optimizing metrics in isolation without considering the interactions between metrics can be detrimental to the whole. The GM ignition switch scandal is an example of this that resulted in 100 deaths, where the fix was $1 per car but lead to criminal charges and a $900M USD settlement. Optimizing for component cost didn’t optimize for organization risk or the cost of human lives lost. Intelligent systems powered by AI hold the promise of understanding these interconnections and making significant progress on optimizing the whole in addition to the parts.

There are many business problems that are much more well-defined with a smaller set of external factors than global financial markets. What’s to stop someone for solving one of these much easier business problems? Nothing. My company is already doing that. We are also solving the much harder problem of complex interdependencies across an entire global enterprise to maximize organizational performance.

A perhaps more useful response to the hype around the fear of AI taking over business decisions is not to tell people they don’t have to worry. Rather, explain where we will feel AI’s influence and where we might not feel it. Then we can thoughtfully step back from the hype and the anti-hype, and plan for the changes headed our way.