In this session, Marie Cattelin, Senior Product Owner at Bettermile, presents an end-to-end, real-world case study on building and deploying AI-powered Smart Insights for last-mile operations—moving beyond dashboards to deliver clear, actionable recommendations that actually change user behavior (UK spelling changed to US for consistency; keep original if UK spelling is preferred).
Marie’s team faced a common but critical challenge: translating massive volumes of heterogeneous operational data into context-aware productivity insights for depot employees and subcontractors, who neither have the time nor the expertise to analyze (corrected from “”who have neither… to analyse””) charts and tables. Compounding the problem, last-mile operations vary significantly by region, culture, seasonality, and operational model—meaning a “one-size-fits-all” insight simply does not work.
The session walks step-by-step through the strategy, trade-offs, and implementation choices behind Bettermile’s Smart Insights initiative:
Establishing data reliability and readiness as a prerequisite for AI
Designing concise, high-signal summaries that surface the single most impactful operational problem
Evaluating architectural choices such as precomputed (corrected from “”pre-computed””) vs. on-demand AI generation, balancing cost, latency, and scalability
Navigating constraints around legal frameworks, data governance, and processing costs
Ensuring the AI delivers the right recommendation to the right user at the right moment
While the product is still evolving, the talk demonstrates how the team measures impact through built-in feedback loops, qualitative user signals, and early adoption metrics—closing the loop between AI output and real operational value.
Attendees will leave with a concrete, production-tested blueprint for turning AI from “interesting analytics” into trusted, actionable decision support—without overwhelming users or forcing AI where it doesn’t belong.
