A flexible yet governed data architecture is crucial for supporting AI/ML innovation, which requires a balance between flexibility to support various data structures and integration styles, and governance to mitigate risks related to data quality, privacy, intellectual property, bias, and explainability. Data architectures encompass three layers: infrastructure, integration, and access, each requiring flexible elements that support the needs of AI/ML projects. A flexible infrastructure supports open table formats such as Apache Iceberg, a flexible integration layer uses tools like CData that integrate multi-structured data using various styles, and a flexible access layer accommodates many types of analytical models through open APIs and easy integration with a vibrant commercial and open-source ecosystem. However, flexibility creates complexity, and complexity raises the risk of mishandling data, so data teams must maintain vigilant oversight and control of data usage through governance across all three layers, including observability, validation, lineage, access controls, and masking.