In a little over a week—on Saturday, March 6—one of the world’s great sporting challenges will get underway: the 36th America’s Cup. And a McKinsey team played an important part in this long-awaited event.
This year’s competition requires teams to sail a 75-foot vessel. But it’s not a typical sailboat. Hydrofoils fixed to its hull lift the entire craft out of the water, “flying” one meter above it, enabling it to reach speeds exceeding 60 mph, or 100 kph. Most importantly, the hydrofoils are one part of the boat where this year’s Class Rule—the race protocol—allows for design modifications, which offer huge competitive advantages for teams that get those modifications right.
The partnership between McKinsey and defending champions Emirates Team New Zealand (ETNZ) began in 2019. What they needed, they soon realized, was a new kind of crew member that could sail thousands of boats at a time—and never need to set foot on shore. The answer to that call was an AI bot, or software robot, that could test new hydrofoil designs by sailing them on Emirates Team New Zealand’s simulator.
The simulator had been key to the team's victory in 2017; the sailors had used it to test new boat designs without having to actually build them. But that simulator required multiple team members using it simultaneously for it to work properly. This was a logistical challenge, given the sailors’ scheduled practices, travel, and competitions.
As Forbes detailed in a recent piece, Firm colleagues helped the team develop a “digital twin” — a digital replica of a sailor able to control the simulator optimally and automatically run simulations before physical devices were built and deployed. The team included colleagues from QuantumBlack, McKinsey’s analytics firm, as well as highly experienced sailors. London-based Jacomo Corbo, QuantumBlack’s co-founder and Chief Scientist, said to Forbes, “McKinsey’s digital twin represents one of the most ambitious and complex applications of deep reinforcement learning ever.”
He went on to explain: “Prior to the development of the digital twin, testing new boat designs required the team’s sailors to perform test runs in a simulator, which was inefficient and costly. McKinsey’s digital twin outperforms the sailors using the same simulator, speeding up testing and the rate at which designs are optimized by a factor of nearly ten.”
As a first step, the team turned to the cloud. Over six weeks, data, analytics, and machine-learning experts from Sydney, Melbourne, and London migrated New Zealand’s simulator and developed an infrastructure to run it in the cloud. The analysts “taught” an AI bot how to sail. After a number of months, the bot began outperforming the sailors in the simulator.
Soon, the sailors were learning maneuvers from the bot, and the bot also began to detect granular errors in the simulator itself, down to the behavior of a single water molecule.
Over the course of 2019 and early 2020, the bot sped up Emirates Team New Zealand’s design process by a factor of ten. And, although the race is just about to kick off, this project has already proven that reinforcement learning can be a transformational tool for process design, with potential applications across industries.