Cohere VP says enterprise AI sovereignty requires control of the full agent stack at VB Transform 2026

2026-07-15 · Source: VentureBeat · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Cloud Computing & IT Infrastructure · Depth: Intermediate, quick

Summary

At VB Transform 2026, Rachad Alao, VP of Product Engineering at Cohere, asserted that enterprise AI sovereignty demands complete control over the full agent stack, from infrastructure to governance systems and agent frameworks. Alao, formerly of Google and Meta, stressed that for mission-critical systems, tight control over data residency and AI operations within understood jurisdictions is crucial. He explained that despite falling inference prices, agentic workloads exponentially increase token utilization due to complex tasks requiring extensive processing and tool interaction. Cohere advocates for model routing, using the "right model for the task at hand" to solve problems privately and securely. This involves deploying smaller models like North Mini Code, which runs on a single Nvidia H100 GPU for 80% of coding use cases, and Command A+, a 218-billion-parameter mixture-of-experts model with a four-bit compressed version and Apache 2.0 license. Alao also highlighted multimodal search as integral to agentic workflows and positioned Cohere's governance layer to prevent vendor lock-in.

Key takeaway

For AI Architects evaluating enterprise AI deployments, prioritizing full control over the agent stack is critical for true sovereignty and data security. You should implement robust governance layers to route workloads based on data sensitivity and intelligence needs, preventing vendor lock-in. Consider deploying smaller, specialized models for common tasks to optimize costs and performance, reserving larger frontier models for highly complex, less sensitive operations. This strategy ensures data residency and operational control.

Key insights

Enterprise AI sovereignty requires full control of the agent stack, including data, infrastructure, and model routing, to manage complex agentic workloads.

Principles

• AI sovereignty extends beyond firewalls.
• Agentic workloads increase token consumption.
• Route tasks to the right model.

Method

Implement a governance layer to route requests among models based on intelligence required and data sensitivity, integrating multimodal search as an agent tool.

In practice

• Deploy smaller models for 80% of use cases.
• Use on-premises models for regulated data.
• Leverage Apache 2.0 licensed models for flexibility.

Topics

• AI Sovereignty
• Enterprise AI
• Generative AI Agents
• Model Routing
• On-premises AI
• Multimodal Search
• Vendor Lock-in

Best for: AI Engineer, Machine Learning Engineer, CTO, Director of AI/ML, AI Architect, MLOps Engineer

Related on AIssential

• What is sovereign AI — and why it will decide the winners and losers of the AI race — Genuine sovereign AI demands comprehensive control over data, infrastructure, stack, rules, and finances, making open-source architecture essential.
• Enterprise AI - Cohere — Cohere's enterprise AI strategy emphasizes partnerships, specialized applications, and addressing governance and maturity.
• Establishing AI and data sovereignty in the age of autonomous systems — Enterprises are prioritizing AI and data sovereignty to mitigate IP risks associated with third-party generative AI models.
• When the sovereign AI diagnosis goes prime time — The global AI landscape is shifting towards national and enterprise control over AI infrastructure, data, and models.
• IBM Report: Most EMEA Executives Don’t Fully Understand Their AI Dependencies — Many EMEA executives lack critical visibility into AI dependencies, risking costs and operational stability.
• Announcing strategic MOUs with Indra Group and Multiverse Computing - Cohere — Digital sovereignty in AI is achieved through collaborative partnerships and flexible deployment options across diverse environments.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by VentureBeat.