DoorDash Rolled Out Claude Code to Every Employee: Small Teams and Architecture Principles Drove a Throughput Surge

As DoorDash marked a major milestone rolling out Claude Cowork broadly, the aim was to raise the floor of AI fluency across the whole company. People outside engineering are now using the terminal, coding, and shipping real work -- while even executives who understand AI's importance often remain stuck at the 'it's just a chat' stage. Many people's eyes light up the moment they realize connecting it to Gmail, Google Calendar, or Slack makes them more efficient knowledge workers -- and that's precisely the effect the Cowork rollout was designed to achieve.

Company-wide Claude adoption has produced dramatic throughput gains. In one striking example, a massive code migration that would historically have taken four engineers a full quarter was completed solo, in just three weeks, by a single engineer. Multiple other major products cut their timelines by more than half, pulling release dates in significantly. The team is now hunting for these pockets of success, analyzing what made them work, and giving a handful of teams a deeper charter to see what they can produce.

From these adoption experiments, one clear lesson emerged: keeping teams small is one of the biggest accelerants, since each person owning a clear domain sharply reduces cross-team coordination costs. Another important learning: it genuinely pays off to invest upfront in making a codebase agent-friendly. In one successful example, the tech lead had written down 50-plus architectural principles they cared about.

For engineering leaders, the advice is to build a culture of sharing written artifacts -- the most scalable way to spread learnings across an organization, with the added byproduct of accumulating assets agents themselves can read. New grads are picking up this new paradigm remarkably fast, and the encouragement for leaders is simple: stay curious, and actually play with the tools yourself.