01
Describe the outcome
Ask to view work, submit your records, or handle an approval using a person, date, project, or task.
Permission-aware AI
Ask naturally. Act with control.
HourOps turns conversational requests into controlled workforce actions. Your role, workspace, reporting hierarchy, record state, and confirmation all remain part of the decision.
Scope aware
Every intent
Tenant aware
Every query
Auditable
Every outcome
HourOps AI Assistant
Protected workspace session
Guardrails active
Show pending timesheets for Maya this week.
I found one eligible pending timesheet.
Employee
Maya Chen
Week
21 Sep
Hours
40.0
Would you like to approve or reject it?
Approve Reject
Approve it.
Approved and recorded
Maya’s pending timesheet was approved. The AI audit entry records the action and its channel.
Ask about work, leave, timesheets, or projects...
How to use it
Conversation without surrendering control
Ask for a view or action in ordinary language. HourOps resolves the request, applies the same operational rules as the product, and pauses whenever your intent needs confirmation.
02
Review the safe interpretation
The assistant narrows ambiguous details and shows confirmation choices before sensitive or broad actions.
03
Receive a recorded result
Authorized actions update HourOps records and appear in the relevant audit trail and prompt history.
Security decision path
The model does not decide your permissions.
Natural language helps identify the requested intent. HourOps remains responsible for identity, tenant selection, scope validation, approval authority, confirmation, record updates, and audit history.
Signed identity
Authenticated user
Tenant context
Active workspace
Scope check
Allowed intent
Authority check
Hierarchy and record
Confirm and audit
Controlled outcome
Explicit AI access
The user must have the Use AI Assistant permission before a request is processed.
Allowlisted actions
A request becomes a supported intent, then that intent is checked against the user’s live scopes.
Tenant-contained data
Queries run inside the active workspace context, preserving account boundaries and database isolation.
Approval authority
Reporting hierarchy and workspace-wide approval authority are validated before changing a request.
Clarification first
Ambiguous names, mixed instructions, and unclear dates are stopped until the user chooses precisely.
Recorded outcomes
AI-originated approvals and rejections are written to operational audit history for review.
Approval and rejection
One conversation. Two controlled outcomes.
Approval is never inferred from a vague request. Rejection is also an explicit status change. Both paths operate only on eligible pending records.
Approve eligible work
Request
Approve Maya’s pending timesheet for the week of 21 September.
Guardrail
HourOps verifies scope, reporting authority, employee identity, date range, and pending status.
Recorded outcome
The eligible record becomes Approved and the AI channel is recorded in the audit history.
Return work for correction
Request
Reject Daniel’s pending timesheet for this week.
Guardrail
If the request covers multiple tasks or weeks, the assistant asks which records should be rejected before continuing.
Recorded outcome
The selected pending record becomes Rejected, so the employee can correct and resubmit it.
Hard boundaries
Some actions stay outside the assistant.
Critical structural and administrative changes remain in the standard HourOps interface. A prompt cannot grant itself more authority.
Create, edit, or delete users
Change organization settings
Create or modify roles
Change subscriptions or billing
Rewrite holidays or leave policies
Approve the requester’s own records
Move work forward without skipping the rules.
Use conversational workflows for everyday operations while HourOps keeps permissions, tenant boundaries, confirmations, and auditability in the path.