Salesforce just took a bold step in its AI strategy. It’s expanding partnerships with OpenAI and Anthropic to embed their advanced models into its new AgentForce 360 platform, aiming to make generative AI a core function of its CRM and enterprise tools Under the expanded agreement, OpenAI’s GPT-5 will be integrated more tightly into Salesforce products. Users will be able to access Salesforce data, build analytics visualizations, and interact with Customer Relationship Management (CRM) functions directly from ChatGPT Simultaneously, Salesforce will offer Anthropic’s Claude models, particularly for regulated sectors like finance and healthcare, hosted in Salesforce’s secure environment.
One standout feature is AgentForce Commerce, which will allow merchants to sell products via ChatGPT’s Instant Checkout but Salesforce retains control over order fulfillment and customer data The integration reaches into Slack as well via ChatGPT and Salesforce’s infrastructure, teams will be able to generate summaries, automate workflows, and embed AI agents in daily communications.
From a journalist’s perspective, three risks and opportunities catch my attention
Opportunity in enterprise lock-in By embedding AI deeply into its core stacks (CRM, Slack, commerce), Salesforce is positioning itself as the platform on which generative AI runs turning AI from a feature into a foundational layer
Regulatory and security exposure For sectors with strict compliance needs (health, finance), handling sensitive data through AI pipelines carries risk. Even if Anthropic’s models are isolated, the architecture must prove airtight
Dependence on external models Salesforce is still relying on models built elsewhere. If OpenAI or Anthropic shift direction, raise prices, or face constraints, Salesforce is vulnerable. The bet is that this integration strength outweighs that dependency
This is Salesforce signaling its ambition to be more than a CRM vendor it wants to be the “AI operating system” for business. Whether it succeeds will hinge on execution, trust, and regulatory climate