How agent-ready is Intercom?

Intercom's overall agentability score of 37–44 across its homepage, pricing, and documentation indicates that autonomous AI agents will encounter significant friction when attempting to parse content, understand system state, and complete workflows. The platform demonstrates strong machine readability (83/100), meaning its HTML structure is well-formed and semantically sound. However, critical gaps in transparency (0/100), defaults (27/100), and shadow UI avoidance (15/100) prevent agents from understanding system confidence, distinguishing authenticated from public features, or reliably interacting with dynamic interface elements.

The low transparency score reflects a complete absence of machine-readable explanations for system outputs—agents cannot determine confidence levels, trace provenance of results, or access audit logs. Similarly, the poor shadow UI avoidance score suggests heavy reliance on client-side rendering or interactive elements that lack accessible labels and semantic hooks. These issues combine to make Intercom's interface difficult for agents to navigate predictably, though the solid machine readability foundation means many of these problems are addressable through incremental improvements to markup and API design.

Intercom can meaningfully improve agentability by addressing the following gaps identified in the audit:

• Add confidence or certainty metadata to API responses and label UI outputs as 'verified' versus 'best-guess' so agents can assess reliability of information.
• Expose an activity log or audit trail in both human-readable (visible feed) and machine-readable (JSON event log) formats, and provide 'why this result' affordances that explain system decisions.
• Clearly mark which features and content are accessible without authentication versus those requiring sign-in, enabling agents to plan workflows appropriately.
• Provide labeled pause, cancel, or stop controls for long-running operations with accessible names like 'Cancel import' instead of unlabeled icons.
• Wrap status messages and updates in ARIA live regions (role=status or role=alert) so agents can monitor state changes programmatically.
• Include source references, input citations, or provenance links on generated outputs, and structure important results as a one-line summary with an expandable drill-down for details.