7.5 AI in the Payments Process: the Payments Agent, Discount Intelligence, Virtual Cards and Fraud Controls
Key Takeaways
The Payments Agent has three assistants: Payment Options (compare programs), Supplier Offers (dynamic discount offer lifecycle) and Payment Execution (schedule and monitor payment process requests).
For dynamic discounting, the agent weighs the early-payment discount against the opportunity cost of paying early; for virtual cards, it weighs rebate potential and extended payment float.
Accepted supplier offers update the related installments, so later payment runs pay on the agreed terms through the normal approvals and controls.
The 26B Insights view lists exceptions such as spoiled payment documents, files pending acknowledgment for more than two days, and payments rejected by acknowledgment failure.
Oracle AI Apps predicts discount opportunities from payment terms, due dates, invoice amounts and annual percentage rates, and produces supplier scores from risk and performance metrics.
Oracle's exam page lists Explain the role of AI in the Payments process as an objective, and the credential overview mentions generative and agentic AI. AI in payments matters most in two areas:
- Working capital: when and how to pay.
- Execution oversight: spotting failures before suppliers complain.
The controls you configured in chapter 6 still apply. AI suggests and automates within them.
The Payments Agent
The Payments Agent for Payment Options, Offers, and Execution was introduced in release 26A and extended in 26B and 26D. Oracle's readiness notes say it changes payments from a focus on just paying suppliers to using the outstanding payment information to evaluate payment options and benefits in light of working capital goals. It is built on the AI Agent Studio for Fusion Applications and works through conversation and guided pages. Once it is enabled in 26A, the existing Payables > Payments link opens the Payments Agent.
It has three assistants:
| Assistant | What it does | Examples of natural-language requests |
|---|---|---|
| Payment Options | Reviews upcoming installments (due in the coming weeks) and compares programs: dynamic discounting, virtual credit cards and supply chain financing. It explains the economic impact and creates draft dynamic discount offers on the user's confirmation. | "Compare the benefit of paying Supplier X early with a discount versus a virtual card" |
| Supplier Offers | Runs the dynamic discount offer lifecycle: draft, review, send to suppliers, track responses, send reminders. When an offer is accepted, the related installments are updated so later payment runs pay on the agreed terms. | "Show expired offers for Supplier X", "Send a reminder for this offer" |
| Payment Execution | Guides scheduling and submitting payment process requests (including with templates), shows the status of payments and payment files, and highlights exceptions. It links to the standard pages for action, while keeping existing validations, approvals and controls. | "What is the status of yesterday's payment file?" |
How the agent compares programs
- Dynamic discounting: the early-payment discount against the opportunity cost of paying cash earlier.
- Virtual cards: the rebate and extended payment float, based on financing duration and program terms.
The results are shown side by side so treasury can choose according to cash flow and working capital goals. Oracle frames the agent as moving AP from "operational specialist" to "governor": the agent handles routine tactical work, and specialists manage by exception.
Insights (26B)
The Insights view gives a prioritized, actionable list. In 26B it covers:
- payment documents spoiled in the last 30 days;
- payment files terminated, pending action, or partially acknowledged because of acknowledgment failures;
- payment files pending acknowledgment for more than two days;
- payments rejected because of acknowledgment failure;
- the most frequent payment rejection errors in the last 30 days.
From each insight, users go straight to the assistant or page that resolves it. These insights match the processes in section 7.2: spoiled checks on Record Print Status, and ISO 20022 acknowledgments with automatic voids. That's why the processes matter even when an agent watches them.
Oracle's roadmap includes Standing Offers (suppliers enrolling permanently in optimization programs), virtual card offer lifecycles, email-based supplier outreach without Supplier Portal accounts, and a side-by-side canvas layout.
Oracle AI Apps: predicted discounts and supplier scores
Before the agent, Oracle offered predictive AI for Payables through Oracle AI Apps, a separate application that ingests large data volumes and uses decision science to predict outcomes.
- Discounts: AI Apps calculates discount opportunities, its predicted outcomes, from payment terms, due dates, invoice amounts and annual percentage rates.
- Supplier scores: AI Apps produces a supplier score, an aggregation of a series of risk and performance metrics. Oracle distinguishes it from a credit score, which analyzes a person's credit file and lending history.
Virtual card payments
Oracle's embedded virtual card program pays supplier invoices with single-use card numbers through participating banks. The 26B update added Banco Bradesco, Emirates NBD, NatWest and RBS. Earlier issuers include Barclays, Brex, First Abu Dhabi, HSBC, J.P. Morgan, Wells Fargo and Westpac. Benefits Oracle lists:
- no need to store suppliers' bank details;
- better buyer cash flow through card credit;
- less check processing;
- acknowledgment notifications in ERP.
Setup needs the Financial Application Administrator role. Oracle recommends an end-to-end "penny test" with the bank.
AI-assisted controls around payments
- Supplier bank account validation (Redwood, 25C). Bank accounts entered at supplier registration are checked against J.P. Morgan's Account Validation Service for account validity and ownership. The result (Verified, Failed or Inconclusive) is shown to approvers to reduce payment fraud. It doesn't block payment (section 2.2).
- Payables Agent anomaly detection (section 5.1) catches duplicate or unusual invoices before they reach a payment run.
- Predictive Cash Forecasting (Treasury, 25C) uses predictive AI on cash positions. That is relevant when deciding whether early payment programs are affordable.
Exam angle
AI questions usually test what the feature does and where its limits are:
- The Payments Agent evaluates programs and helps execute payments, but payment approval rules and validations still apply.
- Accepted offers change installments, not the payment terms setup.
- Insights point to acknowledgment and printing exceptions that you resolve with the standard processes in section 7.2.
A supplier accepts a dynamic discount offer sent through the Payments Agent's Supplier Offers Assistant. What happens next?
The supplier's payment terms in the reference data set are changed for all future invoices
The related installments are updated so later payment processing pays on the accepted terms
A Quick payment is created immediately, bypassing payment approval
The offer is sent to the invoice approval workflow for supervisor approval
When the Payments Agent compares paying early under dynamic discounting with paying by virtual card, which factors does it weigh?
Supplier credit scores, the invoice approval chain and each supplier's payment history
The bank account's GL cash account and the reconciliation tolerances
The payment process profile's grouping rules and the payment document numbers
The discount versus the cost of paying early, and the card rebate plus payment float
Which situation does the 26B Payments Agent Insights view flag for action?
Payment files pending bank acknowledgment for more than two days
Supplier invoices waiting for a manager's invoice approval
Expense reports with missing receipts
Purchase orders not yet received
Sections you finish are checked off in the contents.