Smart Workplace Intelligence data first
Answers first from a dated Smart Workplace Intelligence evidence batch verified against its manifest, and reads the tenant live only when fresher data is needed.
Coming soon
An AI assistant for Microsoft 365 workplace transformation project teams: answers from project documentation with citations, and read-only diagnostics of a test tenant for troubleshooting and migration wave pre-checks.
Status
Smart Workplace Agent is not available yet. Its source code, packages, and documentation are not published, and no release date is announced. This page describes the intended scope and may change before a first release.
The captures above come from a desktop prototype running on fictional Contoso data and documents. Scripted demo answers do not come from a language model, and a model can misread a figure: check every figure against the cited sources.
The agent is designed for project teams first: architects, migration leads, and support engineers who need answers from their own documents and quick read-only checks during identity, messaging, collaboration, and modern workplace projects.
Answers first from a dated Smart Workplace Intelligence evidence batch verified against its manifest, and reads the tenant live only when fresher data is needed.
Indexes project documents such as designs, runbooks, decisions, and lessons learned in Word, PDF, PowerPoint, Excel, and Markdown, and cites the source document and section in every answer.
Reads users, licenses, groups, directory synchronization, Teams, SharePoint sites, and OneDrive through Microsoft Graph with an application certificate and read-only permissions.
Reads Entra devices, Intune compliance and policies, Autopilot, Windows Update, and Windows 11 readiness. Device secrets such as BitLocker recovery keys and LAPS passwords are never requested.
Runs reviewed read-only PowerShell scripts for mailboxes, permissions, forwarding, and on-premises Active Directory, and correlates hybrid identities with deterministic rules.
Guardrails
No tool changes any tenant object. The agent refuses to start against a tenant that is not explicitly allowed, and production environments are refused. A future write pattern with explicit human approval, dry run, and audit is designed but not implemented.
EU data residency is enforced in code: a model provider that does not guarantee processing in the EU is refused unless an explicit, audited override is set. The model provider is a configuration choice, and a local model can be used for offline work.
Every tool call is written to an audit trail with pseudonymized identities. Questions, answers, and tool results are not written to the audit trail, and secrets are never logged.
Relationship
The agent reuses the public WorkplaceCloudHub PowerShell toolkit as a source of tools instead of copying it, and converts its outputs into structured results. The toolkit itself stays available on GitHub today.
An answer from the agent is evidence to review, not a verdict: a read-only check reflects what the tenant reports at that moment and does not replace a validated migration or change process.