Google sees retail success with agentic commerce push

Google is accelerating its push into enterprise customer experience with a unified AI platform designed to manage the full retail journey, from product discovery to after-sales support. The Gemini enterprise suite for customer experience, launched earlier this year, is already being deployed in live retail environments, pointing to a shift in how businesses approach conversational commerce.

Unified platform

"During the NRF in January of this year, Sundar announced the Gemini enterprise for customer experience, which what we did was we took search, which we knew was a really good tool, Vertex AI search. We took service, a service suite, which comes with an agent studio to build service agents for really large call centre applications that use AI. And then we took sales, our sales agent or shopping agent, and mashed all of this together into one single suite, so that regardless of whether you're on top of the funnel trying to search, discover, add to cart, buy the product, or after you bought a product, you have concerns that you want to solve, that entire journey we collapsed into one single suite, and we called it Gemini enterprise," said Darshan Kantak, VP of Product, Applied AI, Google.

The platform brings together search, service and sales into one system, alongside pre-built agents for retail, food ordering and automotive use. Businesses can also build custom agents within the same environment.

Kantak said the aim is to simplify deployment, allowing organisations to orchestrate pre-built tools or extend them with their own workflows. The approach reduces fragmentation across customer-facing AI systems.

Multimodal shift

"Another one is multimodality, which is very important. So whether you're using audio, we stream audio directly into the LLM, and audio out of the LLM. There's no speech to text translations, which adds lag, which adds transcribing errors, which adds robotic nuances or tone to the speech. We've removed all of that because it's audio to audio directly," said Kantak.

The system replaces text-heavy chatbot interfaces with structured, interactive elements.

"So that carousel is now embedded as a widget that comes automatically into the chat so you don't have to parse through a wall of text. It synthesises, it formats, it makes it actionable for you," added Kantak.

It also supports image and video inputs, allowing users to upload photos to identify products or diagnose issues, with responses tailored to the context.

Enterprise integration

"Customers want to know that the existing ecosystem they can connect to, they don't have to bring their data over. We give them over 100 enterprise grade connectors to all of the tools they're using currently," said Kantak.

More than 100 pre-built integrations, including platforms such as Shopify and Salesforce, enable organisations to connect existing systems without migrating data.

This allows customer data, policies and workflows to be accessed directly within AI-driven interactions.

"Security, governance, all of the hardening so that you don't have a daisy chain of vendors you have to rely on, all of that is available to us and is available as Gemini enterprise for customers," said Kantak.

Commerce protocol

A key element of the platform is support for the Universal Commerce Protocol (UCP), designed to standardise how AI systems interact with retail services.

"Once you build your agent with us, you're also provided UCP readiness as part of it. UCP is universal commerce protocol which allows you to talk this common commerce language because all of the differences in the API are abstracted away," said Kantak.

The protocol is intended to make businesses discoverable across AI interfaces, including search and conversational assistants, enabling transactions within those environments.

"Your agent now becomes discoverable in certain contexts. When somebody searches for your business from Gemini or in the search mode, they can transact using you as a merchant of record directly from these entry points," added Kantak.

Adoption remains early.

"When you're talking about UCP, it's small, but it's starting to grow, and the back end agent AI is starting to grow," said Kantak.

Retail deployment

Retailers such as Bunnings have already deployed the platform in production, with implementations completed in a matter of weeks.

"That's the speed with which we've moved, and it's in production," said Kantak.

The system combines traditional search with conversational AI, allowing users to describe needs in natural language rather than rely on keywords.

"When somebody goes to your site and searches for an item in that catalogue, it is able to understand the context and serve the answer very quickly. When you type a conversational thing or speak it, we synthesise what it is you're looking for, break it down into context, send that to the LLM, get the response back, and personalise it for you," said Kantak.

The shopping agent supports discovery, recommendation and checkout, as well as post-purchase queries.

"These AIs are actually talking the language of humans, and not making humans talk their language. You're talking a natural language, and it understands you," said Kantak.