To resolve data bottlenecks and pursue its AI ambitions, ICA architected the Business Intelligence Cloud Analytics (BICA) platform on Google Cloud, centered on BigQuery. This shift represented a move away from traditional, restrictive Extract, Transform, Load (ETL) tools toward an engineering-led approach. By building their own code-based frameworks to automate ingestion, security, and connectivity, ICA’s IT teams gained the flexibility to scale repeatable processes and adapt as business needs change.

ICA now streams high-volume store data, including real-time receipts and wholesale movements, through Kafka into BigQuery. By performing transformations in a unified SQL environment, the team applies software engineering best practices—such as version control and modular modeling—to their data pipelines. This reduces technical debt and allows analysts to iterate faster using a common language, accelerating the delivery of data insights to drive better business decisions.

Security was built into the architecture from the start to meet strict regulatory requirements. The team engineered a custom encryption service to protect sensitive data at the column level. To allow data scientists to work with real information without compromising privacy, ICA implemented specialized decryption views. These views decrypt data only in transit when viewed by an authorized user, meaning the unencrypted data is never stored at rest in BigQuery. This approach eliminates the need to create mocked data when real data is too sensitive to access, which can often lead to inaccurate results. For added resiliency, ICA uses a hybrid key management strategy where a master key on an external cloud provider is required to unlock the key sets on Google Cloud.

This secure data foundation has significantly increased ICA’s ability to innovate with AI. The company migrated its machine learning workloads—including its personalized recommendation engine—to Gemini Enterprise Agent Platform, improving training and deployment speeds while lowering operational costs. Gemini Enterprise Agent Platform provides a managed environment that handles the underlying infrastructure, allowing ICA’s scientists to focus on refining models rather than managing servers.

This architectural agility has transformed ICA’s ability to prototype. A recent proof of concept with ICA Bank was completed in just two and a half months—a process that would’ve taken approximately 18 months previously. This development speed increase is possible because the data and technical capabilities are now readily available in a shared environment.