Case study
Case studies

Real stories. Numbers, not promises.

We only publish results we have measured. When we do not have a number we say so, and when we measure one it ends up on this page. Every case can be tried out.

Case 1 · with the numbers
E-commerce · food retail · 2,000+ SKUs

From 40 minutes a page down to under 5.

The problem

A large and growing catalogue, with pages written by hand one at a time. Forty minutes of work per product across attributes, description and SEO, with two people stuck on data entry instead of selling. New products sat in the warehouse because the page was not ready.

What we did

A pipeline that takes photos, supplier PDFs and management-system exports and produces the complete page: attributes, description, SEO and translations, with the client's rules and glossary. The quality check runs before a human looks at anything, and data that is missing is declared rather than filled in.

What we learned

The bottleneck was not writing, it was deciding: human review is needed, but it is needed on a few pages, the ones the check flags. The rest go through. That is why the approval step stayed, and why we are not removing it.

<5′per page
−65%data entry
3 daystime to market

measured on 2,000+ SKUs · food retail · quarter on quarter

Try it now
Raw documents going in, publishable product pages coming out
from the supplier export to a publishable page · AI-generated image · for illustration
Cases 2 and 3 · qualitative
Dental practices · WhatsApp chatbot

Reception no longer answers the same questions.

The problem. The phone rang all day about opening hours, treatments, prices and bookings. Reception answered the same ten questions, and anyone calling about an emergency got a busy line.

What we did. An assistant that answers on the website and on WhatsApp using only the practice's own information: hours, treatments, published prices, how to book. When the answer is not there, it says so and hands over instead of inventing an opening time.

The result. We do not have a publishable number and we are not inventing one: what changed is that the repetitive questions no longer reach the phone, and reception sees them only when they genuinely need a person.

Marketing · video and visuals with AI

A steady flow of content.

The problem. New content was needed every week, and every piece meant a shoot: set, crew, a full day, post-production. The cost per piece made the frequency the channels demand impossible.

What we did. The product stays real, the scene is built in generation: stills and motion, landscape for the site and vertical for social, from brief to finished file inside a day.

The result. Six pieces to look at below, each with one line on how it was produced. The number that matters is the cost per piece, and that one we work out together on your case.

The next case

The next number is yours.

The ones above are measured on somebody else's catalogue. On yours they get measured from the start: before we begin we photograph what the process costs today, so at the end the comparison exists and nobody has to take our word for it.

Talk to us