Google on Wednesday announced the public beta of Recommendations AI, a fully-managed service that enables retailers to use AI to give customers personalized product recommendations. Personalization is one way the retail industry has adapted to the data age and evolving customer expectations.
Recommendations AI is based on technology that Google says it uses to deliver recommendations on its own products, including YouTube and Google Search. The cloud company says retailers can use the tool to replace or complement existing recommendation models.
To get started, an organization imports catalog and user events data. Then, they can choose a model type, identify their optimization objective and begin training the model. The initial training takes two to five days, Google says, before it can start giving customers recommendations.
The service taps into a consumer's shopping history, putting more emphasis on individual customer data than on product data. It uses "context hungry" deep learning models that tap into item and user metadata, according to a Google blog, "to draw insights across millions of items at scale and constantly iterate on those insights." The service also corrects for bias with popular or on-sale items, seasonality or items for which there is sparse data. Retailers can retrain models daily.
Recommendations AI directly competes with Amazon Personalize, the service that AWS brought into GA last year. Earlier this year, Adobe also released an AI-powered product recommendations tool.
Since Google Cloud CEO Thomas Kurian made his debut as cloud chief last year, Google Cloud has taken an industry-focused sales approach, zeroing in on six key verticals, including retail. The industry is a natural fit for the cloud company, given its wariness of Amazon.