How Sharing AI’s Time Savings Drives Productivity

Joe Manok, VP for Advancement, Clark University.

In 1926, Ford Motor Company publicly announced its five-day, 40-hour workweek without reducing weekly pay. Workers had fought for that formula long before Ford’s announcement, but the anniversary offers a reason to revisit a century-old arrangement that has proven remarkably durable.

At Clark University, where I lead fundraising and alumni engagement, we tested whether giving employees a stake in productivity gains could help produce those gains in the first place. Employees would share in the gains through time off, while the university would see stronger results. The larger question for leaders is whether sharing productivity gains with employees can create a stronger incentive to find gains in the first place.

When an employee uses AI to finish a day’s work in an afternoon, for example, a manager can usually give them more to do. That makes sense for the organization, but it creates an incentive problem for the employee: If greater efficiency simply produces more work, why invest the effort required to find the next efficiency?

Ron Shaich’s book, Know What Matters, helped shape our thinking about incentives. Under his leadership, Panera developed a Joint Venture Partnership Program, which he described as “a kind of company-owned franchise model that made store managers and assistant managers part-owners of the cafes they managed.” For leaders, the principle is broader than a four-day week: If employees help create productivity gains, consider whether they should share in some of the value created.

After two years of productivity gains, our progress had plateaued despite process changes and new AI tools.

We put that proposition to the test within our division. From August 1 through September 19, we completed a 50-day experiment called “Project Rouge,” named after Ford Motor Company’s River Rouge complex. The formula was 100% of pay, 80% of traditional working hours and a target of at least 120% of baseline output.