AI Sovereignty: Control Your Data, Your Stack

At VB Transform 2026, hundreds of enterprise leaders gathered to discuss generative AI agents driving business outcomes. Cohere's VP of product engineering, Rachad Alao, shared insights on achieving AI sovereignty.

Alao defined AI sovereignty as more than just running applications behind a firewall. It involves tight control over data residency, AI operations, and the entire technology stack, from GPUs to governance systems. This control is paramount for organizations managing mission-critical systems like banks, hospitals, and governments.

He also addressed the economic argument for smaller models, noting that while token prices fall, overall consumption is rising rapidly due to complex agentic use cases. These agents require significant processing and multiple steps to complete tasks, leading to exponential increases in token utilization.

Cohere's strategy, Alao explained, focuses on helping enterprises solve complex problems privately and securely, emphasizing the use of the right model for each task. This approach contrasts with providers who prioritize maximizing token utilization. Model routing allows organizations to direct sensitive or regulated workloads to appropriate models, enhancing efficiency.

The discussion also touched upon the evolution of search, which is becoming an integral part of agentic workflows, moving beyond simple text retrieval to multimodal capabilities. Data control and the ability to avoid vendor lock-in remain key drivers for enterprises seeking greater autonomy in their AI deployments.