DoorDash published a conversation with Mariana Garavaglia, its Chief People Officer, about how the company is using AI so teams can learn, automate, and create solutions without always relying on engineering. The central idea? AI isn't here to replace people, but to expand human capacity in everyday and strategic tasks. (openai.com)
What DoorDash said and why it matters
In the interview, Garavaglia explains that DoorDash approaches AI adoption in three layers: access and literacy for every employee; integration of internal data to break down silos; and exploration of agents that can execute tasks reliably. That strategy aims not just to automate, but to improve workflows and the work experience. (openai.com)
DoorDash is rolling out ChatGPT Enterprise in technical and non-technical areas and uses the OpenAI APIs on its customer support platform, which processes millions of chats each month. It's a clear signal: large companies are already embedding AI into core operations, not just isolated experiments. (openai.com)
"The goal is not just to automate but to empower workflows so people have a better day-to-day." This sentence captures the human-centered approach the company describes. (openai.com)
Practical examples that explain the change
Can you imagine someone in HR building a small tool that would previously have required the engineering team? At DoorDash they tell the story of a People Ops person who created a script to upload documents automatically—something they used to do one by one. That kind of democratization of creation is what changes productivity in practice. (openai.com)
Think about Venezuela: if your small business could automate invoicing or payroll sheets with a template generated by AI, you'd save time and reduce errors. It's not science fiction—these are repetitive flows AI can ease today.
How they measure impact and literacy
DoorDash starts with simple metrics: adoption and frequency of use. They also analyze license distribution and how to include literacy in performance frameworks—for example, valuing willingness to adopt new tools. That's a practical way to turn usage into a recognized skill. (openai.com)
They also recognize that different functions progress at different paces: engineering goes first, other teams join later. The idea is to use durable metrics while tools evolve. (openai.com)
Limits and safeguards: AI that supports, not decides
Garavaglia insists AI should expand human judgment. Concrete examples: synthesizing feedback in performance reviews and analyzing surveys to identify themes and generate action plans. That way, managers receive actionable summaries, but final interpretation remains in human hands. (openai.com)
They are also developing predictive models to support decisions like promotions or executive hires, but as assistance, not verdicts. That blend of data and human judgment is key to minimizing bias and error. (openai.com)
What's coming in the next 12 to 24 months
HR teams are exploring agents for tasks like answering internal policy questions, supporting professional development, and helping managers with operational work. The big bet is mass personalization: development plans sized to each employee, instead of one-size-fits-all programs. (openai.com)
For organizations without dedicated engineers on the People team, Garavaglia emphasizes that technical capacity is an accelerator: it lets you move from strategy to rapid delivery. But she also stresses there are ways to start without large tech teams, for example with pilots and internal training. (openai.com)
Final reflection
The practical lesson is simple: AI is no longer only for specialists. If you use it well, it can turn repetitive tasks into time for higher-impact work. What can you apply tomorrow in your team? Start by identifying a recurring process that steals your time and imagine how an automated flow or an AI-generated summary could free up hours.
DoorDash shows a path: access, well-integrated data, and human care in decision-making. It's not magic—it's organized design so technology helps people do their jobs better. (openai.com)