CategoryIQ
IceCream Labs built AI-first products for retail. CategoryIQ applied machine learning to structure and categorize product catalogs at a scale manual processes couldn’t touch - categorising SKUs, tagging images, attribute enrichment.
Brick and mortar retailers who were setting up their e-commerce divisions, were realising that their catalogs were inadequate when it came to, not just number of SKUs, but also the quality of information (images, categorisation, attributes) about their SKUs. This was costing them across the conversion funnel from acquisition to conversion to retention.
Applied ML only earns its keep when it slots into a real workflow.
Focused the model on the decisions retailers actually needed to make, not on accuracy for its own sake.
Designed the product so ML output was reviewable and trustable — a human could see why a categorization was made.
Packaged it as something a retail team could adopt without a data-science department.
Award-winning AI tech required to be paired with frontier engineering to make it adoptable and a SaaS business. The win came from letting customers focus on their workflows without having to worry about how AI/ ML models worked. The product was what made it usable. An early lesson in what would later become the AI-product thesis behind Action Feed.