Salesforce this week outlined a series of new products and capabilities that are centered on a “headless” framework the company is now exposing to multiple artificial intelligence (AI) agents.
Launched at the company’s TDX event, one of the highlights of the presentation was the display of cross-vendor orchestration between agents. Reps for Salesforce highlighted, for example, a customer inquiring via Slackbot about the status of an imminent flight and receiving a direct response from a customer service chatbot from the airline itself within the same app, in this case Finnair.
Rather than simply returning search results, the demo showcased how multiple AI systems can coordinate responses within a single interface. The approach moves beyond today’s assistant-style interactions toward more integrated, task-driven workflows.
Similarly, a focus of the presentation was the growing role of agents in customer service. This is part of what makes agent-to-agent communication and orchestration increasingly important for enterprises. As one of the premier CRM companies in the world, Salesforce has a large role to play in how agents will be adopted into the space.
The company is emphasizing a “headless 360” approach, making its platform accessible via APIs so developers and AI agents can access and interact with customer data without relying on traditional interfaces. Headless commerce refers to the separation of front-end user interfaces from back-end data and systems, allowing either to be changed without unintentionally impacting the other. This is important for agents because they don’t operate through graphical interfaces, instead requiring direct, programmatic access to systems in order to retrieve data, take actions and coordinate workflows across applications. Joe Inzerillo, president of enterprise and AI technology at Salesforce, also framed “headless 360” as part of the next generation of tools for developers, adding that “code agents have gotten really capable.”
One of the key insights was the general acceleration of development across the industry. “What used to take weeks now takes, in some cases, hours,” said Inzerillo. This dovetailed with a section of the webinar focusing on deterministic versus probabilistic approaches to AI, highlighting a broader shift from traditional software that follows fixed rules to systems that generate outcomes based on patterns and probabilities. As a result, development cycles are becoming less about defining every step in advance and more about managing and refining AI-driven behavior over time.
Ultimately, Salesforce wants to be more than a company that provides AI tools. It wants to be the platform where these tools are built, connected and managed. The success of this plan depends on how these tools can work together across different systems, vendors and business workflows. For now, the vision is clear, even if the path to consistent execution remains a work in progress.
