Salesforce is accelerating the realization of the “Agentic Enterprise”—an era where software evolves from a tool that assists humans into a “teammate” that autonomously makes decisions and completes tasks—centered on its autonomous AI agent platform, “Agentforce 360”. This strategic shift follows a significant industry crisis in January 2026, when major SaaS companies saw approximately 15 trillion yen (roughly \$100 billion) in market capitalization vanish, signaling a structural transition from traditional per-user subscription models to a new era where AI creates value via APIs.
- Launch of the Atlas Reasoning Engine for Execution
- Design Balancing Deterministic Rules and AI Reasoning
- Open Platform Strategy Connecting with External Systems
- The 2026 Shock to Traditional Seat-Based Pricing
- “Headless 360” – Liberation from UI and Increased API Value
- Reliability Governance and the Importance of Audit Logs
- Reaching the “Rule of 40” and Re-Accelerated Growth
Launch of the Atlas Reasoning Engine for Execution
In the second quarter of 2026, Salesforce launched the General Availability (GA) of the “Atlas Reasoning Engine,” the brain of its AI agents. This represents a fundamental shift from traditional “Runbook” models that follow predefined manuals to “Reasoning” models that autonomously determine the best actions based on input. Specifically, it analyzes customer inquiries, refers to CRM data within Salesforce, and autonomously decides and executes tasks such as generating next actions, creating tickets, or sending follow-up emails. The following diagram illustrates the four layers of AI agent task delegation.
Design Balancing Deterministic Rules and AI Reasoning
The core capabilities supporting AI agent construction are “Agentforce Builder” and “Agent Script,” which became generally available in the spring of 2026. Salesforce advocates a concept called “Guided Determinism” (Hybrid Reasoning), allowing developers to differentiate between parts where the AI has autonomy and parts where it must strictly follow specific rules within a single script. For instance, a rigid business rule such as “offer free shipping if the order total exceeds \$100” is controlled by code, while the natural greeting and tone of the conversation are left to the AI. This approach mitigates “hallucinations” typical of generative AI while enabling flexible customer interactions. Developers can design agents via the “Agentforce Studio” app through three methods: natural language instructions, a block-based Canvas view, or by directly editing .agent files in the Script view. This flexible design philosophy allows even non-IT personnel to rapidly build specialized “AI butlers” tailored to their company’s specific needs.
Open Platform Strategy Connecting with External Systems
Instead of remaining within a closed ecosystem, Salesforce is promoting an “Open Agent Platform” strategy to connect with external tools and third-party AIs. Central to this is the adoption of the “Model Context Protocol” (MCP), a common standard for linking AI with external tools. In April 2026, Salesforce Hosted MCP Servers were made generally available, allowing Agentforce to securely utilize data from external applications like Google Drive and Slack. Furthermore, the company is involved in defining the “Agent-to-Agent” (A2A) standard, aiming to lead a multi-agent world where its own agents can delegate tasks to specialized agents from other providers. Symbolizing this ecosystem expansion, the latest models like OpenAI’s GPT-5 and Anthropic’s Claude were integrated into Agentforce 360 in October 2025. This enables client companies to select the optimal AI model for their needs and operate it in combination with Salesforce’s robust data infrastructure.
Redefining the SaaS Model and Recovering from the “15 Trillion Yen Crash”
The 2026 Shock to Traditional Seat-Based Pricing
In January 2026, a seismic shift occurred in the software industry, often referred to as the “Death of SaaS”. The catalyst was a demo by the AI developer Anthropic for its AI agent “Claude Cowork,” which showed an AI recognizing computer screens and performing mouse operations just like a human to complete tasks. Consequently, four major U.S. SaaS companies, including Salesforce, saw their combined market capitalization drop by approximately 15 trillion yen in less than a month. Investors feared that if an advanced AI agent could perform the work of 100 people, the “number of users (seats) x monthly unit price” model—the pillar of SaaS revenue—would collapse. In a world where users no longer interact with “screens” to operate systems, traditional differentiators like UI/UX usability become valueless, risking the relegation of SaaS to mere data storage. This crisis was not just temporary panic; it was a historical turning point where the primary user of software shifted from humans to AI.
“Headless 360” – Liberation from UI and Increased API Value
Salesforce’s response to market concerns was the announcement of “Salesforce Headless 360” in April 2026. This platform allows CRM functions to exist beyond human-oriented screens, enabling AI agents to call them directly via APIs, MCP, and CLI. Salesforce’s value has shifted from “sophisticated input screens” to “business logic and trusted data” that AI agents can access with verified grounding. Manual data entry by users is being replaced by background autonomous execution, with some sales teams already reporting a 90% reduction in SFA (Sales Force Automation) entry labor. As software value moves from “human-facing GUIs” to “AI-facing interfaces,” the company is positioning itself as an orchestrator that allows AI agents to handle CRM foundations securely and efficiently.
Corporate Governance and New Paths for Revenue Growth
Reliability Governance and the Importance of Audit Logs
In an era of autonomous AI agents, ensuring “reliability” is paramount for enterprises. To visualize AI decision-making, Salesforce implemented “Reasoning Trace” for its Enterprise Tier and above, recording “what was referenced and why an action was chosen” in a Who/When/What/Why/Result format. The following diagram highlights four critical points in AI agent governance.
Reaching the “Rule of 40” and Re-Accelerated Growth
Strategic AI pivots are beginning to show positive effects on Salesforce’s financial metrics. Revenue in 2026 is projected to reach approximately \$46 billion, representing roughly 11% year-over-year growth. This indicates a re-acceleration from 10% growth in 2025, and combined with an operating margin of 33%, the company is expected to exceed the “Rule of 40″—a key metric for healthy software companies where growth and profit margins total 40% or more. As of the third quarter of fiscal year 2025, the annual run rate for Agentforce reached \$540 million, with over half of new contracts coming from additional adoption by existing customers. The market has re-evaluated the company, shifting from an “AI loser” to a practical and profitable AI leader, with many analysts expecting stock recovery throughout 2026. Moving forward, transitioning to outcome-based pricing models—where monetization is tied to the volume of work processed or the value created by AI agents—will be key to sustained corporate value.
Reference Pages
-
[Digital Agency of Japan] Trends in AI Agent Governance https://www.digital.go.jp/assets/contents/node/basic_page/field_ref_resources/96d0f18a-9860-479e-9917-8e622b10a26e/82877546/20260714_policies_ai_governance_report_01.pdf
-
[Salesforce] Agentforce Official Site https://www.salesforce.com/jp/agentforce/
-
[Salesforce] Atlas Reasoning Engine https://www.salesforce.com/jp/agentforce/what-is-a-reasoning-engine/atlas/
【#AIAgent #SaaS #Salesforce #DigitalTransformation #AIGovernance #Agentforce #Headless360 #Software】
コメント