From 14-hour batches and poor documentation to AI-ready data: SafetyCulture's dbt rebuild

SafetyCulture has been a long-time user of dbt. Our adoption of dbt started in 2020. Since then, we have had great experiences, but our project was far from ideal for people's usage and AI usage. Our execution time was slow, 14 hours for a daily batch; our documentation was poor, and our data quality was questionable. We decided to take a journey to rebuild our data model, ensuring quality, performance and AI readiness. This has been one of the most successful projects internally in SC, where we reduced execution time by 90% and enabled AI BI for the whole business, with adoption exceeding 50%. We were able to save more than US$140K in infrastructure costs.