ASOS Enhances Customer Experience with Azure AI
February 04, 2025
UK-based ASOS, a leading fashion destination for young adults, consistently seeks to leverage innovative technologies to elevate customer experiences. With a vast catalog and a powerful recommendation engine built on Microsoft Azure, the company delivers a personalized fashion experience to its customers. Recognizing the potential of generative AI platforms, such as ChatGPT, ASOS identified an opportunity to expand its business, enrich its technology infrastructure, and meet the evolving tech expectations of its target demographic.
To achieve these objectives, ASOS implemented Azure OpenAI Service and Azure AI prompt flow with ChatGPT language models. The goal was to rapidly develop and test an AI-powered customer experience. This new experience helps customers explore and discover new looks with a conversational interface.
“Our customers come to us for great fashion. Having a conversational interface option could get us closer to our goals of fully engaging our customers and personalizing their experience by showing them the most relevant products at the most relevant time.” – Cliff Cohen, Chief Technology Officer, ASOS
ASOS recognizes the importance of creating engaging and inspirational experiences. Its offerings aim to provide customers with trends and offerings that suit their individual preferences. With hundreds of brand partners and its own labels, ASOS seeks to be a part of every aspect of the customer’s life.
ASOS actively invests in digital solutions that leverage emerging technologies like generative AI to enhance the customer experience and deliver uniquely tailored offers. The company strategically partnered with Microsoft to build a new AI-powered feature that could interact with customers as soon as they arrived on the ASOS website or app.
Building a Natural Language Experience
“We’ve worked closely with Microsoft in the past and we’ve been early adopters of technology all along, so it made sense to explore using Microsoft technology coupled with our team’s existing knowledge of natural language processing to develop a new kind of experience to engage shoppers,” says Cohen.
The primary objective was to develop an AI-powered experience that could curate selections tailored to each customer’s preferences. This strategy proved essential for rapidly acquiring knowledge about the latest industry trends. The ASOS team used prompt flow within Azure AI Foundry to explore language-model-powered generative AI options. This accelerated the secure development of the solution.
“Prompt flow helped streamline our development and testing cycles, which established the groundedness we required for making sure the customer and the solution were interacting in a realistic way,” says Fabon Dzogang, Senior Machine Learning Scientist at ASOS. “In addition, we used Azure OpenAI Service models and prompt flow to communicate and orchestrate workflows between various services using our internal authentication APIs.”
With the capabilities of these Azure services, ASOS created a prototype within weeks. The AI experience engages with customers to understand their fashion preferences, integrates trend data drawn from external sources and ASOS designers, and allows shoppers to effortlessly find relevant purchase options from the company’s extensive catalog.
“Our customers come to us for great fashion,” says Cohen. “Having a conversational interface option gets us closer to our goals of fully engaging the customer and personalizing their experience by showing them the most relevant products at the most relevant time.”
Training AI to Track Trends
To ensure the solution’s effectiveness, the AI must remain aware of current trends. “A big part of fashion is newness and keeping up with trends,” says Cohen. “So it was critical that we develop capabilities for the experience to meaningfully demonstrate newness in a great way.”
ASOS leveraged AI to gather and embed the latest trends into its knowledge base. The experience then curates and shares items based on insights of the latest fashion trends.
“For this project, a key part is the tone of voice,” says Vanessa Spence, Senior Creative Director at ASOS. “We’ve worked hard to make sure that when the customer is engaging with the experience, they feel like they are talking to ASOS.”
Fast Development and Responsible AI
ASOS has kept its engineering investment minimal, allowing developers, including those new to AI, to onboard quickly.
At every stage of development, ASOS has prioritized responsible AI to reduce cultural biases in search results. They successfully eliminated malicious AI “jailbreaks” during initial testing. Using responsible AI principles from Azure AI Content Safety helps ensure top-quality interactions and output within the natural language experience, consistent with the company’s high ethical standards.
“Responsible AI was actually the starting point for our decision to use prompt flow capabilities,” says Papinder Dosanjh, Head of Data Science and Machine Learning at ASOS.
Dzogang adds, “We tested with 150 internal users, plus a sampling of real customers, and got zero red flags in terms of safety and security for our customers or for our brand. This gave us the confidence that we could proceed to scale up.”
The next step for ASOS is customer testing with a limited group of users. Once launched, the AI-powered natural language experience will integrate with ASOS’s existing suite of AI-enabled tools. These tools support the customer experience and assist in making effective business decisions.
“Without Azure AI prompt flow, we would have been forced to invest in quite significant custom engineering to deliver a solution. Instead, we were able to achieve great speed by easily integrating our existing microservices into the prompt flow solution, and easily publishing to our managed endpoints.” – Papinder Dosanjh, Head of Data Science & Machine Learning, ASOS
