2020 hasn’t just challenged humans–even AI got frustrated

McKinsey & Company

This week, we look at how pandemic-related data shifts tripped up artificial intelligence, with implications for its predictive capabilities. Plus, South African banking faces new client needs, and a very fishy McKinsey project.

Photo of fiber network

Even AI was surprised by 2020. This year of crisis has trained a spotlight on the power of analytics and artificial intelligence. Across industries and geographies, analytics have enabled leaders to more effectively handle the challenges presented by these unprecedented times, from supporting and protecting workers to engaging increasingly digital customers and managing fragile supply chains.
Achilles’ heel. One of the most widely used advanced-analytics techniques—machine learning—relies on the principle that patterns and behaviors from the past will likely repeat in the future. However, recent data inputs reflect profound changes in human behavior: physical distancing in daily life, less travel, and altered spending habits to name just a few. For models built before the pandemic—ones that relied on historical data—COVID-era inputs might severely reduce predictive capabilities. And new and lasting patterns, such as higher consumer spending on digital channels, will likely emerge, invalidating or reducing the predictive power of pre-COVID-19 data as well.
Yes, but. Despite these issues, pre-COVID-19 models have tremendous capacity to give executives important insights to help them navigate the crisis and the next normal—if leaders are prepared to take the steps needed to shore up those models. The challenge here is not simply a technical one for data scientists to solve. While analytics professionals will play an important role, stabilizing critical models will depend equally on efforts from leadership to recalibrate business strategies for the changing landscape, forge new data partnerships, convene interdisciplinary teams with sufficient diversity, and more.
Speaking of diversity… While AI can reduce human bias, it can also subtly bake in and scale bias by earning our misguided trust in its fairness. To avoid this pitfall, we often rely on fairness proxies to evaluate algorithms. A model might be deemed “procedurally fair,” but does that necessarily equal outcome fairness in AI-based decisions? As we apply stricter scrutiny to AI, it’s worth considering how those fairness proxies can hold humans accountable, too. Running algorithms against human decision making is the best way to expose biases on both sides.
Postpandemic demand. Our new global survey, “The state of AI in 2020,” suggests that organizations are using AI as a tool for generating value. A small number of respondents from a variety of industries attribute more than 20 percent of their organizations’ earnings before interest and taxes (EBIT) to AI, and more than two-thirds of respondents who reported adopting AI in 2020 found that its adoption increased revenue. These companies plan to invest even more in AI in response to the pandemic and its acceleration of all things digital.

Aneel Bhusri photo

Workday’s founder on software and the next normal
Aneel Bhusri, co-CEO of Workday, the finance and HR powerhouse, talked with McKinsey about serving customers as the COVID-19 crisis began, data’s importance for employee diversity, and the limitations of remote work. “Being in person is really critical. We are social beings, and we generate great ideas, and I think we inspire each other,” he says. “So as we go through the next months of the COVID-19 crisis, I think people are going to be surprised and say, ‘No, we actually like office environments more than we had thought.’”

image of a fish

Down on the fish farm
If you’ve ever wondered what McKinsey consultants do all day, you’re not alone. Basically, we solve problems, like how to improve education, address climate change, or design new products. We recently wrote
this interactive story
to show kids (and adults!) an example of an interesting problem we recently helped solve.
It all started in a beautiful coastal region of Latin America, where aquaculture farms raise fish. Unlike farmed land animals, farmed fish are more difficult to monitor because they exist in larger numbers and grow at varied rates depending on their environment. Local farmers found fish tracking to be especially troublesome when deciding how much fish food to buy. A regional fish-food provider—McKinsey’s soon-to-be client—noticed this challenge. The food company wanted to be a better business partner to fish farmers by better estimating the necessary amount of food each aquafarm needed in a more customized way. Cue McKinsey.
While McKinsey is filled with all kinds of experts, from scientists to economists to writers, every expert has their own specialization. This ensures that each unique project is staffed with consultants that are best suited to the area of work. For this particular fish-farming engagement, a McKinsey team assembled experts in agriculture, business operations, sustainability, analytics, and more.
McKinsey consultants pooled their expertise to build an analytics model that could measure factors that influence fish eating habits like size, weight, and water temperature. This model would allow the client, the fish-food company, to do more than just sell food: the company could now provide accurate and customized estimates for how much food each aquafarm should purchase. After the model’s implementation, fish farmers saved money and eliminated waste, and McKinsey’s client saw increased revenues.
Check out how the work—and the fish—flowed here

— Edited by Caroline Dudlak

Have feedback or other ideas? We’d love to hear from you.

McKinsey & Company

Copyright © 2020 | McKinsey & Company, 3 World Trade Center, 175 Greenwich Street, New York, NY 10007


More Posts

Send Us A Message