AI Use Policy
How Workplace.io uses automated analysis to generate culture insights, how we limit exposure of sensitive data, and the guardrails
we apply to reduce misuse. This page intentionally excludes proprietary model and taxonomy details.
Last updated: December 16, 2025
At a glance
- Purpose-limited: automated analysis is used to generate organizational insights and trends.
- Short-lived content: message content is processed briefly for analysis and then removed.
- Derived metadata retained: structured results and aggregates are retained to power reporting.
- Guardrails: role-based access and acceptable-use restrictions reduce surveillance risk.
What we mean by “AI”
Workplace.io uses automated classification and scoring techniques to map collaboration signals into structured, research-informed metrics.
We do not disclose proprietary model architecture, training data, or internal signal taxonomy in public documentation.
What automated analysis is used for
Automated analysis supports product features such as metric trends, aggregate summaries, and structured signal outputs used for reporting.
It is designed for organization-level insight rather than individual monitoring.
- Generate aggregate metrics and trends at the workspace and group level.
- Produce structured outputs used to power dashboards and comparisons.
- Support operational quality checks and consistency of reporting over time.
Inputs and outputs
Automated analysis may temporarily process message content and related metadata needed for analysis. Outputs are structured results and
aggregates used for reporting.
- Inputs (short-lived): message content and limited metadata required for analysis.
- Outputs (retained): derived metadata (structured results) and aggregated reporting data.
Where inference runs
Classification and culture inference workloads run on AWS using Amazon SageMaker, separated from the application serving layer hosted
on DigitalOcean. We restrict access paths and credentials used to operate these workloads.
Interactive chat (optional)
Workplace.io includes an optional interactive chat feature that helps users interpret dashboards and metrics. When used, this feature may
send a user’s prompt and limited dashboard context to a third-party AI provider (currently OpenAI) in order to generate a response.
This chat feature is separate from our culture classification system, which runs on AWS using Amazon SageMaker.
- User-initiated: chat runs only when a user submits a prompt or interacts with chat features.
- Minimized context: we limit what is sent to what is needed to answer the user’s request.
- Sensitive input: we recommend users avoid entering sensitive personal data into free-form prompts.
Guardrails and intended use
Workplace.io is intended to support organizational health insights. We encourage configurations and access patterns that reduce the
risk of individual surveillance and misuse.
- Role-based access: administrative controls are restricted to authorized roles.
- Aggregation: reporting emphasizes group/workspace-level results.
- Customer control: admins choose which sources to connect and can disconnect integrations.
Quality and change management
We manage changes to scoring and outputs to preserve consistency and reliability. We perform evaluation and operational checks appropriate
to the platform’s maturity and customer needs.