01What you answer for
| You answer for | How an AI agent puts it at risk |
|---|---|
| Payments | Fake or inflated invoices paid |
| Fraud | Bank details changed on a supplier’s say-so |
| Budget | Purchases agreed past the budget |
| Supplier relationships | Genuine invoices refused or delayed |
02Questions to ask your team
- Can the agent pay anything that does not match an approved purchase?
- Is the budget enforced by the payment system or only by the prompt?
- What does it do when a supplier says its bank details changed?
- How often did it pay too much in testing, and by how much?
- Are its payments reconciled, and by whom?
03Evidence to ask for
- Fake, inflated and over-budget payments in testing: rate, range and cost.
- Genuine invoices it failed to pay.
- The supplier tactics that worked.
- Payment logs that tie each payment to its conversation.
04Where to go next
- AI purchasing and invoice agents: what a purchasing test checks
- Finance and accounts payable playbook: the threats and the rules
- The enterprise guide: the readiness checklist for your review
Questions
Does an agent change our audit obligations?
The obligations are yours either way. An agent adds a new party that can be persuaded, so the controls should be enforced by systems, tested, and its payments logged with the conversation behind each one. Ask your auditors how they want that evidenced.
See where your agent leaks money. Tested against your own rules, with the money at stake.