Meet the scaleup tackling AI's forgotten challenge: your camera roll

Everyone building AI right now wants to generate something new — an image, a video, a paragraph. Popsa, a London scaleup, is going the other direction. It wants to make sense of the pile of photos you already have. The numbers back up why that's a real problem: people now take an average of 551 photos a month, more than 6,600 a year, and according to Popsa's own research, 70 per cent of them never get looked at again. They just sit there, buried in a camera roll, quietly multiplying.

Co-founder and CEO Liam Houghton knows the feeling better than most. He's built a personal library of more than 200,000 photos, partly by digitising his entire childhood — old computer files, printed photographs, even backfilling locations and timestamps by hand. He calls it a complete photographic record of his life from birth to today, and he expects younger generations to end up with that record automatically, without ever meaning to create it. That, in his view, is what makes photo organisation such a large opportunity over the next few years.

Popsa's platform analyses the people, places, activities and relationships in a user's photos, then organises them, strips out duplicates and produces a curated version of someone's life story, largely for printed photo books rather than screens. The app currently generates around 89,000 smart albums. What's notable is where the analysis happens: on the device itself, not in the cloud, so Popsa says it typically never sees a user's photos unless that person chooses to print them. Hundreds of separate AI models run behind that single interface — one scoring composition and lighting, others spotting recurring people versus background strangers, others reading activities. The company has also pushed into storytelling, turning images into descriptive tokens that capture relationships, locations and emotions rather than just objects, so the system can tell the difference between