Understanding the analysis
Safety score
A safety score in the Analysis View is a generated indicator about the transcript, not a survey of an employee or a measured property of a workplace.
What appears in the product #
The conversation analysis can include a safety_score and related evidence, defensive-behaviour information, and a qualitative confidence label. The interface displays the number when it can parse it and may show an evidence phrase from the analysis. A separate phrasing review can assess HR wording and offer alternatives.
These values originate in language-model output based on the transcript. They are not collected through a validated safety questionnaire and are not a clinical or organizational assessment.
What changes the result #
The transcript's accuracy, completeness, speaker attribution, language, cultural context, and the specific topic discussed may all affect what the model returns. A single conversation cannot represent every interaction or a person's experience over time.
The interface may show an unavailable state when the relevant result is missing. Do not interpret an absent score as either safe or unsafe, and do not convert the number to a probability without an independently validated basis.
Discuss the underlying experience #
Check any cited phrase and ask the employee open questions about how the conversation felt. A score can help you decide what to explore, but the employee's account and the surrounding facts matter more than the display.
Use care with sensitive information and avoid making adverse decisions from this indicator alone.
Use this guidance thoughtfully
This page describes the current product workflow; it is not an HR policy or a promise about every deployment. Before acting on information about an employee, verify the source record, consider context the system cannot see, and use your organization's process. If a control behaves differently from this guide, check the current workspace UI and confirm the saved result rather than assuming an action completed.