DoorDash Deploys AI Analyst for 10,000 Employees
DoorDash’s new AI agent Vera answers 10,000 employees’ business questions with 90% accuracy after reorganizing data.
In brief
DoorDash’s new AI agent Vera answers 10,000 employees’ business questions with 90% accuracy after reorganizing data. DoorDash improved AI accuracy by organizing data instead of upgrading models. Originally reported by Pymnts.
Vera Answers Business Questions
DoorDash built Vera to answer employee questions like why orders dipped in California.
“Vera answers questions in conversation using company data.”
Vera’s Testing Progress
Vera’s pass rate rose from 43% to 90% on 1,900 questions across sales, finance, and operations.
Data Challenges
DoorDash’s data spans 350 petabytes across 200,000 datasets with fragmented sources.
Hard Questions Remain
Vera struggles with cause-and-effect analysis and forecasts, showing plans for employee approval.
AI Performance Test
Vera passed 73% of questions versus 48% for off-the-shelf tools on the same AI model.
Finance Impact
Vera improved strategic finance questions with 90% accuracy after data reorganization.
In short
DoorDash improved AI accuracy by organizing data instead of upgrading models.
