Salesforce Chairman and CEO Marc Benioff posted on his X account, welcoming Anthropic Co-founder and CEO Dario Amodei to the Salesforce annual conference Dreamforce 2026.

During the meeting, Amodei and Benioff jointly announced that the partnership between Salesforce and Anthropic will further expand, integrating the Claude large model more deeply into Salesforce's product ecosystem; it was disclosed that currently, 7,000 Salesforce employees are using this integrated tool in their actual work, and as of early June this year, Salesforce's stake in Anthropic was valued at approximately $5 billion.

Amodei assessed the current level of AI technology adoption, stating that the "actual value that has penetrated and been used" is only 5% to 10% of its total current value, indicating that the application of AI technology in the entire economic system is still in its early stages, with huge untapped potential. Benioff noted that the progress of companies in adopting AI technology is "uneven," and promised that Salesforce will "do more to help customers accelerate this transformation."

At the conference, Salesforce also launched two new products: one is a new system called AIforce, which allows customers to access data through AI agents instead of relying on traditional interfaces, enabling users to build custom interfaces in real-time; the second is a new sales inference model named Koa, developed over six months based on NVIDIA's Nemotron foundational model, trained using proprietary synthetic data derived from thirty years of corporate workflows.

Notably, this Dreamforce conference itself has become a focal point in the current debate in the AI industry on whether to slow down the pace of development—Amodei had just called for the industry to "urgently slow down" AI development, warning of serious consequences if not controlled; while Salesforce President Patrick Stokes responded that "as builders, slowing down a bit is actually good news because we have a lot of catching up to do, and we can use this time to catch up." Meanwhile, NVIDIA CEO Jensen Huang refuted concerns about the pace of AI development at the All-In conference on Monday, during which he received a supportive call from President Trump.

From a market mechanism perspective, Salesforce's holding of approximately $5 billion in Anthropic equity and the deep integration of Claude into enterprise-level products essentially represents a "equity binding + product integration" dual bundling between an application layer software giant and a cutting-edge model lab: Anthropic gains a stable distribution channel covering a large number of enterprise customers and real workflow training data (like the synthetic data used for the Koa model), while Salesforce leverages Claude's model capabilities to launch new AI-centric products like AIforce, seizing the entry point for enterprise-level AI applications. Amodei emphasized that the AI value "penetration rate is only 5%-10%", which objectively also advocates for a longer commercialization window and greater imaginative space for the entire industry. Who benefits: application layer companies like Salesforce that deeply bind top model labs through equity and product dual binding can share both the model capability dividends and their own distribution channel advantages; who is under pressure: application layer competitors that hold different positions in the debate on "whether AI should slow down" and have not established similar deep binding relationships with top labs may be at a relative disadvantage in the competition for enterprise-level AI application entry points.

Source: Public Information