Gopuff launches Go, a Grok-powered AI agent that takes voice, photo and text

Gopuff and xAI have launched Go, an AI agent built into the instant-delivery service's app. The shopping assistant runs on multimodal Grok models and accepts requests in any format: text, voice or a photo.

What the Go agent can do

Go is an AI assistant integrated directly into the Gopuff app interface. Its goal is to rethink how people search for and choose goods for fast delivery.

Traditional e-commerce search demands precision: the shopper must know the brand, the product name and sometimes the item number. Go works differently, because it understands context and intent. A user can type "something sweet to go with tea this evening," dictate a shopping list by voice or photograph an empty fridge shelf. The agent parses the request, matches it against the Gopuff catalog and suggests specific options that can be added to the cart right away.

Three Grok models run under the hood:

• Text: understands free-form requests and takes the shopper's context and intent into account
• Audio: recognizes voice commands without a separate transcription step
• Vision: analyzes photos of products, packaging and labels

Combining modalities lets Go fit different scenarios, from quick voice input while cooking to a photo of the fridge's contents late at night.

Why this matters for Gopuff

Gopuff is one of the pioneers of quick commerce in the US. The company built a network of its own micro-warehouses in dozens of American and British cities and specializes in delivering goods in 15-30 minutes. Unlike classic aggregators such as DoorDash, Gopuff keeps inventory in its own warehouses, which gives it more control over speed and price.

Many quick-commerce purchases are impulsive. Shoppers don't want a specific product; they want a problem solved: "what to snack on," "what to clean the floor with," "what to drink tonight." Traditional search handles this poorly, returning literal results rather than contextual ones. Language models cope with vague queries far better.

Beyond convenience, an AI agent is a tool for growing the average order value. It can recommend complementary products, suggest substitutes for items that are out of stock and personalize offers based on purchase history.

xAI moves into the enterprise market

For xAI, the partnership with Gopuff is one of the first publicly announced commercial use cases for Grok outside the X platform (formerly Twitter). Elon Musk's company launched Grok in late 2023 as a ChatGPT competitor, but for a long time the model was available only to X Premium subscribers. By embedding Grok in a third-party app, xAI shows it wants to become a B2B platform, following the model of OpenAI, which sells API access to GPT to thousands of companies.

This is a shift in positioning: xAI is no longer just a "competitor to Anthropic and OpenAI" and is claiming the role of AI infrastructure supplier for business.

Gopuff's competitors are already moving in the same direction. Instacart launched AI search together with OpenAI, and Amazon is embedding its Rufus voice assistant. Against that backdrop, choosing Grok over GPT is a deliberate bet on differentiation.

What this means

An AI agent in an instant-delivery app signals that language models are becoming a standard component of consumer products. If Go delivers a real increase in conversion and average order value for Gopuff, it will accelerate the race: similar solutions will appear in other retail apps much faster.

For xAI, this is the first serious step toward turning Grok from a public chatbot into a commercial B2B platform.