A logistics company is operationalizing an AI solution to optimize delivery routes.

In CPMAI and PMI-aligned AI lifecycles, integrating diverse data types from multiple operational systems is typically handled through robust data engineering pipelines, most commonly implemented as ETL (extract, transform, load) or closely related ELT patterns. For a logistics optimization use case, the AI system needs to bring together traffic patterns (often from external or sensor feeds), internal delivery schedules, and vehicle performance/telematics data into a consistent, analyzable structure.

An ETL pipeline is designed precisely for this: it extracts data from heterogeneous sources, transforms it into common formats and schemas (handling units, timestamps, geocodes, data quality rules), and loads it into a target store (data lake, warehouse, or feature store) that downstream AI components can consume. CPMAI emphasizes that this integration work is a core part of the Data Understanding and Data Preparation phases, because AI models depend on unified, high-quality inputs rather than fragmented, siloed feeds. While real-time frameworks, federated models, or warehouses may play additional roles, the primary method explicitly focused on integrating diverse data sources into a coherent whole is an ETL pipeline, making option B the best fit.