DoorDash’s AI Push: From Chatbot to Text Agent, Betting on Conversation to Drive Orders

DoorDash once built its business on the promise of convenience. Tap an app. Pick a restaurant. Food arrives. Yet even that simple flow started to feel like work. Customers grew tired of scrolling menus. They wanted to say what they craved and have the right items appear. So the company responded with AI.

Last June it rolled out Ask DoorDash. The in-app chatbot let users type natural language requests, upload recipe photos or describe a mood. Results followed quickly. TechCrunch reported that nearly half of restaurant orders placed through the tool went to places the customer had never tried before. Grocery carts built with it ran more than 35% higher in value. Users finished them five times faster than through standard search.

Co-founder Andy Fang traced the idea back to 2023. The models simply weren’t ready then. By mid-2026 they were. DoorDash combined models from OpenAI, Anthropic and Google with its own data on orders, preferences and merchant catalogs. The result felt less like search and more like a helpful friend who already knew your tastes.

But one chatbot inside the app wasn’t enough. On September 30 DoorDash went further. It unveiled a text-to-order AI agent that works inside Apple Messages. No app opening required. Just send a message like “order my usual protein bowl to the office” and the system matches your phone number to your account, reviews history, builds a cart, shows food photos and checks out with saved payment details.

Bloomberg noted the move positions DoorDash ahead of rivals Uber Eats and Grubhub in the race toward personal AI agents. These agents promise to handle tasks without forcing users to navigate apps or websites. DoorDash opened a waitlist for U.S. customers. Early tests suggest people order more frequently when friction drops this low.

The original article from The Information had laid out the case months earlier. Consumers increasingly expect conversational interfaces. Traditional search demands specificity. AI handles vagueness. It surfaces options for “filling dinner for a family of 4” or “kid-friendly vegetarian spots with mild options.” Then it refines. The piece argued DoorDash could not afford to lag as chat-based commerce spreads.

So far the numbers back the bet. Ask DoorDash helped users discover more than 40,000 restaurants in its first three months. Grocery orders placed via the assistant often involve routine restocking. DoorDash added one-tap reordering and suggestions for cheaper alternatives to speed that up. One analysis showed consumer memory features improved grocery checkout conversion by about 24% and lifted basket sizes 17%.

Yet the rollout isn’t limited to hungry customers. DoorDash also launched DashBuddy, an AI assistant for its delivery workers. New Dashers can text questions over SMS and get answers instantly. More than 200,000 have already used it. The tool resolves around 95% of signup and first-dash questions. Those who chat with DashBuddy before their initial delivery prove more likely to complete one within three days.

And. The company didn’t stop at consumers and Dashers. It introduced tools for enterprise AI assistants to place bulk orders. Corporate Slack bots or other agents can now query catalogs, assemble group carts that respect mixed dietary needs and quantities, then track deliveries. All without anyone touching the DoorDash app.

Engineers at the company described their architecture in technical posts. An orchestrator coordinates specialized agents for restaurant discovery and grocery tasks. Persistent memory stores user preferences across conversations. The system avoids calling large language models for every deterministic action to keep latency low. Confirmation steps guard against mistakes before carts reach checkout.

Early user tests produced mixed but mostly positive feedback. One Business Insider reporter snapped a photo of a recipe and received a complete grocery list for borscht from nearby stores. Another asked for shrimp ideas and got five distinct recipes. Accuracy isn’t perfect. DoorDash warns users to review carts. Prices in messages sometimes differ from the app. Still, the speed and personalization impress.

Competition looms. Other delivery platforms experiment with their own AI features. General-purpose agents from big tech could eventually handle food ordering as one small capability among many. DoorDash counters with its massive data advantage. Thirteen years of delivery records inform recommendations, routing and even new hardware like its delivery drones announced alongside the text agent.

Those drones, tested first in Northern California with partners including Chipotle and Popeyes, aim to cut delivery times to under five minutes for many orders. The AI ordering tools complement that vision. Faster discovery and checkout pair with faster physical delivery. The entire loop tightens.

Wall Street has taken notice. DoorDash shares reacted positively to the September announcements even as the broader market wavered. Analysts see potential for higher order frequency and larger baskets to lift revenue. Merchant adoption could grow too if the assistant drives traffic to smaller restaurants that customers might otherwise overlook.

Challenges remain. Hallucinations in recommendations could frustrate users or merchants. Privacy questions arise when phone numbers link directly to order histories in messaging threads. Regulatory scrutiny of AI in consumer apps continues to build. DoorDash says it monitors performance closely and iterates based on feedback.

For now the strategy looks clear. Turn every interaction into a conversation. Reduce steps between hunger and delivery. Collect even richer data in the process to train better models. The original thesis from The Information has been tested in market. Results suggest it was right. DoorDash didn’t just need an AI assistant. It needed several. One for customers. One for Dashers. One for enterprises. And it built them.

The next test will show whether these tools move the needle on retention and spending enough to justify the engineering investment. Early signals point yes. Customers complete orders faster. They try new places. They order more often through text. Dashers start quicker and stay longer. If those trends hold, conversational commerce may become the default way people think about dinner. DoorDash intends to own that shift.