Gopuff teamed up with SpaceXAI to launch Go, an AI shopping assistant powered by Grok, designed to learn shopping habits and suggest or add items automatically.

Gopuff and SpaceXAI have built Go, a Grok-powered shopping agent that aims to shrink the time you spend deciding what to buy. The tool is built to remember your recent purchases, pick up on eating patterns, and suggest carts without a lot of input from you. It launched on June 3rd and is already available inside the Gopuff app for users to try. This partnership signals another step toward agent-driven commerce where software does more of the thinking for shoppers.

Go is more than a recommendation engine; it actively adds items to your cart based on memory and inferred needs. The assistant can suggest meals or entire carts tailored to budgets, dietary preferences, or time of day. It also watches for context cues like weather or calendar events to pick moments when a suggested cart would be useful. The interface emphasizes mood and presentation, so suggestions can come with visuals and a vibe meant to match the occasion.

https://x.com/axios/status/2062102219084706297

The design goal is to reduce decision friction, the brief pause before you commit to a purchase that often stalls commerce. Instead of digging through categories or searching keywords, Go should surface practical options—think quick breakfasts, protein-heavy lunches, or low-sugar snacks. Users can ask for niche items such as gluten-free options and receive curated picks without endless scrolling. Voice control and conversational prompts make it behave like a kitchen assistant rather than a cold catalog.

From a technical angle, Grok supplies the underlying AI agent that lets Go learn and remember at the user level while making suggestions at scale. That memory includes recent buys and inferred habits, so the assistant can predict what you’ll need next. The system also factors in external signals so timing and context feel relevant, not random. The practical upshot is fewer checkout decisions and more timely replenishment of essentials.

Go also leans into aesthetics to enhance engagement, showing contextual scenes—like poolside fruit on a hot day—to make suggestions feel immersive. These presentation choices are designed to help users visualize how items fit into real moments, which can speed the decision to buy. That visual layer complements practical prompts rather than replacing them. It’s an attempt to make shopping feel less transactional and more situational.

The move follows a larger industry trend: shops and platforms are rolling out agentic commerce features that actively manage carts for customers. These agents promise to automate routine tasks—suggesting, bundling, and even adding items—so the user’s cognitive load drops. For companies, the upside is higher basket conversion and more predictable demand signals. For customers, the promise is convenience, though it raises clear questions about control and oversight.

Gopuff framed the partnership as solving the friction before the transaction, the part where people think, decide, and remember. In a statement to Business Wire, Gopuff’s founder said the following

 “Thirteen years ago, we bet that instant delivery would change the world and built the infrastructure to enable that vision. Today, we believe the greatest friction left in commerce is not delivery or instantaneous access to the essentials customers need. It’s the moment before: the thinking, the deciding, the remembering.” 

That emphasis on pre-purchase friction explains why Go focuses so heavily on memory and timing rather than raw search accuracy. By anticipating needs, the agent aims to turn planning into a simple confirmation step. Users still control the final purchase, but the suggestion-to-checkout pathway gets shorter. It’s a product bet that consumers will trade a bit of manual searching for curated convenience.

Privacy and user control will determine how smoothly features like Go scale. Any system that tracks purchases and infers habits needs clear settings and transparency about data use. Gopuff will have to balance convenience with guardrails so recommendations don’t feel intrusive. If they get that balance right, Go could be a handy assistant; if they don’t, it risks becoming noisy or presumptive.