Dreamforce 2025 delivered everything you’d expect—big keynotes, polished demos, and plenty of excitement about AI. But this year, the tone was more grounded. Salesforce balanced ambition with depth, pairing its vision for AI with the engineering work that makes it real: stronger architecture, observability, and control.
When Agentforce launched last year, it marked Salesforce’s proof of concept for a new kind of AI—intelligent, autonomous agents operating safely inside enterprise systems. That debut proved the vision was possible, building on Salesforce’s existing trust layer of data security, governance, and reliability.
This year, the focus shifted from trust to confidence. Salesforce built on that foundation to ensure that what teams build actually performs as expected with repeatable results. Confidence means AI that behaves predictably, not just safely; it’s the difference between a compelling demo and dependable production use.
That evolution—from trust to confidence—defined nearly every update to Agentforce 360 and Data 36O (formerly known as Data Cloud: see Salesforce’s many new product names here). Agentforce 360 builds on last year’s Agentforce debut, extending the platform with deeper architecture, observability, and control to provide more robust options for:
- How teams engage: Agentforce Voice is now GA, and conversational agents are expanding into chat and collaboration environments.
- How teams build: The new Agentforce Builder enables business users to create agents through natural language (Vibe Coding) and Agent Script, giving developers direct access to the underlying logic.
- How teams pay: New Flex Credit models adapt to how organizations actually consume AI.
Together, these updates point to a single goal: AI that meets organizations where they are—ready for real work, not just proofs of concept.
Engaging with Confidence: Voice, Slack, and the Agentic Interface
Salesforce knows adoption depends on how naturally people can engage with AI inside their existing workflows.
Agentforce Voice, now part of the freemium foundation, gives teams a safe, real-world way to build trust in conversational AI. They can design and test agents in live workflows—seeing how logic holds up under real dialogue and unpredictable phrasing—without risking customer experience.
That access extends to Slack, which Salesforce is positioning as the new agentic command center. Agents now live where people already collaborate—inside Slack. That matters because adoption no longer requires behavior change. Teams don’t have to learn a new interface to benefit from AI; they can use it in the same conversations they’re already having.
Building with Clarity: Agentforce Builder, Vibe Coding, and Agent Script
The next source of confidence is clarity in how AI is built. Agentforce 360 introduces the redesigned Agentforce Builder—a unified workspace where business users can design and refine agents directly, while developers maintain full access to configuration and logic. It’s Salesforce turning what used to be experimentation into engineering.
A Unified Workspace for Builders and Developers
The new Builder brings natural language into the development process. Users can describe what they want an agent to do in plain language, and the system translates those descriptions into structured logic, data connections, and actions. It’s fast, intuitive, and accessible—lowering the barrier for non-developers while keeping outputs transparent and auditable.
The Builder supports multiple editing views—natural-language, low-code canvas, and script—and allows for real-time simulation and debugging. That dual design matters because it bridges a long-standing gap between business speed and technical control. Business users can prototype quickly, while developers can refine with precision—all in one environment. It’s how AI development becomes sustainable inside complex organizations.
Extending Control with Agent Script
For developers, Agent Script extends that control. It introduces a human-readable scripting language that defines how agents behave, combining conditional logic, structured rules, and deterministic hand-offs through the hybrid reasoning engine. The result is a development model that supports both speed and precision—business users can move quickly, while engineers can refine with full control.
Reliability Through Intelligent Context and Hybrid Reasoning
Underneath, Intelligent Context and the hybrid reasoning engine give that foundation its reliability. Intelligent Context preserves the structure and meaning in complex data—including documents, tables, and images—so agents interpret questions correctly. The hybrid reasoning engine balances what should be handled through fixed, programmatic logic with what benefits from AI’s adaptive reasoning. That balance limits variability, improves consistency, and helps teams predict how an agent will behave in production.
Together, Intelligent Context and the hybrid reasoning engine make agent behavior explainable. They help teams understand why an AI made a choice—and ensure it makes that same choice the next time. That’s the foundation of confidence.
Migrating to the New Builder
In the near future, Salesforce will encourage teams to migrate existing agents into the new Builder. While Salesforce may provide a migration tool, Coastal sees this as an opportunity to revisit early designs and strengthen governance and observability. Think of it as a “spring cleaning” for your new and evolving agents. Confidence here comes from transparency: knowing exactly how an agent thinks, decides, and improves.
AI Agent Economics: Flex Credits
Salesforce’s new Flex Credit model extends the same flexibility to how organizations pay for AI. It replaces rigid product licensing with fungible credits that can be applied across Agentforce 360, with Data 36O and other consumption-based services projected to be included later this coming year.
That gives teams the freedom to innovate without fear of overconsumption. They can align spend to actual value instead of rigid license tiers, scaling intelligently as their AI footprint grows.
But flexibility still needs discipline. At Coastal, we emphasize consumption efficiency as a design principle. True confidence includes understanding where value is created, how resources are consumed, and how architecture influences cost. Salesforce has made flexibility possible; the work now is using it wisely.
Data 36O: The Backbone of AI Confidence
If Agentforce 360 is the engine, Data 36O is the fuel. Rebranded from Data Cloud, it now serves as the unified data backbone for the entire Salesforce ecosystem—connecting structured and unstructured sources into one governed foundation. That consistency gives every agent and app the same context—a single version of truth that makes AI performance measurable and repeatable.
Integrate Once, Model Once, Activate Many Times
Integrate once, model once, activate many times is a strategy that will make AI sustainable. By keeping models consistent across departments, Salesforce eliminates the fragmentation that drives bad decisions and bloated costs.
Data 36O’s Intelligent Context capabilities preserve relationships in unstructured data and deliver that nuance directly to agents. The result is more precise reasoning and fewer mismatches between data and intent.
AI doesn’t need perfect data—it needs consistent access and governance. Data 36O is how Salesforce delivers both.
What This Means for Enterprise Teams
This year’s updates mark Salesforce’s evolution from vision to operation. Agentforce 360 and Data 36O make AI more predictable, explainable, and observable—ready for enterprise scale.
Salesforce’s work this year also acknowledges a deeper truth: building AI agents is still hard. By investing in architecture, observability, and usability, Salesforce is applying what it’s learned from its own experience and from thousands of customers to make that work easier and more reliable.
Teams can now build AI-first processes however they want. Backend automation, customer-facing experiences, and internal employee agents all run on the same platform—whether through prompts in a Flow, headless agents called by API, or assistants embedded in collaboration tools.
Salesforce started with a focused vision for agentic AI and is now positioning itself as the platform for custom business AI, making it possible to build with precision without needing specialized machine learning developers or prompt engineers.
At the same time, Salesforce isn’t letting the excitement of AI overshadow the fundamentals. Data 36O reinforces access and consistency not just for people, but for the agents themselves—ensuring every system draws from the same governed context.
Salesforce calls this the Agentic Enterprise—a model where human teams and intelligent agents work together in governed, observable systems. It’s not a destination; it’s an operating state that evolves as teams mature.
How Coastal Helps
At Coastal, we’ve been part of this evolution from the start. We serve on numerous Salesforce Partner Advisory boards, including Data 36O, giving us direct insight into how Salesforce is adapting its platform for real-world complexity.
That access matters—but what matters more is how we use it. We help clients translate Salesforce’s architectural progress into operational performance, connecting data, process, and governance so AI can deliver meaningful outcomes, not just functionality.
Salesforce has made AI easier to consume. But ease of use isn’t the same as readiness. Our work starts with what makes AI achieve results: strong data foundations, clear design choices, and aligned architecture that balances flexibility with control.
Through our complimentary Data Strategy Lab, we help teams identify high-value use cases, shape the supporting data models, and map the next steps in building the frameworks that turn AI experimentation into dependable performance at scale.
Salesforce has built Agentforce 360 for confidence. Coastal helps you operationalize it.
