Google aims to add generative AI to retail. To attempt.
To coincide with the National Retail Federation’s annual convention in NYC, Google Cloud launched new GenAI tools to help businesses customize online purchasing and improve back-office processes.
Eltrys wasn’t able to evaluate the new tools before this morning’s announcement, so this writer can’t tell how they function. The launch is scheduled for Q1. However, Google’s disclosures demonstrate its aggressive GenAI customer acquisition strategy.
The new Google Cloud Conversational Commerce Solution enables merchants to embed GenAI-powered agents on their websites and mobile applications, like a brand-specific ChatGPT. Natural language chats with shoppers provide personalized product recommendations from agents.
Branded chatbots are old. Google claims the agents are powered by “sophisticated” models like PaLM and can be tweaked with merchants’ data (catalogs and websites).
Google Cloud’s new Catalog and Content Enrichment toolset uses GenAI models like PaLM and Imagen to generate product descriptions, metadata, categorization suggestions, and more from a single product photo, complementing the Conversational Commerce Solution. The toolkit also allows merchants to create fresh product pictures from current ones or utilize product descriptions to produce AI-generated shots.
When eBay released a comparable AI-powered product image-to-description feature a few months ago, merchants quickly complained about misleading, repetitious, and even untruthful content.
I asked Google Cloud’s managing director of retail, Amy Eschliman, what Google had done to address reports of hallucinations. Google is “continuously improving” its systems, and human review is essential to the catalog and content enrichment procedures, she said.
When the stakes are great, I hope for human assessment. For example, a deceptive AI-generated picture or description in a product catalog might get a store in trouble with consumers or false advertising accusations.
Eschliman said human-in-the-loop is a recommended practice for corporate use cases to assure high quality, eliminate bias-related risk, promote trust and transparency, develop and train the model, and comply with regulatory and company policy.
Google unveiled a retail-specific Distributed Cloud Edge device today to “reduce IT costs and resource investments” around retail GenAI. (Google has always supplied Distributed Cloud Edge, but it’s now targeting shops.) Google said the edge cluster, which powers GenAI applications in convenience marts, gas stations, fast casual restaurants, and grocery shops, comes in single-server and multi-server configurations.
According to Eschliman, Google Distributed Cloud Edge enables shops to maintain operations even when their location is out for a few days, thanks to the local control plane. Retailers may now deploy a small cluster of Google Cloud-managed nodes in almost any shop. This completely managed hardware and software lets retailers operate current software with distributed AI for 24/7 mission-critical store operations.
Google claims Q1 will bring price and availability.
After being informed, I wondered: Are stores truly demanding GenAI?
Perhaps. At least retail titans.
Walmart said yesterday that it is actively investing in GenAI search to better comprehend queries and allow customers to search by use case (e.g., “unicorn-themed toddler birthday party”). Amazon uses GenAI to summarize user reviews, assist sellers in creating product descriptions and picture captions, and help shoppers select garments that fit.
According to a Google poll, 81% of retail decision makers feel “urgency” to adopt GenAI, while 72% are ready to deploy GenAI technology today in customer service automation, marketing support and product description generation, creative assistance, conversational commerce, and store associate knowledge and support.
But given some of the bumpy GenAI rollouts in retail lately (see Amazon’s review summary inflating negative comments), I’m not persuaded the retail sector would hurry to embrace GenAI from Google Cloud or any other supplier. We must wait and see.
