In a recent discussion at VB Transform 2025, AI leaders from General Motors, Zoom, and IBM shared critical insights on the ongoing debate between open, closed, and hybrid AI models for enterprise use. The choice of model architecture is not just a technical decision but a strategic one, with each option presenting unique trade-offs in terms of cost, security, and performance.
Barak Turovsky, GM’s first Chief AI Officer, highlighted the constant buzz surrounding new model releases and leaderboard shifts. Drawing from his experience in launching early large language models (LLMs), Turovsky emphasized how open-sourcing AI weights and training data has historically driven major breakthroughs, yet cautioned that enterprises must weigh innovation against control.
Representatives from Zoom and IBM echoed the sentiment, noting that closed models often provide enhanced security and proprietary advantages, which are crucial for sensitive enterprise applications. However, they can come with higher costs and less flexibility compared to open models, which foster collaboration and customization but may pose risks in data privacy.
The panelists agreed that a hybrid approach might offer the best of both worlds for many organizations. By combining elements of open and closed systems, enterprises can tailor solutions to specific use cases, balancing innovation with the need for robust governance and security protocols.
As generative AI matures, companies are moving beyond simple chatbots to deploying intelligent, autonomous agents. This shift demands a deeper evaluation of model selection to ensure scalability and alignment with long-term business goals, a point underscored by all three industry leaders.
Ultimately, the decision on AI model architecture will vary based on an organization’s priorities, resources, and risk tolerance. The insights from GM, Zoom, and IBM serve as a valuable guide for enterprises navigating this complex landscape, urging a thoughtful approach to AI adoption.
Maria Lourdes
