The Conversion Shift: Why Customers Trust Autonomous AI over Legacy Filters

 Ask any beauty consumer and they'll tell you that buying foundation online has always been a gamble. It isn’t like picking out a new lipstick or eyeshadow palette, finding the perfect foundation match is deeply personal, and difficult even on the hight street with expert lighting and a fair amount of trial and error.

According to research from YouGov, 86% of UK beauty consumers regularly purchase the same products, with only 24% actively trying new brands and products. This is a missed opportunity for consumer acquisition, cross selling and increased AOV.  

In August 2023, we designed and deployed a patent-pending, AI-powered Foundation Finder for LOOKFANTASTIC, combining high-precision physics with advanced machine learning. This didn’t just create another widget on the LOOKFANTASTIC site, it was a highly scalable agentic commerce solution that now drives a 5% revenue uplift across client sites, while fundamentally altering the mechanics of ecommerce fulfilment and inventory demand.  

And this was just the beginning. 

Moving beyond the guesswork

Many conversational product finders in the beauty ecommerce space rely on basic, static questionnaires that filter products based on a handful of subjective user inputs. These systems fail because they do not account for the complex science of human skin tone, product texture, and lighting.

We wanted to bring the in-store consultation experience online with complete accuracy. This meant approaching the project from a highly scientific, data-led perspective:

• Spectrophotometer Measurement: Engineers analysed over 3,400 foundation samples under laboratory conditions. A specialised spectrophotometer captured the exact colour profile and finish of each product under multiple simulated lighting environments.
• Advanced Colour Algorithms: The tool cross-references user-provided data (including current products used and preferences on coverage, finish, and skin type) with its database of over 6,100 shade options across nearly 300 products and 86 brands.
• Product Attribution Modelling & Semantic Search: Rather than relying on simple keyword matching, THG Commerce used data-led generative AI to build deep product profiles, enabling the finder to understand semantic queries like "dewy, medium-coverage for dry skin in natural light."

The results were validated through 400 Glossybox subscribers who tested the tool; 80% confirmed that the shade recommendation was exactor extremely close to their actual skin tone. 

The hard metrics: conversational AI driving real financial performance

Engagement metrics are valuable, but the ultimate validation of any project like this lies on the bottom-line impact. The Foundation Finder has delivered exceptional commercial performance:

• Despite the typical consumer reluctance to switch cosmetics brands online, 73% of customers purchased a new foundation brand through the finder.
• 96% of customers purchased a product they had never bought before.
• Nearly a quarter (22%) of users made their first-ever foundation purchase on the site through the tool, proving its efficacy as a customer-acquisition engine. 

The rise of agentic commerce

Agentic commerce is moving quickly, and in 2026 it is bigger and more intelligent than ever before. What started as a highly sophisticated diagnostic quiz for LOOKFANSTASTIC has paved the way for a whole new generation of agentic commerce experiences.

Today we are moving far beyond questionnaires. We have recently launched full-scale AI Shopping Assistants and interactive Virtual Try-On (VTO) solutions. These cognitive engines do not just ask you about your skin type; they visually overlay products onto your face or body in real time, letting you see exactly how a shade of lipstick, foundation, or eyeshadow reacts to your specific features. It bridges that final gap between digital browsing and physical trial, giving consumers the absolute confidence to tap "buy".  

This isn’t the end, we are constantly developing these tools, talking about what is next for Agentic Commerce at THG Commerce. The space is developing constantly, think: interactive, real-time, live. Dynamic, hyper-personalised personal shopping taking conversational commerce to the next level. We are moving from simple transactional interaction to an era of immersive, collaborative digital experiences where the storefront feels alive and personalised - no different to a flagship high street store. 

The operational reality of agentic commerce

While the financial benefits of conversational AI are clear, implementing these systems has its own challenges. These difficulties should be discussed openly:

• Data onboarding: Mapping thousands of SKUs via spectrophotometer is an incredibly resource-intensive process. While many conversational commerce products are ‘plug and play’ automatically capturing all your website data, for those that aren’t, maintaining a standardised data library as products are constantly reformulated or repackaged requires continuous operational discipline.
• Real-time scaling: Conversational AI must process image uploads and rich dataset queries within milliseconds. Any lag in response time directly correlates to cart abandonment. This requires substantial investments in robust API architectures and high-performance cloud environments.
• The operational strain of peak periods: The warehouse needs to be able to keep pace of the demand of peak periods. An exceptional front-end user experience means nothing with backlogs, delayed deliveries, and negative customer reviews that quickly erode the brand trust built by the AI finder.

The era of Agentic Commerce

The success of the Foundation Finder proved that the future of ecommerce was personalisation at scale, and it was just the starting line of a continuous innovation race. The agentic commerce landscape is developing daily, moving at a relentless pace with brands constantly competing to pioneer the next breakthrough. Resting on yesterday’s success is not an option.

We are actively pushing the boundaries of what is possible, from scaling cognitive AI to developing increasingly intelligent solutions to bridge the gap between human intuition and machine precision. The breakthrough of the Foundation Finder was only the beginning; the most intelligent chapters of this journey are right around the corner…

Subscribe to our newsletter to stay up to date with all of our latest agentic commerce product launches.