How agent-ready is Freshworks?

Freshworks received an overall agentability score of 19/100 on its homepage, 45/100 on pricing, and 37/100 on documentation. These scores indicate that autonomous AI agents will face significant friction when attempting to parse content, navigate workflows, and operate the platform programmatically. The site currently lacks the structural signals and control patterns that enable reliable agent interaction.

The audit identified critical gaps in transparency (0/100) and defaults (0/100), meaning the site provides no machine-readable activity logs, confidence indicators, or sensible starting states for automated workflows. Control scored 13/100, reflecting missing confirmation patterns and undo mechanisms for destructive actions. Machine readability (40/100), shadow UI avoidance (30/100), and chunking (20/100) show foundational issues with semantic HTML structure and content organization that limit agent comprehension.

Improving agentability will require systematic changes to how Freshworks exposes system state, structures markup, and handles user actions. These improvements would benefit both autonomous agents and assistive technologies used by human visitors.

To improve agent-readiness, Freshworks should prioritize the following changes:

• Add semantic HTML landmarks (<header>, <nav>, <main>, <article>, <section>, <footer>) in place of generic <div> and <span> elements to give agents clear structural cues about page organization.
• Expose an activity log or audit trail in machine-readable format (JSON event log with a visible activity feed UI) so agents can track system actions and understand workflow state.
• Gate destructive actions behind explicit, distinctly-labelled confirmation steps rather than same-styled buttons, and add labelled pause/cancel controls (e.g., 'Cancel import') for long-running operations.
• Implement undo/override capabilities for completed actions, particularly destructive changes, to allow agents and users to recover from errors without full workflow restarts.
• Attach source references or input citations to generated outputs, and surface confidence indicators on key results (via API response fields or UI labels like 'verified vs best-guess').
• Provide one-line summaries with expandable drill-downs for important results, making it easier for agents to decide whether to examine details or proceed to the next step.