Intersolia had a queue of 30,000 documents. We built a pipeline in which the exceptions go to reviewers. The queue was empty after six months. Since then the pipeline handles more than 1,000 documents a day, three times the manual rate, and more than 96 percent pass the quality check.
Supplier product specifications arrived in at least twelve languages. A five-week proof of concept turned them into finished product pages in the right language. Processing time went from hours to minutes. One person published what used to take several. The pages passed the retailer’s own language review.
The chain wanted an ordering assistant. Over a three-month discovery, the prototypes showed the gap between expectation and what the technology could do. We recommended waiting and wrote the roadmap instead.
“While we didn’t get the AI we initially wanted, we got something more valuable — a clear understanding of what’s actually possible and a trusted advisor who could explain it without the technical jargon.”
— Digital Innovation Manager
The op-ed in Dagens Industri, “HR är missförstått men inte oskyldigt”, was written together with our client House of People.
Data pipelines and retrieval over your own material, language models and agents that carry out tasks. Everything runs in your environment and is yours after handover.
Large document volumes go through a supervised pipeline, and your people review the exceptions.
Your team builds with Claude Code for one day, starting from your own work. A survey before the day sets the content, and the know-how stays with you.