How Zipcar and DoorDash leaned on A.I. during the pandemic

Alok Gupta, head of data science and machine learning at DoorDash, speaking at Fortune Brainstorm A.I. in Boston. Nov. 8, 2021.
Rebecca Greenfield for Fortune

When people started working from home and changing their consumption habits during lockdown, Zipcar and Doordash used A.I. to determine their next moves.

Freedom Dumlao, chief technology officer of Zipcar, said that the company’s short term car rental patterns changed dramatically last year. Before the pandemic, people would typically rent a car to go to the office and do chores. After the initial standstill of lockdown, the data showed people would rent a car in the middle of the day. 

“All of a sudden, everyone was driving again….for us, that came as a shock,” Dumlao said at Fortune Brainstorm A.I. in Boston on Monday. “Once we saw it happening, we noticed right away the types of drives and driving behavior were very different.”

Before the pandemic, the Boston-based car sharing company had made significant investments in its system to position vehicles in areas and at times when it predicted demand, Dumlao said.

That investment helped Zipcar to meet the moment. 

They were also able to “leverage comparable models, for example holidays and vacation weeks,” which Dumlao said “helped us bridge quite a bit.”

Alok Gupta, head of data science and machine learning at Doordash was also busy figuring out how to best serve customers, restaurant owners trying to stay afloat, and gig workers who were delivering orders during lockdown. 

On the customer side, he noticed “a really big demand from users to access grocery and convenience items.” 

For that, Gupta said DoorDash used A.I. to predict inventory and also recommend substitutions, which is a problem that typically doesn’t occur if someone is ordering takeout from a restaurant.

The company also leaned on computer vision to dynamically update the opening and closing hours of a restaurant. 

When a DoorDasher went to a restaurant and saw it was closed, the could submit a photo to verify the hours. Computer vision was able to also read the updated hours on the door and dynamically update them on the app without human intervention.

“We are able to improve our hyperlocal understanding,” Gupta said.

Other uses include telling DoorDashers the best place to park when they’re picking up an order. While this benefits the customer, Gupta also pointed out that it helps the hard-working delivery people “shave five seconds here, ten seconds there” so they’re able to work more efficiently.

More tech coverage from Fortune:

Subscribe to Fortune Daily to get essential business stories straight to your inbox each morning.

Read More

Brainstorm HealthBrainstorm DesignBrainstorm TechMost Powerful WomenCEO Initiative