Salesforce Unveils the Foundation for Building Trusted AI

• 03 Oct 2025
• TM Team
• 0 Comments

Delivering accurate and trustworthy AI is non-negotiable for enterprises, but fragmented data and inconsistent governance stand in the way. In fact, more than 80% of AI projects fail to deliver value, often because of poor data quality, weak governance, and fragmented integration, according to a RAND study. To combat this, Salesforce is expanding data, governance, security, and semantics capabilities across the Salesforce Platform — so every AI agent operates with the same accuracy, context, and control. These innovations help companies move from cautious experimentation to transformative action, scaling AI responsibly across every workflow to power the Agentic Enterprise.

The Salesforce Platform provides the essential scaffolding for the Agentic Enterprise — a new model of work where humans and AI agents collaborate seamlessly across every workflow, decision, and interaction. The platform delivers three core capabilities that make this model accurate, explainable, and secure:

• Context and Accuracy — ensuring outputs are grounded in unified business data and knowledge.
• Built-in Trust, Security, and Compliance — embedding visibility, control, and compliance into every workflow.
• Open and Unified — connecting agents, data, and semantics across ecosystems to avoid lock-in and ensure consistency.

What’s New

Building on its commitment to context, trust, and openness, Salesforce today introduced ‌new innovations that enable enterprises to scale reliable agentic AI across every workflow:

• Data Cloud Context Indexing: A new indexing pipeline in Data Cloud helps AI agents interpret unstructured content — like contracts, diagrams, and tables — through a business-aware lens. This surfaces precise details from vast, disconnected information, making it easier to get the right answers faster. For example, a field engineer can upload a schematic and have the agent guide them step-by-step through a troubleshooting decision tree — getting to the right fix in minutes, not hours.
• Data Cloud Clean Rooms: Now generally available, Data Cloud Clean Rooms let companies securely share, collaborate, and analyze data to unlock intelligent insights without moving or exposing raw data. The offering extends the power of a typical clean room by leveraging zero copy connectivity, eliminating the need to duplicate sensitive datasets — reducing security and compliance risks and cutting storage costs. For example, a group of global banks are using Clean Rooms to compare transaction patterns and detect fraud rings in hours instead of weeks — without exposing sensitive customer records to one another. Additionally, companies can securely collaborate with partners through our privacy-enhanced, native integration with AWS Clean Rooms.
• Tableau Semantics: Tableau Semantics is an AI-powered semantic layer, natively integrated with Data Cloud, that translates raw data into business language. To accelerate time to value, Tableau is delivering an out-of-the-box Customer 360 Semantic Data Model (SDM) that unifies cross-cloud data and metadata, simplifying modeling, enforcing consistent metric governance, and providing the business context AI and BI need for accurate, trustworthy insights. To facilitate this, Tableau is co-creating the industry’s first open semantic interchange. We’re also partnering with leaders like Databricks, dbt Labs, and Snowflake to enable standardized semantics across platforms — ensuring analytics and AI remain consistent, explainable, and grounded in shared business definitions.‌ For example, companies that define “ACV” differently across teams can now unify that logic, so every agent and every dashboard uses the same definition of revenue.
• MuleSoft Agent Fabric: As companies accelerate their AI adoption, they’re facing an explosion of agents across teams, platforms, and vendors. This surge introduces a new form of fragmentation — agent sprawl — where disconnected workflows, redundant automations, and compliance blind spots emerge. MuleSoft Agent Fabric gives enterprises a single place to register, orchestrate, and govern every AI agent, regardless of where it was built. For example, this allows a global retailer with an agent for tracking inventory, another for updating prices, and a third for detecting fraud to work together, not operate in silos, to make pricing adjustments automatically and fraud checks happen in real time.
• Embedded AI Security and Compliance: New AI-powered security capabilities will be embedded across the Salesforce Platform, streamlining the maintenance of security and compliance posture. This includes new integrations industry leaders like CrowdStrike and Okta, ensuring that security threat identification and compliance management are proactive and integrated into the agentic enterprise’s operating mesh.
• Informatica Acquisition Unlocks Enterprise Metadata: The planned acquisition of Informatica (subject to close by Q1 FY27) will bring together Informatica’s data catalog, integration, governance, quality, privacy, and Master Data Management (MDM) capabilities with the Salesforce Platform to create a unified data architecture for agentic AI. This will give enterprises the metadata intelligence needed to ensure AI agents operate safely, responsibly, and at scale across the agentic enterprise.

 “At Salesforce, we believe AI is the most powerful enabler of business transformation today. But for it to deliver on its promise, it must be grounded in deep business context to power the true force multipliers: intelligent automation and actionable insights,” said Rahul Auradkar, EVP & GM, Unified Data Services. “That’s why we’re building a unified foundation that harmonizes all of a company’s structured and unstructured data to surface these insights and fuel this automation, all grounded in the governance and security needed to operate with confidence. With these innovations, we’re helping our customers move faster, scale responsibly, and unlock the full potential of the Agentic Enterprise.”