AI Agents β overview
OwnExpense is part of the Own360 System of Action, so you can describe an outcome in plain English and have AI carry it out as real actions in the app β not just chat about it. A central runtime, OwnAgents, accepts a task, pins it to the apps it is allowed to touch, and a supervisor routes the expense parts of the work to OwnExpense's headless surface (the same create/read/approve operations the UI uses). Every step reasons through OwnIQ, Own360's sovereign AI gateway, which keeps inference inside your perimeter, runs bound by your permissions, and writes an audit log of what was decided and done. OwnExpense also ships its own in-app AI you use directly: OwnScan (receipt capture and OCR), the PolicyAgent reasoning shown on capture and on every expense, and OwnIQ-backed analytics and CFO insights. This section explains what the agents can do, how to use them, and how to verify the results.

Step-by-step
- 1Think in terms of outcomes: an agent task is a sentence describing what you want done (for example, "Submit an expense report 'Customer trip BLR' totalling $1,250 with three itemsβ¦"), not a sequence of clicks.
- 2When dispatched through Own360, the OwnAgents runtime pins the task to the allowed apps and the supervisor routes the expense work to OwnExpense's headless surface β the same operations (create report, add line items, submit, approve) that the screens use.
- 3The agent reasons through OwnIQ, which stays in-perimeter, runs within your permissions, and logs every decision and action for audit.
- 4To use AI yourself inside the app, open New expense (OwnScan + PolicyAgent), any expense's detail (PolicyAgent reasoning + FraudAgent), or Analytics (Ask OwnIQ) β covered in the topics below.
- 5Always confirm the result on the real screens (the created report, its line items, and the policy/audit trail) before treating a task as done β see "Verify agent results".
Tips & pitfalls
- β’Agents act through the same permission-checked operations you do β they cannot do anything your own account is not allowed to do.
- β’If OwnIQ is not configured for a deployment, live model answers degrade gracefully (Analytics shows an honest "OwnIQ is not configured" state and offline answers); the deterministic PolicyAgent and the curated surfaces still work so you are never shown fabricated AI output.
















