Agentic Enterprise Architecture
6 etapas1 peças5 Fontes
The concept of "Agentic Enterprise Architecture" was introduced by Salesforce at Dreamforce 2026 to address the gap between probabilistic AI reasoning and deterministic business execution. The core argument is that deploying AI agents in actual business workflows fails not due to model intelligence, but because of a missing connection layer that translates judgment into compliant actions governed by existing permissions, approvals, and data structures.
A Deloitte survey cited in the material indicates that only 5% of U.S. companies consider themselves "very ready" for AI agent integration, with 72% struggling to connect scattered data across systems. To resolve this, Salesforce proposed a four-layer architecture: Data 360 (unified data foundation), Customer 360 (business logic and processes), Agentforce (digital workforce executing tasks), and AIPulse (interface connecting humans and agents). This structure ensures AI agents operate within established security policies and approval workflows rather than acting autonomously without guardrails.
Industry adoption shows mixed results. Anthropic uses Salesforce as its CRM while integrating Claude into Salesforce's interface; Siemens reduced lead response times from days to minutes using SiePortal AI; L'Oréal cut average consultation handling time from 7 minutes to 2.5 minutes. However, challenges remain regarding governance, trust, and continuous monitoring. Kathy Baxter, Salesforce’s Principal Architect of Ethical AI Practice, emphasizes that consequential decisions must route through deterministic layers (like Apex or Flow) to ensure auditability and human oversight, warning that pre-launch testing alone is insufficient for probabilistic systems.
Como chegamos aqui
Salesforce introduces 'Agentic Enterprise Architecture' at Dreamforce 2026, defining a four-layer model (Data 360, Customer 360, Agentforce, AIPulse) to bridge probabilistic AI and deterministic business rules.
byline.network ↗