The increasing integration of artificial intelligence and automation into business processes, particularly within Salesforce environments, necessitates a re-evaluation of traditional cybersecurity and governance practices. Existing security frameworks, which typically focus on identities, permissions, access controls, and configurations, are insufficient to address the complexities introduced by AI agents, APIs, and interconnected systems.
A new framework, called Trust Mapping, has been proposed to help organizations understand and manage the evolving security landscape. This framework defines "trust" as the belief that individuals, systems, information, and connected services will operate as expected within a business workflow. It recognizes that these trust relationships can extend across various entities, including human users, AI agents, APIs, and external platforms.
Each trust relationship has a defined responsibility, scope, and boundary, outlining what is being relied upon, where that trust applies, and its limitations. These relationships are also underpinned by assumptions about the conditions that enable the reliance. As AI and automation take on more tasks without direct human intervention, these trust relationships can span multiple connected tools and processes.
The Trust Mapping Framework examines trust across five key domains: entities (who or what participates), information (what data is used), connections (how trust is established or extended), actions (how trust is exercised), and system outcomes (what results the workflow produces). This framework is applicable to various Salesforce processes, including Agentforce, Headless 360, third-party SaaS applications, and AI-assisted workflows.
The framework proposes a two-stage process: Discovery and Governance. Trust Mapping Discovery aims to identify the specific trust relationships that enable a business workflow to function. This involves defining their responsibilities, scope, boundaries, and supporting assumptions. For instance, scenarios might include a salesperson using an AI model like Claude through Headless 360 to analyze Salesforce data, a customer support request handled by Agentforce and reviewed by a human, or a discontinued Salesforce integration with active credentials.
Following Discovery, the Governance stage assesses whether these identified trust relationships remain appropriate, justified, and aligned with business intent, and if they introduce risks requiring mitigation. This assessment considers the five domains of the framework, focusing on visibility, ownership, purpose, monitoring, and review. Organizations evaluate if the right people and systems have appropriate authority, if information remains reliable, and if connections and actions stay within their intended limits. They also examine workflow outcomes and their potential impact on other processes.
Based on this assessment, a trust relationship can be maintained, modified, restricted, or removed, and additional controls or increased monitoring can be implemented. Governance relies on evidence such as changes in permissions or OAuth scopes, security incidents, audit findings, threat intelligence, AI behavior and model evaluations, business process changes, and operational data. The level of review is proportional to the workflow's complexity, criticality, regulatory requirements, connectivity, and autonomy.
A key concept is "trust drift," which occurs when a trust relationship deviates from its original purpose, scope, limits, or supporting assumptions. Examples include unused but active credentials, excessive access, outdated information, and unvalidated AI-generated recommendations.
Trust Mapping is an ongoing process, as relationships can change with the introduction of new integrations and AI agents, vendor replacements, employee role changes, project retirements, or shifts in information reliance. Therefore, Discovery and Governance may need to be repeated. This approach complements existing security practices like security posture management, threat modeling, and identity governance by adding a workflow-level perspective on dependencies, their rationale, boundaries, and underlying assumptions.
