Enterprise Data Platforms Require Design Changes to Support AI Workloads

Many organizations are adding AI to their analytics and data systems for predictive insights and generative applications. Govinda, Senior Manager at Cognizant with 15 years of experience in SAP and non-SAP data analytics, stated that AI initiatives often fail because underlying data platforms were not designed for AI at scale.

Most enterprise data platforms were built to aggregate structured data, support dashboards, and enable historical analysis. These systems frequently encounter data fragmentation across operational systems, cloud warehouses, and data lakes, making it difficult to create consistent datasets for AI models.

Semantic inconsistencies that are tolerable in reporting environments can produce unreliable AI outputs. Batch processing common in legacy architectures also restricts real-time or near-real-time AI applications such as anomaly detection.

Govinda said organizations should build unified data foundations using platforms such as SAP Datasphere, Snowflake, Databricks, and Google BigQuery. These systems combine data from multiple sources while maintaining governance and scalability. Shared business definitions across teams help ensure AI models and analytics reports operate on the same logic.

Event-driven architectures and streaming pipelines allow data to be processed as it is generated. Governance must be embedded in data pipelines, metadata layers, and access controls to reduce the risk of unreliable AI outputs.

Even with updated platforms, scaling AI beyond initial use cases often depends on organizational alignment and continuous model monitoring. Govinda described a utilities analytics project where differing interpretations of billing metrics across platforms required standardized definitions and clearer ownership to improve trust in results.

Technology leaders should focus on building data environments that remain consistent and reliable as business needs evolve, according to the account.