Answers with sources
Questions about working time, remote work, development, private device use, or workplace services are answered from curated SharePoint content.
Governed People Operations · Microsoft 365
Clear guidance. Safe handover.
The solution connects employee self-service, manager support, and structured intake with roles, audit, and human approval. It is not an autonomous HR decision-maker, but a controlled service layer for Microsoft 365.
Copilot StudioSharePoint backendHuman in the loop
People Ops Intake CockpitReview Queue
Workplace · W004Parking request · new
Feedback · AFB2anonymised · review
Manager check-indraft · approval
Policy Q&Aanswered · source
Controlled handling
Review workplace request
The agent prepared the category, summary, and next step. A human takes ownership and decides on the action.
Role and permission checkedSource and policy linkedAudit record preparedHuman approval pending
The purpose
Many People Ops questions repeat. Accountability remains human.
Employees need fast, reliable guidance. Managers need sound preparation. HR and Workplace Services need structured cases rather than incomplete free-text messages. The agent takes on the repeatable work—not the sensitive decision.
Prepare managers better
Structures and drafts for one-to-ones, goals, probation, or performance conversations—with a clear line between preparation and personnel decisions.
Route requests cleanly
Workplace tickets, sensitive HR requests, and anonymous feedback are categorised, minimised, stored traceably, and passed to the right role.
Interactive case flow
From request to controlled action.
Select a stage. The flow shows where knowledge, safety checks, the backend, and a human decision work together.
01Request & roleEmployee · manager
02Knowledge & policySharePoint · rules
03Safety checkPrivacy · limits
04Intake & auditLists · flows
05Human reviewApproval · handoff
Understand request and role
The agent recognises the mode, topic, and required context. It only gathers information that the selected process genuinely needs.
Six controlled working modes
One agent. Different roles, data, and outcomes.
Policy Q&A
Answer employee questions from curated policies, including source and escalation when the answer is unclear.
Manager check-in
Prepare conversation structure, goals, and notes—without performance judgement or hidden assessment.
Sensitive request
Capture only necessary facts, mark sensitivity, and hand the case to HR under controlled conditions.
Feedback & speak-up
Minimise content, assess re-identification risk, and promise only the anonymity the process can actually provide.
Workplace Services
Capture requests about parking, catering, charging points, or equipment as traceable work items.
Goals & development
Structure drafts and check-ins while keeping ownership and approval with people.
Microsoft-native architecture
Knowledge, identity, process, and governance on one platform.
The demo uses a dedicated SharePoint site, curated knowledge, operational lists, role models, and a Copilot Studio agent. A separate intake cockpit exposes generated records to operators.
SharePoint
Curated HR, manager, employee, and workplace knowledge rather than uncontrolled web search.
Copilot Studio
A conversational layer with dedicated instructions, knowledge, and controlled action paths.
Lists & Automate
Cases, feedback, policy rules, check-ins, goals, tickets, and audit as structured data.
Entra & Purview
Roles, groups, permissions, privacy, and governance as core design inputs.
Microsoft 365SharePointCopilot StudioPower AutomateEntra IDTeams-readyHuman in the loopAudit
Hard boundaries
No HR automation at any cost.
The more sensitive the context, the more important data minimisation, traceable sources, and real accountability become. The agent may prepare, structure, and route. It may not judge people.
- No personnel decisionsNo recommendations on dismissal, disciplinary action, pay, promotion, or individual suitability.
- No covert monitoringNo Teams sentiment analysis, individual risk scoring, or behavioural profiling.
- No false anonymityAnonymous processes are described only as far as technology, access, and re-identification risk allow.
- No action without ownershipSensitive cases remain in review queues and require an explicitly accountable human role.