How agent-ready is Xero?

Xero's overall agentability score of 48/100 indicates moderate readiness for autonomous AI agents. The platform performs consistently across its homepage, pricing page, and documentation, with all three sections scoring in the 42–48 range. This suggests systemic patterns rather than isolated page-level issues.

The platform demonstrates strong machine readability (81/100) and solid control mechanisms (73/100), meaning agents can generally parse the HTML structure and navigate key workflows. However, Xero faces significant challenges in shadow UI avoidance (15/100) and transparency (18/100), indicating that important functionality may be hidden in client-side JavaScript or presented without adequate context for agents to interpret reliably.

Mid-range scores in defaults (33/100), clean handoffs (50/100), and chunking (53/100) point to opportunities for incremental improvement in form usability, workflow transitions, and content organization. Status visibility (65/100) is adequate but could be enhanced to better support agent-driven interactions.

To improve agentability, Xero should prioritize the following changes:

• Replace generic <div> and <span> elements with semantic HTML5 landmarks such as <header>, <nav>, <main>, <article>, <section>, and <footer> to strengthen structural machine readability.
• Add source references or input citations to generated outputs, linking back to the underlying data or configuration that produced each result.
• Pre-fill form inputs with sensible defaults wherever a reasonable value exists, reducing the decision burden for agents and improving form completion rates.
• Surface confidence or certainty indicators on key outputs, such as adding confidence fields to API responses or displaying 'verified vs. best-guess' labels in the user interface.
• Restructure content headings to directly answer user questions (e.g., 'How do I cancel?' instead of 'Cancellation'), and ensure each section is self-contained and understandable without surrounding context.
• Gate destructive actions behind explicit, distinctly-labelled confirmation steps rather than relying on same-styled buttons that agents may not recognize as critical checkpoints.