“Despite the awesome predictive powers of AI, measured productivity growth has declined by half over the past decade.”
What’s going on here?
The impact of AI will take time until it is felt, argue Aja Agrawal, Joshua Gans, and Avi Goldfarb, co-authors of Power and Prediction: The Disruptive Economics of Artificial Intelligence. Right now, AI is confined to point solutions, and they call this phase — the introduction and proliferation of technology, prior to tangible benefits being realized — as “The Between Times.” What we have yet to see, they continue, are a “plethora of high-value solutions for AI.”
The two questions that need to be asked of AI include “What is AI really giving us?” and “If we are designing our business from scratch, how would we build our processes and business models?”
For starters, Agrawal and his co-authors advise, stop talking about AI as the avenue to “insights.” Insights “is a trigger word for us because it represents precisely this wrong way of thinking about how an AI advance will create value. For a new AI prediction, ‘insights’ is code for ‘we don’t know what to do with that prediction.’”
Instead, the key to delivering AI value is to understand the decisions that will benefit from predictions. The co-authors state that “AI prediction takes decision-making potentially out of the hands of the decision-maker — a “decoupling” of predictions from human judgement. “The machine does the prediction. The customer exercises judgement. All the customer has to do is judge whether the price is worth the benefit.”
Examples, they add, are telematics that provide insurance pricing based on minute-by-minute driving decisions. “The insurance company prices the risky behavior. The customer judges whether the behavior is worth it.”
Decoupling prediction from judgement “creates opportunity,” they say. “It means that who makes the decision is driven not by who does prediction and judgement as a bundle but who is best to provide judgement utilizing AI prediction. Once the AI provides the prediction, then the people with the best judgement can shine.”
For pondering AI-based transformation, Agrawal and his co-authors, advise, it’s best to start from a blank slate, looking at the decision first, then building the AI-based system. This consists of articulating the mission; reducing the business “to the fewest possible decisions required to achieve the mission if you had super-powerful, high-fidelity AIs;” and “specify the prediction and judgement associated with the primary decisions?