AWS for Industries

Harnessing Generative AI on AWS to Transform Retail Insights

Tapestry, a global luxury fashion holding company, oversees iconic brands such as Coach, Kate Spade New York, and Stuart Weitzman. With over 1,400 retail stores worldwide and more than 18,000 employees, Tapestry sits on a wealth of information that it could use to improve customer experiences and optimize its operations. The company needed an effective system to harness this knowledge at scale, and generative artificial intelligence (AI) emerged as a promising solution.

“At Tapestry, we always look at technology as a driver for our business,” says Muhammad Chaudhry, global head of data engineering at Tapestry. “We are a data-driven company, and generative AI is a new technology. We analyzed it and asked questions like, ‘Is it a driver for us? Does it help improve the lives of our associates and help the business grow?’ Generative AI could clearly help us solve some key business challenges.”

Uncovering the Need for AI in Retail

The first step to implementing effective technical solutions is to clearly define the problem that needs to be solved, then work backward to identify the most suitable approach to solving it. In Tapestry’s case, the existing feedback collection methods were fragmented and could not be scaled across its large retail network. Store visits by corporate teams only yielded anecdotal information that couldn’t be analyzed systematically or used to make effective changes. This led to an incomplete view of customer trends and associate needs, potentially impacting everything from inventory management to store ambience.

“What became very evident was that generative AI could gather and synthesize associate feedback using natural language processing and create a summary,” says Deepak Chandak, senior director of omni-innovations and product management at Tapestry. “Gathering feedback manually from all of our associates was humanly impossible, not to mention summarizing and analyzing that feedback. But generative AI could make this possible.”

The in-house engineering team at Tapestry saw an opportunity to use Amazon Web Services (AWS) to create a generative AI engine that would not only solve these immediate challenges but also serve as a foundation for future AI-driven innovations. With services like Amazon Bedrock—a fully managed service that offers a choice of high-performing foundation models—the company could tap into powerful AI capabilities without the need for extensive infrastructure management or model training.

Constructing a Robust Generative AI Engine

Tapestry’s generative AI engine is built on a foundation of nearly 20 AWS services, with Amazon Bedrock at its core. Amazon Bedrock hosts the large language model (LLM)—initially Claude from Anthropic—that powers these AI capabilities. Amazon Simple Storage Service (Amazon S3), which is built to retrieve any amount of data from anywhere, serves as the central repository for the massive amount of data that Tapestry collects. The company used the generative AI engine to help build two applications that collect and analyze feedback from store associates: Tell Rexy and Ask Rexy.

Tell Rexy, the feedback collection app, is deployed on store devices like tablets and point-of-sale systems. It uses Amazon Transcribe, a fully managed automatic speech recognition service, to convert spoken feedback from associates into text. This facilitates seamless input without the need for typing. Amazon Translate, a neural machine translation service, is integrated into Tell Rexy to support multilingual feedback; this service automatically converts non-English inputs into English for centralized processing.

The collected feedback is then processed and stored on Amazon S3, and Amazon Athena—a serverless, interactive analytics service—is used to create queryable tables so that corporate users can quickly access and analyze the data. Tell Rexy also employs Amazon Comprehend, which helps derive and understand valuable insights from documents, for sentiment analysis. This helps Tapestry gauge the overall mood and satisfaction levels of their associates. The BERTopic neural topic modeling technique is used to group similar feedback. A daily step function, activated by Amazon S3 bucket notifications, runs sentiment scoring and topic clustering on new feedback from associates. The Amazon Athena tables are then updated accordingly.

Ask Rexy, the analytics chatbot, uses a combination of retrieval-augmented generation and text-to-SQL capabilities to answer questions from corporate analysts about the collected feedback. Amazon Kendra, an intelligent enterprise search service, pulls relevant passages from documents stored in Amazon S3 based on the semantic similarity between the user’s question and the stored content. When specific keywords like sentiment or topics are included, the LLM will convert the question into a syntactically correct Amazon Athena SQL query and pull the relevant feedback.

Scaling AI Applications Across Tapestry’s Brands and Teams

Tell Rexy is live across most of Tapestry’s Coach stores in the United States. The application has been used by several thousand associates, who have provided nearly 30,000 pieces of feedback in 1 year. This substantial volume of data is helping Tapestry gain unprecedented insights into store operations, inventory management, and customer preferences. The company is now in the process of expanding these applications to its Kate Spade brand.

The development of this generative AI engine has significantly accelerated Tapestry’s ability to create new AI-powered applications. The company reports that it can spin up new applications 10 times faster thanks to the reusable components and extensible architecture of the engine. This increased efficiency is opening doors for Tapestry to explore additional use cases across its brands and corporate functions. The company is already seeing interest from other business units, such as corporate communications and investor relations, in using the generative AI engine for their specific needs.

“When we build systems in house, we are guided by three principles: scalability, elasticity, and extensibility,” says Chaudhry. “We adhered to these when we built Tell Rexy and Ask Rexy. We have created a generative AI engine where if we have to spin up or provision generative AI applications, it has become much easier for us.”

Learn more about how Tapestry and AWS work together to bring innovation and personalization to market.

Tapestry Collects Feedback from Thousands of Store Associates Using AWS
Tapestry Builds a Scalable IaC Platform for Modernized Workloads Infrastructure Provisioning with Built-In Governance and Security
Tapestry Gains 360-Degree View of Customers by Powering Data and Analytics on AWS

Aditya Pendyala

Aditya Pendyala

Aditya is a Principal Solutions Architect at AWS based out of NYC. He has extensive experience in architecting cloud-based applications. He is currently working with large enterprises to help them craft highly scalable, flexible, and resilient cloud architectures, and guides them on all things cloud. He has a Master of Science degree in Computer Science from Shippensburg University and believes in the quote “When you cease to learn, you cease to grow.”

Deepak Chandak

Deepak Chandak

Deepak Chandak serves as Senior Director for Omni Initiatives and Product Management at Tapestry, Inc., where he partners with esteemed brands Coach, Kate Spade, and Stuart Weitzman. He has extensive experience in product management within the consumer retail sector and is known for his ability to drive business growth by identifying customer challenges and creating innovative solutions. Deepak is a dedicated learner who believes that an organization’s strength lies in its people. He builds cross-functional teams and fosters trusted partnerships, both internally and externally, to deliver sustainable business value. He is committed to enhancing employee engagement and productivity and continues to make contributions to Tapestry's success and the overall retail landscape.

Fabio Luzzi

Fabio Luzzi

Fabio Luzzi is a Technology Executive at Tapestry based out of NYC. He has extensive experience in building and leading teams to execute end-to-end Digital and Data transformation strategies by leveraging Data, ML & AI. Fabio has a proven track record of driving profitable growth for enterprises across different industries such as Technology, Payment Services, Entertainment, Advertising and Retail. He has a Master of Science degree in Statistics and Economics from La Sapienza University in Rome, Italy and has global experience working in Italy, the UK and the United States. Fabio believes in the quote “labor omnia vincit.”

Frank Rosalia

Frank Rosalia

Frank Rosalia is an Applied AI Engineering Manager at Tapestry based in the NYC metro area. In his time at Tapestry, he has worked on a variety of greenfield projects using AWS. Frank was at the vanguard for Tapestry's foray into enterprise AI applications and continues to integrate cutting edge solutions as they become available. Most recently he worked on developing an integrated RAG chatbot platform within the Tapestry ecosystem which allows users to use their own personal knowledge base in less than 20 minutes. He holds an MSc in Data Science from the London School of Economics and a BA in Mathematics and Statistics from Columbia University.

Muhammad Chaudhry

Muhammad Chaudhry

Muhammad Chaudhry is a technologist with 15+ years of experience designing and building cutting-edge technology solutions and data platforms . He has extensive background in building cloud-native, hybrid and on-premises data solutions serving the strategic needs of the business. Over the years his career has evolved from a hands-on data engineer to an IT leader focused of solution architecture, value driven systems delivery and creating strategic alignment of IT and business teams. He holds an MBA with a concentration in global business management from University of Pittsburgh (Pittsburg, PA) and currently heads the Data Engineering group of Tapestry (NYC, United States).