Understanding the analysis
What the AI can and cannot tell you
VooHr's generated analysis is decision support, not a diagnosis, investigation, or source of verified facts. It can help organize a conversation, but it can also misunderstand it.
One conversation is limited evidence #
The analysis is based on a single transcript selected for the request. It does not independently interview the employee, verify events, observe workplace conditions, or know what was left unsaid. A score or pattern may reflect only this conversation and may change with different text or language.
Transcripts above 50,000 characters are truncated before they are sent for analysis. The full saved transcript remains separate, but the model does not receive text beyond that request limit. Review whether important context falls near or after the cutoff.
No diagnosis or clinical label #
VooHr does not provide a clinical diagnosis through this analysis. Generated labels are not a mental-health assessment, personality assessment, or professionally validated instrument. A risk indicator is not a statement about what an employee will do.
Even if a generated paragraph sounds certain, its claims remain probabilistic. The model can produce plausible but incorrect interpretations, unsupported causes, or suggestions that do not fit your workplace.
Human judgement stays with HR #
Verify statements against the transcript and reliable context, ask the employee for their perspective, and follow your organization's policies and applicable requirements. Do not base a consequential employment decision on a model output or one conversation score alone.
When no evidence is available, 'No data' is intentional: the interface is not substituting a zero or claiming that no concern exists. Treat missingness as a reason to check inputs, not draw a conclusion.
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.