Product · Platform capability
Spend controls
AI spend behaves like any other budget line: allocated in advance, observable per task, and enforced at the boundary you set.
Procurement · March
department budget, enforced per task
SAR 8,120
of SAR 12,000
- Tender responsesSAR 4,860
- Quotation comparisonsSAR 2,340
- Supplier onboardingSAR 920
- Alert at 80%ARMED
Why it matters
Agent work is metered work, so cost deserves the same governance as access and data. Spend controls make it a property of every mission, set before the work begins and visible while it runs.
Three levels of control
Per workflow
Each workflow carries a cost ceiling per run. A mission that would exceed it pauses and asks its owner.
Per team
Teams hold monthly allocations with alert thresholds. Usage is visible task by task, not as one opaque bill.
Per department
Departments set hard limits that no workflow can cross. Finance sees spend in the same ledger as everything else.
How enforcement works.
Limits are checked before work runs, not reported after. The router treats cost as a first-class constraint.
01
Budgets set up front
Allocations and ceilings are declared per workflow, team and department before anything runs.
02
Cost-aware routing
Model routing weighs cost alongside accuracy and privacy, so expensive models run only where they earn it.
03
Stops at the boundary
A task that would cross a limit pauses, alerts its owner and waits for a decision before continuing.
04
Observable per task
Every task carries its cost in the audit log, so finance can attribute spend to workflows, teams and outcomes.
Cost is one governed property.
Permissions, masking and audit apply to the same missions, from the same plane.
Governance & audit →