Cultivating Communities — Make Atelier Numérique Learn Faster

A four-round, advanced knowledge-and-innovation-management simulation set inside Atelier Numérique S.A., a 1,600-engineer French engineering-and-technology services firm in Toulouse (EUR 210M revenue, 11% margin) that grew by acquisition into rigid, project-siloed business units. After a EUR 4.2M rail-signalling loss — caused by a real-time-determinism problem the Toulouse aerospace group had already solved and archived eighteen months earlier — the CEO appoints a Head of Innovation & Knowledge with a EUR 1.2M annual budget, a 5%-of-engineers'-time cap, and 12 months to demonstrably accelerate organisational learning without disrupting billable delivery. The firm re-invents ≈30% of new-project hours (≈EUR 11M/year of recoverable waste), its tacit expertise lives in three near-retiring senior architects, and a previous top-down 'Knowledge Management' portal earned 9% compliance. Through Dupouët's lens of community dynamics, Wenger's communities of practice, the SECI tacit/explicit model and absorptive capacity, you (1) map the knowledge flows and classify domains by tacitness and breadth, deciding which need cultivated communities versus codified hierarchy; (2) allocate the EUR 1.2M across structural options — communities of practice, a codified repository, expert directories, rotation/shadowing, an internal conference and digital tools — matching structure to knowledge type without repeating the repository reflex; (3) cultivate engagement and legitimacy — resource community coordinators, ring-fence billable time and defend it when a delivery director demands the 5% be cut to 2%, secure sponsorship without over-formalising the community into bureaucracy, and draw the retiring architects' tacit knowledge into circulation before they leave; and (4) commit a 12-month learning architecture with quarterly KPI checkpoints and pitch the CEO a credible structure-to-outcome chain. The math rewards genuine, cultivated, well-matched community-and-hierarchy design and punishes the five classic errors — the repository reflex, mandating communities into existence, one model for all knowledge, no protected time, and caving to delivery pressure. Final KPIs track Knowledge-Sharing & Reuse Rate %, Community Engagement, Innovation Output and Value Recovered against the EUR 11M waste.