At Dreamforce, business leaders say older AI models are enough

Key Points:

• At Salesforce's Dreamforce conference, CEOs from Anthropic, OpenAI, and Nvidia discussed AI safety concerns, with Nvidia's Jensen Huang advocating for rapid AI development despite calls for slowing progress from other leaders.
• Many attendees and Salesforce customers expressed challenges in fully utilizing existing AI technologies, with a focus on adopting older, more affordable models rather than cutting-edge ones like Anthropic's Claude Fable 5.1 or OpenAI's GPT-6 Astra.
• Companies such as Docusign and Nice are employing strategies like model routing to balance AI performance and cost-efficiency, highlighting the shift from traditional SaaS subscription models to token-based pricing for AI services.
• While frontier AI models offer advanced capabilities, most current business applications rely on previous-generation AI, indicating a gap between the technology's potential and practical adoption in sales and customer service workflows.
• The conference atmosphere reflected a mix of excitement and caution, with some firms waiting months before integrating the latest AI models, underscoring a measured approach to AI adoption amid ongoing debates about safety and regulation.