Salesforce Agentic Enterprise Design Principles

A framework for implementing AI agent architectures in enterprise environments, emphasising modularity, data harmonisation, observability, trust, and scalable infrastructure.

  1. Design for modularity

    Build reusable, composable components that can be reconfigured for different agent functions without rebuilding entire architecture elements from scratch.

  2. Harmonise data with metadata-driven understanding

    Provide agents with unified data and metadata, business glossaries, and ontologies so they can understand context and make informed decisions across systems.

  3. Enable unified observability

    Maintain real-time visibility into agent actions, reasoning, context, governance, and business outcomes across both IT and business functions.

  4. Build with trust

    Embed governance and trust principles throughout the architecture, including agent identities, task-based permissions, and policy compliance checks.

  5. Design for strategic human intervention and oversight

    Implement appropriate human-in-the-loop engagement at decision points, with smooth handoffs that preserve context when escalation is needed.

  6. Enable event-driven processing

    Support real-time decision-making through always-on agents that respond to triggers across any channel, including texts, emails, calls, and APIs.

  7. Ensure infrastructure can scale growing AI workloads

    Build resilient infrastructure with compute that scales based on workload needs, distributed loads, and storage that handles unpredictable data access patterns.

  8. Prioritise open ecosystems and standards

    Design for interoperability using standard interfaces, open protocols, and portable workflow definitions to avoid vendor lock-in and enable flexibility.


Tags: Organisations, Software, Infrastructure, Specific