[ ◀ BACK TO BLOG ]
VISIONATOR · SEP 16, 2026

A boring problem in an AI costume

What a returns file taught a US ecommerce client

A US ecommerce client asked me to build a recommendation engine.

"Like Amazon."

Sure. Everyone wants the Amazon one.

Before quoting, I asked for one file. Their returns.

Roughly one in ten of everything they sold came back. And a third of those returns said the same thing in different words: it wasn't the size I expected.

You don't fix that with a recommendation engine. You fix that with a size chart that tells the truth and a photo that shows the product on a real body.

We did that first. Cheap. Two weeks.

Returns dropped. Margin went up. The owner asked when we'd start the AI project.

I told him we just finished it.

Most companies don't have an AI problem. They have a boring problem wearing an AI costume.

My job is to undress it.

Also on LinkedIn. Join the conversation →
$ ls ./other_articles
▶Why we're letting people in slowlySEP 25, 2026▶The modem screamed, then nothingSEP 23, 2026▶A company with no screen in itSEP 21, 2026▶Whoever answers firstSEP 18, 2026▶Pick them up at BlockbusterSEP 14, 2026▶The forty-five eurosSEP 11, 2026▶Don't start with the tool. Start with the hour.SEP 9, 2026▶Sell the faceSEP 7, 2026▶Three missed calls and a ladderSEP 4, 2026▶I find your MondaySEP 2, 2026▶The company that was fineSEP 1, 2026▶Lesson learned: embrace the momentJAN 18, 2024▶Revolutionizing CX through hyper-personalization strategiesFEB 14, 2023▶The future of e-commerce meets the power of social mediaJAN 31, 2023▶The impact of AI on e-commerceJAN 27, 2023