Predictive Asset Intelligence
A predictive maintenance dashboard that uses SAP RPT-1 to forecast remaining useful life across a building estate, with no model training, no labelled pipeline, and a single API call.
Built for the SAP UK Experience Centre to demonstrate SAP’s RPT-1 model doing something most customers assume requires months of ML engineering work.
The scenario is a facilities management contractor responsible for a multi-building government estate: pumps, boilers, lifts, HVAC units, generators. The dashboard opens with all ten assets unscored. No predictions, no risk ratings. The question: which one is going to fail first, and when? RPT-1 answers by receiving the current assets alongside 200 historical maintenance records (with known failure outcomes) in a single API call. No training, no fine-tuning, no labelling pipeline. The model learns the relationships between age, service history, operating load, and fault frequency from the historical context and applies them to the current estate. As predictions populate one by one, healthy assets appear first. The critical pump lands last, red border pulsing: fourteen years old, three faults in twelve months, last serviced nearly a year ago.
What follows shows the full action loop. A plain-English explanation of why the model flagged the asset. A what-if slider showing how a recent service would change the prediction. A work order raised into SAP Plant Maintenance. The shift from reactive threshold alerts (“this machine is overdue for a service”) to actual remaining useful life estimates is the business case. For asset-intensive industries, that difference is the gap between a planned maintenance window and an unplanned failure.
The entire stack runs on SAP BTP, with all AI models accessed via SAP AI Core.