DoorDash Deployed Internal AI Agent for Data Analysis

The internal agent, named Vera, has been deployed to 10,000 employees to parse 350 petabytes of company data.

Updated on Sept. 21, 2026 in Artificial Intelligence

Isometric editorial illustration of a clean, minimalist terminal pedestal in an architectural office space, representing enterprise data analysis. — DoorDash has launched an internal AI agent named Vera to assist 10,000 employees in analyzing over 350 petabytes of corporate data. AI Illustration. Upload story photo >

Live Poll

Do you trust AI agents to accurately analyze data for important business or financial decisions?

DoorDash has deployed an internal AI agent named Vera to 10,000 employees for querying corporate data. This live internal deployment follows performance testing where the agent achieved a 90% pass rate.

Why it matters

The deployment addresses the fragmentation of the company's 200,000 datasets by providing a tool tailored to internal structures that off-the-shelf models fail to navigate. It underscores a broader trend in enterprise automation to handle large-scale, messy internal data.

Vera achieved a 90% pass rate across a test set of over 1,900 questions, significantly outperforming a standard off-the-shelf agent which scored 48%. The agent manages 350 petabytes of data across 200,000 distinct datasets.

The players

DoorDash

A logistics and food delivery platform that manages a massive 350-petabyte stack of internal data across hundreds of thousands of datasets.

The details

DoorDash organizes its 350 petabytes of data specifically for the model to improve performance against the company's unique information architecture. Vera utilizes a two-stage evaluation process in which a secondary AI model grades the primary agent's responses against expert-verified answers. For complex planning tasks, the system requires human approval before executing any data-driven actions.

September 2025: 7% of U.S. enterprise CFOs used AI agents in finance workflows.
September 21, 2026: Official article publication regarding the Vera deployment.

The Tech Race

DoorDash's focus on tailoring models to messy internal datasets highlights the competitive shift away from general-purpose LLMs toward domain-specific agentic tools. The company is now attempting to bridge the gap between simple chatbots and autonomous analysis agents for complex corporate planning.

This deployment currently affects 10,000 DoorDash employees who use the tool to query company data without manual intervention. The agent is strictly internal and requires human verification for complex tasks, limiting its impact to established corporate workflows.

The takeaway

The deployment demonstrates that success with large-scale internal data requires specialized staging rather than relying on generic AI models. Future developments will focus on the agent's ability to navigate the most complex analysis tasks, marking a key milestone for the company.