You are Claire Vasseur, newly appointed Head of Innovation & Knowledge at Atelier Numérique, an engineering firm in Toulouse. Your firm pays every year to solve problems it has already solved. Before you spend a euro, you have to decide which knowledge to go after first.
The situation
It is June 2026. A rail-signalling project has just closed four months late and €4.2 million over budget. The post-mortem finds the painful detail: the team spent six weeks on a real-time-determinism problem that the Toulouse aerospace group had solved, and written up, eighteen months earlier. The answer was inside the firm. It never moved.
Atelier Numérique has 1,600 engineers across four sites (Toulouse, Lyon, Nantes and Berlin), turns over €210 million at an 11% operating margin, and grew by buying engineering shops that still run as sealed units. An internal sample puts about 30% of the hours on new projects into re-solving what was already solved: roughly €11 million a year. Proposals that combine two sectors, the firm's best-margin work, have fallen from seven to two in three years. And three of the most senior architects, whose judgment has never been written down, retire within eighteen months.
This is not the first attempt. A mandatory documentation portal, with its lessons-learned template, settled at 9% compliance: engineers bill by the hour, and a template nobody reads costs them an hour they could charge. The CEO has given you €1.2 million a year, a cap of 5% of engineers' time and twelve months. Your dashboard starts with the reuse rate at 9%, against a target of 55%, and voluntary participation at 8 out of 100.