# AI landscape and Product Intelligence: How Hello Retail builds its personalization foundation > Hello Retail's Product Intelligence is a proprietary AI model built over five years that combines text and image signals to underpin personalized search and recommendations. **Author:** Rasmus Leth Skjoldan **Published:** May 21, 2026 **Tags:** product-intelligence, ai, personalization, ecommerce --- AI has been part of Hello Retail's platform since its founding, and for the past five years a dedicated internal team has done nothing but build proprietary AI models. The centerpiece of that work is Product Intelligence - a holistic model that combines text and image signals to understand products through a product-first lens. It sits beneath nearly every feature Hello Retail ships, from ranked recommendations to personalized search results. ## Why shopper expectations keep climbing Two words dominate every ecommerce conversation right now: artificial intelligence. The pressure that creates for online retailers is real and measurable. According to Salesforce's State of the Connected Customer report, 88% of customers say the experience a company provides is as important as its products and services. As AI models get better at reading individual preferences, shoppers start to assume that every store, every search result, and every recommendation already knows who they are. The practical consequence is that personalization has moved from competitive differentiator to baseline expectation. McKinsey research found that 71% of consumers expect companies to deliver personalized interactions, and 76% say they get frustrated when that does not happen. For ecommerce operators, the question has become less about whether to personalize and more about how deeply the underlying models actually understand the catalog. ## Product Intelligence: A five-year build Hello Retail's answer to that challenge started well before generative AI entered mainstream conversations. Over the past five years, an internal team has focused exclusively on building proprietary AI models. The most important output of that sustained effort is Product Intelligence. Product Intelligence approaches the world through a product lens. Rather than applying a single model to all signals, it draws on multiple underlying models working in concert: text models that process product descriptions, copy, and attributes alongside image models that interpret visual signals from product photography. The result is a richer, more holistic representation of each product - one the platform can reason about in context, understanding purchase patterns, natural pairings, and catalog positioning in ways that neither text nor image alone could support. That layered architecture matters because real product catalogs are messy. A title alone does not capture whether a running shoe is built for trail or road. An image alone does not convey that a jacket is fully waterproof. By combining signal sources, Product Intelligence builds a picture of each product that reflects how shoppers actually experience it, going beyond the static description captured at import time. ## What Product Intelligence actually powers Product Intelligence operates as the shared foundation across nearly everything Hello Retail ships. Product Recommendations uses Product Intelligence to rank products per shopper based on behavioral signals captured across the storefront - surfacing items that fit each visitor's specific pattern of browsing and buying, shaped by the platform's understanding of both the shopper and the product. Search draws on the same foundation to return ranked results shaped by both the query and the shopper's behavior, so two people searching for "boots" can receive meaningfully different result sets depending on what the platform knows about each of them. That shared foundation is a deliberate architectural choice. When recommendations and search draw from the same product intelligence layer, the personalization stays coherent across touchpoints. The same understanding of a product informs whether it surfaces in a search result, a recommendation widget, or a triggered email. There is no seam where one part of the platform has a different model of what a product is from another. For teams interested in how the models work, Hello Retail publishes an overview at [helloretail.com/en/product-intelligence/](https://helloretail.com/en/product-intelligence/). ## AI belongs in the foundation The business case for getting this right is substantial. Epsilon research found that 80% of consumers are more likely to make a purchase when brands offer personalized experiences - a figure that puts real revenue weight behind the quality of the underlying models. AI models that drive that personalization need to be built, tested, and refined against genuine catalog complexity before they deliver reliably at scale. Hello Retail's position is that AI belongs in the foundation, embedded in the core models from the start instead of bolted on as a late-stage feature layer. Five years of dedicated work on Product Intelligence reflects that view. The models powering personalized experiences in this Q4 release were built ahead of the current AI moment, which means they have had time to be validated and hardened across thousands of product catalogs and real shopper behavior patterns. For ecommerce teams evaluating personalization platforms, that history is worth pressing on. A platform where AI has been load-bearing infrastructure for five years has encountered failure modes - and corrected them - that a more recently assembled system has yet to face. That difference surfaces in recommendation quality, search relevance, and the consistency of experience across every channel where a shopper meets the catalog. The full Q4 fall release session shows how Product Intelligence connects to Retail Media, Search Word Boosting, and Recommendations in practice - the chapters that follow this one each demonstrate a specific capability built on this foundation. --- *This content is from the Hello Retail blog. For the full experience with images and formatting, visit [helloretail.com/en/blog/ai-landscape-product-intelligence](https://helloretail.com/en/blog/ai-landscape-product-intelligence)*