Instacart has deployed an embedded checkout experience within ChatGPT through the emerging Agentic Commerce Protocol.
With this rollout, Instacart becomes the first partner to launch a ChatGPT app that enables a complete shopping cycle from query to payment without requiring users to leave the conversation interface.
Operationalising agentic commerce
The integration addresses a long-standing failure point in conversational commerce: the “handoff.” Historically, AI models could recommend products or generate meal plans, but the execution phase required deep-linking to a separate app or website a step that often led to cart abandonment.
Under this new deployment, users can plan meals conversationally and have the system automatically build a cart based on local retailer inventory. The real differentiator is the checkout process. By leveraging the Agentic Commerce Protocol, transactions are completed directly within the chat interface using a Stripe-powered credit card flow.
According to Nick Turley, VP and Head of ChatGPT, the goal is to link AI suggestions directly to real-world services.
With the Instacart app embedded in ChatGPT, users can move from meal planning to checkout within a single, seamless conversation,” Turley said. “It’s another step toward realizing our vision where AI not only provides helpful suggestions but also connects directly to real-world services, saving people time and effort in their daily lives.
This integration goes beyond standard API usage. Instacart participated early in the OpenAI Operator research preview, offering feedback to ensure the technology could handle real-world constraints while following established norms.
This “preview” participation indicates that Instacart’s complex data environment with tens of thousands of SKUs and constantly changing stock levels acted as a testing ground for OpenAI’s agentic capabilities. Instead of merely using the tool, Instacart helped shape how an AI agent interacts with external fulfillment logistics.
The Instacart deployment highlights the importance of structured, real-time data when integrating with large language models (LLMs). An AI agent is only as effective as the data it can access; in a commercial context, hallucinations such as recommending out-of-stock items pose both financial and reputational risks.
Anirban Kundu, CTO at Instacart, emphasizes that enabling shopping within an AI agent requires technology capable of interpreting highly local and constantly changing inventory. Instacart addresses the “hallucination” risk by grounding the AI’s responses in its extensive dataset, which includes over 1.8 billion product instances across 100,000 stores.
“Instacart and ChatGPT are redefining what’s possible in AI-powered shopping,” said Kundu. “Built on the Agentic Commerce Protocol, this experience delivers intelligent, real-time support for one of the most essential aspects of daily life: getting groceries to feed your family.
“Together, we’re creating a seamless and secure way for people to turn simple conversations into real-world actions helping customers move from inspiration to a full cart delivered from the store to their door with ease.
Dual adoption: Customer-facing and internal efficiency
While the embedded checkout grabs headlines, Instacart’s broader plan involves extensive internal deployment. The company utilises ChatGPT Enterprise to streamline internal workflows, aimed at accelerating the development of customer experiences.
Furthermore, they have deployed OpenAI’s Codex to power an internal coding agent.
This dual approach – using AI to sell (Agentic Commerce) and AI to build (Codex) – offers a model for operations. It moves beyond isolated pilots into a holistic stance where generative models drive both revenue and R&D efficiency.
The deployment points to a change in how brands view digital storefronts. Instacart’s approach appears to accept that consumer entry points are fragmenting.
Rather than forcing all traffic through a proprietary app, the company is positioning its infrastructure as the backend fulfilment layer for third-party AI platforms.
Instacart’s availability and implementation in ChatGPT
The experience is currently active for users on desktop and mobile web platforms, while native mobile availability for iOS and Android applications is rolling out shortly.
To access the feature, users must invoke the specific Instacart application within the ChatGPT interface (for example, by prompting “Instacart, can you help me shop for apple pie ingredients?”) and link their accounts.
This opt-in mechanism ensures that data sharing is consensual, a requisite governance step for enterprises deploying consumer-facing AI agents.
This integration serves as a case study of agentic AI for commerce. For retail and technology execs, the Instacart model demonstrates that the next phase of digital adoption involves preparing API structures and data pipelines to serve “non-human” customers (AI agents) as reliably as human ones.
The focus must remain on data accuracy and real-time availability. Without these foundations, agentic workflows will fail to deliver return on investment.
