DoorDash makes its AI grocery reorder 8 times faster

DoorDash has cut the wait for its AI grocery reorder from about 16 seconds to 2 seconds. The trick is simple. The AI no longer studies your past orders when you tap. It does that work earlier, so the list is ready when you ask. DoorDash says views of shopping lists doubled and orders from this flow rose about 34%.

What changed

The feature lives inside Ask DoorDash, the company's chat-style shopping assistant. Before, the system rebuilt your likely recurring purchases with each request. Now it prepares them in advance. InfoQ reports the result as a drop from 16 seconds to 2 seconds, which is 8 times faster.

How it works

DoorDash looks at 90 days of your purchase history, along with signals from your profile. A large language model, the kind of AI behind chatbots, runs in batches to spot items you likely buy again and again. Then a set of fixed checks reviews the results. These checks follow set rules and give the same answer every time. Only then is the list saved.

What stays live

A saved list could go stale, so DoorDash checks it against the current catalog when you open it. For each item, the system first tries to find the exact store item you bought before. If that fails, it tries a product identifier from the catalog. Last, it falls back to a search term. This helps when listings change or a product is out of stock.

The AI is still used when a request needs real thinking. One example is finding cheaper options based on prices, promotions, package sizes, inventory and your preferences. Simple edits, like changing a quantity or removing an item, update the list directly with no new AI call. DoorDash treats your own edit as the final word.

Why it matters for business

A DoorDash engineer, Twisha Jain, put the lesson this way: better AI systems are not always about more reasoning. They are about deciding what belongs in the live loop and what can happen ahead of time. For any business using AI tools, the same idea applies. Slow work done in advance can make a tool feel quicker, and people may use it more.

DoorDash also reports that adding a new Reservations agent on its shared platform took one week. That is about ten times faster than building its first agents of this kind.

What is not known

The material does not say how often the prepared lists are refreshed, what the approach costs, or how it compares in speed with rivals such as Instacart, which also uses AI for shopping help. The 34% and doubled-views figures are reported by DoorDash.