How to build the AI platform your company needs

Why it matters

This article explains why companies building their own AI features need an internal platform, using DoorDash as its central example. Swaroop Chitlur and Siddharth Kodwani described in a talk how, after DoorDash signed its first OpenAI contract, teams began duplicating the same technical plumbing, prompting the creation of a shared platform layer that now serves thousands of internal users and handles enormous daily agent activity. The piece draws a key distinction between using consumer products like ChatGPT or Claude, where OpenAI or Anthropic bears responsibility for model behavior, and shipping AI features to customers, where the business chooses the model, controls what it says, and answers for mistakes. It cites the Air Canada chatbot case, in which a tribunal held the airline responsible for its bot's false promise, to illustrate that liability risk. The article organizes a platform around access, trust, context, and people, and positions DoorDash, which owns Wolt and Deliveroo, as a rare company that has shared its AI journey publicly, including missteps.