Learn how DoorDash automated more than 35,000 calls per day.
reduction in average handle time
automated calls per day
weeks to scale
DoorDash saw an opportunity to drive down costs by automating the tedious process of calling merchants, which would also ensure customer orders were placed faster than before, leading to higher customer satisfaction. The Global Operations team at DoorDash leveraged Replicant’s state-of-the-art voice AI to place 100% of eligible phone orders to restaurants on behalf of DoorDash customers. Within six weeks of launching, Replicant’s Thinking Machine™ went from making zero outbound calls a day to over 35,000 calls.
George McConnnell Global Operations Manager at DoorDashReplicant was able to meet or exceed our goals across all major metrics but more importantly, our partner restaurants preferred the AI solution. During our initial pilot we spoke with numerous merchants who explained they preferred speaking with the Thinking Machine since the kitchen is often noisy and the robot is clear.
Within weeks after launching, the results spoke for themselves. DoorDash realized dramatic savings by moving away from offshore call center agents that had higher costs per task and per minute.
The food delivery business is subject to greater fluctuations in demand throughout the week and even day, with most orders being placed during meal times and on weekends. Order volume can fluctuate dramatically minute by minute. In order to be prepared for potential peaks in volume, many businesses overstaff call centers. This results in unused agent minutes when demand is low, and consequently agent turnover.
DoorDash no longer needed to pay for unused agent capacity, since they only pay for the minutes the Thinking Machine is on the phone. By creating elasticity in their costs, DoorDash not only minimized wasted dollars, but could serve more customers.
While a call center agent can only handle one call at a time, the Thinking Machine handles hundreds of calls concurrently. For the first time ever, DoorDash was meeting order volumes without skipping a beat — regardless of how demand changed throughout the day or seasonally throughout the year.
For DoorDash, it was important that the Thinking Machine’s performance could come close to that of human agents. Within the first six weeks, the placement rate of the Thinking Machine was 89% as effective as the offshore agents. By the third month, it was 94% as effective. Outperforming its human counterparts meant 16% fewer calls were escalated to DoorDash’s corporate customer service team.
Replicant accomplished this by continuously retraining its machine learning models. The Thinking Machine quickly became more effective at clearly and succinctly communicating customer orders to merchants. These improvements to the conversational experience reduced hangup rates and reengaged merchants that previously preferred to speak to a human agent.