Governed AP Exception Workflow
Every invoice here is modeled on Oracle Fusion Cloud ERP's Payables flow — it fails three-way match against a purchase order the way a real Fusion Payables invoice would land on a matching hold. Watch an agent propose a fix, a deterministic policy decide whether it executes alone, escalates to a human, or gets blocked outright — and every decision, including reversals, land in an audit-grade ledger. No chatbot in the decision loop; the policy is under 60 lines and reads like a checklist.
Model spend
$0
of $50 cap
Ledger entries
0
every decision, not just successes
Pending approvals
0
escalated, awaiting a human
Propose an exception resolution
Click a scenario, or fill in the form and pick your own number.
Or propose your own — pick a PO and an amount:
Value ledger — this session
Computed live from the actions above. Baseline time and hourly rate are starting points — edit them to match your own team's numbers, per Tioga's value-report methodology: never an invented industry figure.
0
Actions processed
0
Escalated to human
0
Blocked by policy
0.0 hrs
Net hrs saved
$0
Value this session
Only completed actions (auto-approved or escalated-then-approved) count toward hours/value — escalated and blocked actions are tallied above as evidence the governance layer works, not folded into the total. A rolled-back action's hours are removed once reversed. Net hours saved already subtracts human review time for escalated actions — see the exported report for the gross-vs-net breakdown per row.
Scheduled reconciliation
Compares every agent self-reported status update against the ledger's actual decision — runs on a fixed cadence in production, not only when something looks wrong. Click it any time, whether or not anything here looks suspicious.
Audit ledger
Blocked and escalated actions are logged with the same fidelity as executed ones.
No actions proposed yet — try a scenario above.
Policy: actions under $5,000 execute autonomously; up to $25,000 require human approval; above that, blocked entirely — no override exists at this layer, on purpose. Master-data changes (scenario 6) skip the dollar ladder entirely and are gated on a documented authorized-change record instead — every other layer of agent security can pass, and change control is still what catches it. Scenario 7's claimed-vs-actual gap is invisible until reconciliation runs — that's deliberate, and the point. Everything on this page runs in your browser; nothing is sent to a server.
Want the full picture for your environment?
A discovery call gets you a scoped assessment from the person who builds these engagements — not a form, a conversation.
Book a discovery call