Feature
White Paper
The Future of Leadership Development: AI-Driven Manager Enablement
When Salesforce unveiled Headless 360 at its TDX developer conference in April, the message was blunt: Agents do not click buttons, so the screen should no longer be the primary interface.
The company stripped the browser UI from its platform and exposed every capability as an API, MCP tool or CLI command. It is the clearest public commitment yet from a legacy enterprise vendor that it is redesigning its platform around the agent as the primary user.
The company is also hedging on the protocol underpinning the move. Salesforce's engineering leadership has publicly acknowledged that Model Context Protocol may not remain the de facto standard for AI agent communication, an interesting statement from a company rebuilding its platform around the protocol. Analysts are now asking whether Headless 360 represents architectural reinvention, or a reframe of what Salesforce already had.
When APIs Become the Interface
Salesforce has always had APIs. The difference with Headless 360 is that the interface is no longer the center of the product.
The shift is valuable to specific workflows rather than the platform as a whole, said Terra Higginson, a principal research director at Info-Tech Research Group. The practical change is significant, she said: "You will no longer need to go to the actual record, find the contact and then try to figure out which piece of communication contained a specific attachment or detail.” A natural-language query does the work instead.
Higginson is most interested in the Agentforce Service portfolio. If Salesforce makes service, sales, marketing and contact center workflows feel native to the same platform, it could reduce the need for separate point solutions for customer experience. That is where the platform decision changes for buyers, she said. "They are no longer just comparing CRM features or CCaaS features,” she said. “They are comparing which vendor can execute across the customer journey with the least friction."
The Seat-License Problem
Beyond the architectural question, there’s the commercial one. Salesforce built its business on licenses sold to humans. Its pricing model, training ecosystem, partner network and administrator workforce all assume a person in front of a screen. Headless 360 gets closer to consumption-based pricing.
Enterprise SaaS pricing was always calibrated to human usage, said Cache Merrill, founder of software development firm Zibtek. As agents run workflows around the clock, the definition of value, and of liability, change in ways that incumbent pricing models weren’t designed to accommodate.
Buyers are no longer comparing CRM features in isolation, but asking which vendor supports the customer lifecycle most easily, Higginson said. That is a different purchase decision than per-seat licensing, and one that legacy vendors are only beginning to address.
Governance Isn’t Keeping Up
The biggest unresolved question in Headless 360 is operational. When an agent runs a consequential workflow autonomously and gets it wrong, it’s the enterprise’s problem, not Salesforce’s.
"Once you start allowing software to be truly autonomous and on a large scale, weak governance and sloppy operational logic gets a lot more hazardous,” Merrill said.
"AI can help people work faster and better, but enterprises should not let autonomous agents execute consequential actions without governance, permissions, auditability and escalation paths,” Higginson agreed.
Agentic AI still needs human oversight in most enterprise contexts, Higginson contended. "I am still not seeing many places where agentic AI should just fly solo," she added. "Ever."
One specific problem is agents bypassing controls or deleting data. Human users have always served as informal error-correction layers, catching mistakes, skipping irrelevant steps and working around faulty processes, Merrill said. Because agents don’t, weak governance and sloppy operational logic become an organizational hazard, he said.
Agents Cannot Fix What’s Already Broken
The biggest problem with Salesforce's agent-first pitch is that many customers aren’t getting the full value from the platform they already have.
The problem is rarely that users cannot navigate the UI, but that underlying workflows have become complicated, fragmented and poorly maintained over years of accumulated technical debt, Merrill said. Portions of the Salesforce platform continue to run on decades-old logic, pushing the market toward niche, best-of-breed plugins as native capabilities fell behind.
More generally, CRMs have overpromised and underdelivered for a decade, Higginson contended. Customer experience does not live only in sales; it moves continuously across marketing, service and back again.
With Agentforce Service, Salesforce has its best opportunity to fix that by connecting the customer lifecycle.
Whether Headless 360 delivers on that opportunity depends less on the architecture than on the organizations deploying it. "The most successful users of these systems will be ones that already have clean processes, solid data and high internal ownership of their automation,” Merrill said. For everyone else, the agents may just execute poorly.
Conference
Jun
2
Gartner Application Innovation & Business Solutions Summit Las Vegas 2026
Register
Webinar
Jun
17
Engaged, But Not Convinced: What's Eroding Beneath Your Strongest Engagement Numbers
Perceptyx EX Benchmark '26: leadership trust, change confidence and conviction gap truths at large enterprises.
Register
Webinar
On demand
Retention Is a Communication Problem: New Data from 6,000+ Employees
This session gives leaders a clear diagnosis and practical path to treat retention as an infrastructure problem.
Watch Now
Webinar
On demand
Do Learning Programs Really Work? How to Turn Education Into Engagement In Healthcare
See how leaders are using learning programs to build trust with healthcare professionals and create measurable engagement.
Watch Now
Editor's Note: How else are vendors updating their offerings to support AI agents?
- Asana Buys StackAI for $75M to Put AI Agents to Work Across the Enterprise — The $75 million deal gives Asana a no-code engine to build AI agents that act across Salesforce, Oracle and other core enterprise systems.
- Microsoft Built the Agent — Now Enterprises Have to Build the Guardrails — Copilot Cowork is Microsoft's most capable agent yet. The gap isn't the tech: it's the governance, accountability and data hygiene most enterprises don't have.
- SAP Launches Unified AI Platform, 50+ AI Assistants and Autonomous HR at Sapphire 2026 — SAP used Sapphire 2026 to make its biggest push into agentic AI, launching unified AI platforms and autonomous agents across HR, finance and ERP.