Main content

Real-time data and retail - a match generated in AI heaven

Peter Pugh-Jones Profile picture for user Peter Pugh-Jones June 26, 2024
Retailers may be ready to invest in generative AI to improve the customer experience - but one of the biggest challenges is quality data. Confluent's Peter Pugh-Jones shares some differentiating examples.

Robot Holding Red Shopping Cart © AndreyPopov -
(© AndreyPopov -

Generative AI is reshaping the retail industry – fast. Retailers are realizing returns on their AI investments within 14 months, making back an average of $3.45 for every $1 spent. Customers, too, are reaping the benefits of AI-enhanced customer experience. Recent consumer research shows that they’re not only open to using generative AI tools to improve their online shopping experience – they’re excited about them.

But while the promise of AI is filling retailers and their customers with enthusiasm, there’s one critical barrier that stands in the way — data.

According to a 2023 global trends report into AI, one of the biggest challenges to innovation is access to clean and trustworthy data. Without it, a well-intentioned AI enhancement can end up leaving customers disillusioned and damaging a brand's reputation.

The critical differentiator for e-commerce players in the next decade won’t be whether or not they’re investing in AI. It will be whether or not they’re investing in real-time data to power it.

How can I help, really?

Advancements in generative AI are closing the window on frustrating, dead-end chatbot conversations. Instead of generic answers governed by a predefined script, AI-powered chatbots can provide instant support, answering questions and guiding customers along their shopping journey – all in dulcet conversational tones.

How? Uninterrupted, real-time data streaming feeds the AI model with ongoing conversational data, continuously improving every interaction going forward with more timely and personalized responses. This kind of in-flight data insight is enabling retailers to revolutionize the online shopping experience, bringing all the operational efficiencies of AI to customer interactions without the limitations of impersonal, rules-based answers.

For example, Microsoft Cloud for Retail recently announced a back-end template that retailers can build on to provide customers with a virtual personal shopper, enabling them to search and view products through natural language. A customer could type “I’m going camping in Yosemite this March, and I’ve never camped before. Help me find the right gear.” They’ll then get a friendly answer with tailored recommendations, as well as extra items to encourage them to spend more.

Walmart, too, has been experimenting with AI to create more supportive customer journeys by making it faster to find what they want through AI-powered search. If you’re planning a safari-themed birthday party for your kid you could search for the theme and find all the necessary products in one go rather than hunting around for the animal-adorned balloons, cake, napkins, and gift bags.

Then there’s lingerie and beauty retailer Victoria’s Secret. They’ve teamed up with Google Cloud to create a new conversational chatbot that provides shoppers with a more “inclusive” experience through tailored product recommendations and advice based on life experiences,  such as having a mastectomy or breastfeeding.

The last example of Victoria’s Secret shows how AI chatbots can mimic human store assistants in ways that extend beyond product search function into conveying – and connecting over – brand culture and values.

Avoiding the pitfalls of bad data

Imagine an online fashion retailer implementing a generative AI system to provide personalized clothing recommendations to its customers. By analyzing user preferences, purchase history, and browsing behavior the AI would suggest products to interest individual shoppers, increasing basket size and strengthening customer loyalty.

But then customers report receiving irrelevant or inappropriate recommendations. The system begins to degrade. Customer engagement and satisfaction go down. The underlying issue? Bad data.

Incorrect or incomplete product information, biased training data that doesn’t accurately represent the needs of your customer base, outdated information about purchase histories and browsing patterns. These all contributed to an AI shopping assistant that – in human terms – felt rude, dismissive, and ignorant.

Retailers wouldn’t want physical store employees like this. Virtual tools are surely the same.

Next-gen AI demands a next-gen data strategy

The retail industry generates 40 petabytes of data every hour, equivalent to eight million two-hour-long movies. But retail data is typically siloed, meaning companies gain insights from only a small fraction of it – resulting in an incomplete picture of their customers.

To derive actionable insights to leverage AI’s potential, seamless integration along with standardized data are needed across different systems and applications. And data needs to be fresh to avoid having to integrate it from multiple sources in various formats.

Confluent’s Data Streaming for AI – and more recently our AI Model Inference initiative – was set up to solve exactly these challenges. With Confluent, retailers can easily connect, process, integrate, and scale the data needed to support their generative AI use cases.

Confluent can combine data from multiple sources to create a unified view for generative AI models no matter where the data lives. Plus, by providing gen AI models with real-time access and the ability to process data from various sources, it improves their performance. It’s also simple to use and manage, making it easier to build and deploy generative AI applications.

Right here, right now

Generative AI enables retailers to offer more personalized experiences and create innovative solutions that meet the evolving needs of customers in a competitive market. Looking ahead, applications could extend to visual merchandising in-store, emerging trend analysis to enhance product lines, and virtual try-on features in augmented reality.

But whether we’re talking about today or tomorrow, retailers need the right data, right now. That’s why Confluent’s data streaming platform makes the perfect partner.

A grey colored placeholder image