‘The way we interact with search boxes has changed’: What GenAI and RAG mean for retailers

If you’ve used Amazon’s Rufus or Walmart’s search bar, you’ll be familiar with new forms of AI-powered site search.

As more retailers are adding AI-powered search, we caught up with the co-founder of one such solution to find out how this tech is changing the relationship between shopper and online store, and what it means for the future of retail.

Borja Santaolalla is Co-founder and Chief Experience Office at Empathy.co, an enterprise AI search solution made for commerce, which works with the likes of Kroger, Inditex and Carrefour. He told us about the rise in mission-driven queries, how brands can be found in answer engines, and what it means to build responsible AI.

Econsultancy: Is the shift to new search functionality happening in earnest? What are the motivations for retailers?

Borja Santaolalla: Generative AI, together with RAG (Retrieval Augmented Generation), has been a disruptive force in the search industry. The way we interact with search boxes has changed. Our old habit of synthesising our search intent in a handful of keywords has rapidly moved towards more descriptive, natural language phrasing.

When faced with a search box, people are starting to behave as if they were speaking to a shop assistant, rather than a dumb catalogue retrieval search box.

The way we interact with search boxes has changed.

This new customer behaviour signifies a huge opportunity for retailers. Retailers can now design conversational search experiences that are not purely driven by product findability but, more importantly, by discoverability, relationship building, and trust.

In your work with retailers such as Kroger – what accounts for the increase in findability/add to basket i.e. what problems are being solved?

Grocers are experiencing this new change in search behaviour firsthand. Customers want to build healthier shopping carts while saving time and money. Mission-driven queries are starting to become increasingly popular: “healthy snacks with no gluten,” “protein shakes after workout,” or even “frozen food for a Friday TV night.”

A blended approach to search is necessary, one that combines traditional keyword-based (i.e. heuristic search) with semantics (i.e. vector search). This approach, commonly referred to as Hybrid Retrieval in the industry, is retailers’ solution to adapt to changing customer search patterns whilst maintaining the precision required for product, category, and brand-based traditional queries.

You recently wrote about product findability in answer engines like ChatGPT – how big a consideration is this for retailers, compared with AI-powered search on site?

Positioning products in answer engines is a huge growth opportunity for retailers. The ability for a retailer to become contextually relevant as part of a conversation in the public space via general purpose LLMs (i.e. ChatGPT, Perplexity, Gemini), or within their own channels via chatbots, conversational search UIs, in-store kiosks, etc. is the new billion-dollar opportunity. This is particularly effective when combined with personalisation and monetization / retail media. The approach is similar in both cases: build branded LLMs and train them for specific use cases.

Mission-driven queries are starting to become increasingly popular…

Brands need to build their own language models that reinforce their value proposition, so their content is displayed contextually within generative responses and influences related prompts for subsequent interactions.

Your CEO has spoken about explicit vs. implicit discovery of products. How do platforms play into this ecomm evolution e.g. TikTok, CRM, marketplaces vs. .com?

We all love discovering new products, as long as we understand how they have appeared on our radar. Serendipity is rare in the age of surveillance. Relevancy rapidly turns creepy when we feel we’ve been targeted without our consent.

As conversational experiences become more present, privacy is more critical than ever. Ensuring that customers are in control of deciding what personal data can be tracked and for what purpose is paramount in terms of eliciting trustworthiness. Consent management platforms like Visible Privacy make it easy for customers to configure their preferences and for brands to demonstrate their commitment to data protection. Relevancy and respect have to go hand-in-hand.

What is a responsible AI product in ecommerce? (To what degree are retailers using PII/cookies for merchandising and personalisation – and where do you think this is problematic or unnecessary?)

As we move away from cookies, tracking and surveillance, we see retailers adopting privacy-by-design strategies. First- and zero-party data profiling and preferences (i.e. nutritional, health, savings/personal finance, sustainability), which are often inferred from past purchases, transparently elevate the shopping experience within My Account. This is rightfully becoming the new norm in responsible personalisation and providing value to customers.

I suggest watching the live chat from the Ethical Commerce Alliance partner S Group: My Purchases: a use case on trust, privacy and data ethics.

Empathy.co has talked about moving away from ‘productised’ tech towards an innovation-led culture. What does that mean for the way your teams work together?

Retailers feel captive to big tech providers, not only as brokers in customer acquisition but also as ecommerce tech platforms and cloud providers.

Customers feel fooled when it comes to protecting their own data, while retailers don’t know how to break away from their technology straitjackets.

Retailers don’t need more products, they need tools and deep domain expertise to freely express their brand and values through co-creation. Open Innovation is our way of helping clients design and build their own privacy-first product discovery experiences, free from commercial platforms and clouds.

A clear example is how big retailers have decided to build their own CMPs (consent management platforms). Owning consent means brands can design granular, progressive consent paths that are understandable and relatable to their customers. No more long legal consent texts at the store frontdoor!

‘Expressive experiences’ is a term you’ve used – can you give us the vision for what ecommerce could look like, 5 years from now?

As transactional ecommerce becomes more and more of a commodity, brands will differentiate in the way they express their purpose and relate to their customers.

Some brands will express these relationships heavily indexed on technological advancements (i.e. immersive shopping experiences, AR, AI personalization), while others will use technology to elevate emotions, connect with the community and nurture human connections. Or both at the same time, needless to say! Brand values and identity will dictate how they choose to differentiate experiences.