Dik Harbrecht

This presentation explores how Shell standardised data product delivery across a large matrix organisation while preserving the autonomy at the heart of agile and data mesh. As we prepare for an AI-driven future, delivering trustworthy AI at scale first requires trustworthy data at scale. Rather than relying on highly technical reference architectures or centrally controlling how teams work, we introduced policies that define what every data product must deliver and standards that define common approaches for how products are named, governed and exposed. Inspired by Zhamak Dehghani’s data mesh principles and engineering concepts such as loose coupling, high cohesion and information hiding, the approach standardises the boundaries between producers and consumers while leaving implementation knowledge with local teams. The presentation shares the challenges that led us to this model, the thinking behind it, and concrete examples from our Lakehouse policies and standards.