Visa has launched a live agentic commerce pilot with financial technology companies BILL and Billtrust, allowing authorized artificial intelligence (AI) agents to initiate and complete approved business-to-business (B2B) invoice payments through supplier portals.
Announced on October 20, 2026, the initiative uses Visa Intelligent Commerce and the Visa Trusted Agent Protocol to help AI agents authenticate themselves, access supplier payment portals and submit payments using tokenized payment credentials. The pilot is already processing end-to-end transactions involving participating businesses that use BILL and suppliers on Billtrust.
The project targets a longstanding problem in corporate payments: even when an invoice has been reviewed and approved, an employee may still need to log in to a supplier’s website and manually enter payment information.
Visa, BILL and Billtrust aim to automate that final step while preserving buyer approval requirements, supplier control and transaction visibility.
Importantly, the AI agents do not independently decide which invoices to pay. They execute eligible transactions that have already been approved by the buyer and remain subject to the buyer’s payment instructions and controls.
How Visa’s AI Agent Payment System Works
The pilot connects buyer-side accounts payable software with supplier-side payment portals, allowing authorized AI agents to complete transactions across separate business systems.
The process begins with BILL’s authorized Payment Agent, which acts on behalf of a business customer. The agent is registered through Visa’s Trusted Agent Protocol and Agentic Directory, allowing its identity and authorization to be recognized.
The agent then uses tokenized payment credentials and transaction controls aligned with the buyer’s approved payment instructions to initiate the invoice payment.
On the supplier side, Billtrust uses Visa’s Trusted Agent Protocol to recognize and authenticate the agent, allowing it to access the relevant payment portal and complete the transaction according to the supplier’s acceptance policy.
This authentication mechanism is designed to address obstacles traditionally created by CAPTCHA checks and other anti-bot protections, which can prevent automated software from submitting payments through websites intended for human users.
The system also avoids the need for businesses to share their supplier-portal login credentials with AI agents.
According to Visa, the approach is designed to provide visibility into who is acting, which payment has been authorized and whether the transaction has been completed.
Why Visa Is Targeting B2B Invoice Payments
Many companies have already automated significant portions of their accounts payable operations, including invoice capture, document processing, approval workflows and payment scheduling.
However, supplier portals can introduce a manual step at the end of the process. Some suppliers require customers to log in to their websites and submit payment information directly, even when the invoice has already passed the buyer’s internal approval process.
That final requirement can create delays, increase administrative work and prevent businesses from fully automating their payment operations.
Visa’s pilot aims to remove this bottleneck by allowing authorized agents to execute approved payments across platforms.
For buyers, this could reduce repetitive data entry and free finance teams to focus on exceptions, reconciliation and financial planning.
For suppliers, accepting authenticated agent-initiated payments could help streamline collections and potentially accelerate cash flow, while retaining existing payment acceptance and reconciliation controls.
The three companies are testing whether agentic commerce can make business payments more efficient without removing the approval and accountability structures that finance teams rely on.
Visa Trusted Agent Protocol Provides Authentication and Control
A central component of the pilot is Visa’s Trusted Agent Protocol, which helps participating platforms identify and authenticate authorized AI agents.
As businesses increasingly use AI to carry out tasks across software applications, payment systems need ways to distinguish legitimate agents from unauthorized bots or potentially fraudulent activity.
Visa’s approach combines agent identification with tokenized payment credentials and buyer-defined transaction controls.
The pilot includes several key safeguards:
- Agent authentication: Participating platforms can recognize an authorized AI agent acting on behalf of an identified business.
- Tokenized payment credentials: The system uses tokenized payment details instead of relying on shared supplier-portal login credentials.
- Pre-approved transactions: Agents execute eligible payments according to the buyer’s existing approval instructions.
- Invoice-specific payment paths: Payment activity is linked to the particular invoice and its approved instructions.
- Transaction visibility: Buyers and suppliers can retain information about authorization and payment completion.
These features are intended to help businesses automate payment execution without giving AI agents unrestricted authority over company funds.
However, authentication and transaction controls do not eliminate all risks. Businesses still need appropriate approval processes, monitoring, security measures and procedures for handling incorrect invoices, disputed transactions or unusual payment requests.
BILL and Billtrust Connect Buyers With Suppliers
The partnership brings together three different parts of the commercial payments process.
Visa provides payment infrastructure and agent-authentication capabilities through its Intelligent Commerce portfolio and Trusted Agent Protocol.
BILL provides accounts payable technology for businesses. Its authorized Payment Agent initiates eligible payments on behalf of customers using the buyer’s approved instructions.
Billtrust provides accounts receivable and payment solutions for suppliers. In the pilot, it enables participating supplier portals to recognize and authenticate the authorized agent so that payments can be completed within the supplier’s existing acceptance rules.
The collaboration is designed to connect invoice approval on the buyer’s side with payment acceptance on the supplier’s side.
This is particularly relevant where suppliers accept card payments through their portals but buyers have traditionally relied on bank transfers, checks or manual card entry to settle invoices.
The pilot could help businesses automate more of these transactions, although the extent of any savings or improvement in payment speed will depend on individual workflows, fees and supplier arrangements.
AI Agents Could Change Commercial Payments
The initiative is part of a broader shift toward agentic commerce, in which AI systems do more than provide recommendations or answer questions. Instead, they can perform authorized tasks on behalf of individuals and businesses.
In consumer commerce, this could involve AI agents helping shoppers find products, compare options and complete purchases. In corporate finance, agents could support invoice processing, payment execution, reconciliation and other repetitive financial tasks.
B2B payments are a particularly relevant use case because businesses already operate within established approval systems, spending policies and supplier agreements.
Rather than allowing an AI system to make unrestricted financial decisions, companies can define what it is permitted to do and require human approval before funds are committed.
Visa’s pilot tests whether this approach can work across supplier portals that were not necessarily designed for autonomous software agents.
If the model proves reliable and expands beyond the current pilot, it could help businesses reduce manual work in accounts payable and make it easier for suppliers to accept payments from automated business systems.
Nevertheless, a live pilot does not establish that the technology is ready for universal deployment. Broader adoption will depend on integration costs, security performance, supplier participation, regulatory requirements and the reliability of automated payment workflows.
What the Pilot Means for the Future of AI Payments
Visa, BILL and Billtrust have confirmed that the pilot is live and processing end-to-end transactions with participating buyers and suppliers.
The next question is whether the model can scale to a wider range of suppliers, payment platforms and corporate finance systems.
Several factors will influence its potential impact:
- Security and reliability: Platforms must consistently authenticate legitimate agents and prevent unauthorized transactions.
- Business adoption: Buyers and suppliers need compatible systems and clear procedures for exceptions and disputes.
- Payment economics: Companies must evaluate card fees, processing costs and the benefits of faster payment execution.
- Compliance and accountability: Businesses need clear responsibility for agent actions, transaction records and approval controls.
- Interoperability: Agent-authentication standards must work across different payment systems and supplier portals.
The announcement does not disclose transaction volumes, processing-cost savings or a timeline for general availability. Those details will be important for assessing the pilot’s commercial impact.
For now, the project represents a practical test of AI agents moving from administrative assistance to executing approved financial transactions in live business environments.
Bottom Line
Visa, BILL and Billtrust have launched a live pilot that allows authorized AI agents to initiate and complete approved B2B invoice payments through supplier portals.
The system uses Visa’s Trusted Agent Protocol, tokenized payment credentials and buyer-defined controls to automate a step that often still requires manual data entry. It is designed to maintain transaction visibility and supplier oversight without allowing AI agents to independently decide which payments should be made.
The initiative highlights a potential new role for AI in corporate finance: carrying out approved payments across different software platforms. Its broader significance will depend on security, adoption, cost efficiency and whether the pilot can expand beyond participating businesses.
Disclaimer: This article is for informational purposes only and does not constitute financial, investment or business advice. The project remains a pilot, and its results should not be interpreted as proof of industry-wide adoption or guaranteed cost savings.
Also Check: Grayscale Lets Eligible Investors Swap Crypto for Shares in 10 ETFs and Trusts
