The global payments industry has always relied on a simple principle: a human decides to make a purchase, and a bank or card network processes the transaction. Every card swipe, online checkout, or mobile payment follows this model, where identity and intent are linked directly to a person. However, that long-standing structure is beginning to change as artificial intelligence becomes more capable of making decisions on behalf of users.
Visa is now testing how payment infrastructure should work when transactions are initiated not by people, but by software. Through its new Agentic Ready programme in Europe, the company is exploring how financial systems can support AI agents that search for products, compare prices, and complete purchases automatically. The initiative involves collaboration with major banking partners, including Commerzbank and DZ Bank, and focuses on adapting existing payment networks to a future where software may act as the buyer.
The shift could represent one of the biggest changes in payments since the rise of online banking. If AI agents begin to handle routine transactions, banks, regulators, and payment networks will need new ways to verify identity, confirm consent, maintain compliance, and manage risk. Visa’s current work is not about consumer apps, but about building the infrastructure required for a world where transactions may start with an algorithm rather than a human.
This article explores how Visa’s programme works, why AI-initiated payments are gaining attention, what challenges banks face, and how the financial system may evolve as software agents become active participants in commerce.
The Traditional Payment Model Is Built Around Human Intent
Modern payment systems are designed around the assumption that a person is always responsible for a transaction. Whether using a credit card, bank transfer, or digital wallet, the process depends on verifying that the customer has authorized the purchase.
Every transaction today involves several steps:
- The user chooses a product or service
- The payment method is selected
- The bank verifies identity and funds
- The network processes the transaction
- The merchant receives confirmation
This model works because identity and intent are clearly linked to a human user. Security systems, fraud checks, and compliance rules are all designed with that assumption.
However, advances in artificial intelligence are starting to challenge this structure. AI systems can now analyze data, make decisions, and perform tasks automatically. In some cases, they can already handle activities that previously required human input, such as scheduling, budgeting, and inventory management.
If software can decide what to buy and when to buy it, payment networks must be able to process transactions that do not begin with a person pressing a button.
Visa’s “Agentic Ready” Programme Tests AI-Initiated Payments
To prepare for this change, Visa has launched its Agentic Ready programme in Europe. The initiative focuses on testing how payment infrastructure should operate when an AI agent initiates a transaction instead of a human user.
The programme is being developed with participation from banks including Commerzbank and DZ Bank. The goal is to make sure that existing systems can handle a new type of transaction flow without breaking security or regulatory requirements.
According to information published by Visa and reported by The Paypers, the project explores scenarios where AI software is given instructions or rules and then completes purchases automatically. Instead of the user confirming each payment, the system acts based on predefined conditions.
For example, an AI agent could:
- Monitor prices for a product
- Compare suppliers
- Check available budget
- Complete the purchase when conditions are met
In this case, the transaction still belongs to the user, but the decision and execution are handled by software.
Visa believes this shift could be as significant as the move from physical payments to online transactions, when banks had to redesign systems to support internet-based commerce.
Why AI Agents Are Becoming Part of Financial Systems
Artificial intelligence is already widely used in banking for fraud detection, risk analysis, and customer support. The next step is allowing AI to perform actions, not just provide information.
Software agents are being developed to handle repetitive or rule-based tasks, such as:
- Paying recurring bills
- Ordering supplies
- Managing subscriptions
- Handling procurement requests
- Executing approved purchases
In large organizations, these processes often involve multiple manual steps. AI agents can reduce time and cost by automating routine decisions.
As these systems become more reliable, companies want them to interact directly with payment networks instead of requiring human approval for every transaction.
This is why payment providers like Visa are preparing their infrastructure now, before AI-initiated transactions become common.
Identity and Authorization Become More Complex
One of the biggest challenges with AI-initiated payments is proving identity and intent.
Today, banks verify that a person authorized a transaction using methods such as:
- PIN codes
- Passwords
- Biometrics
- Two-factor authentication
- Card verification
If an AI agent initiates the transaction, the system must confirm that the agent is allowed to act on behalf of the user.
This raises several questions:
- How does the agent prove it represents the customer?
- How much autonomy should the agent have?
- Can the agent spend money without approval?
- What limits should be set?
Visa’s programme focuses on defining rules that allow software agents to operate safely while still meeting banking regulations.
The goal is to create a system where the payment network can verify not only the user, but also the software acting for that user.
Security and Compliance Remain Critical
Any change to payment systems must follow strict regulatory rules. Banks are required to maintain detailed records, prevent fraud, and ensure that customers give proper consent.
Commerzbank and DZ Bank, which are involved in early testing, are exploring how AI agents can be integrated without breaking compliance standards.
Key areas being tested include:
- Fraud detection
- Audit trails
- Customer authorization
- Transaction monitoring
- Dispute resolution
These controls are essential because financial transactions are highly regulated. Even small changes to how payments are initiated must be carefully reviewed.
A report from RepRisk notes that banks are already dealing with more frequent incidents related to AI, some of which have resulted in multi-million-dollar losses. This makes regulators cautious about allowing software to take a more active role in financial decisions.
As a result, Visa’s work is focused on infrastructure and system design rather than consumer products.
Payments May Be One of the First Areas to Use Agentic AI
Many industries are experimenting with AI, but payments could be one of the first areas where software agents operate with real autonomy.
This is because payment processes are structured and rule-based. Transactions follow clear steps, which makes them easier to automate compared to more complex decisions.
For example, an AI agent could be given rules such as:
- Buy only from approved vendors
- Stay within a spending limit
- Compare at least three prices
- Require approval for large purchases
With these limits in place, the agent can handle routine transactions safely.
Visa’s research suggests that as long as rules are defined clearly, AI can reduce manual work without increasing risk.
Enterprise Procurement Is a Key Use Case
Large companies often have complex purchasing processes that involve multiple approvals. Even simple orders may require several steps before payment is completed.
AI agents could simplify this workflow by handling routine purchases automatically.
For example, a company system could:
- Detect low inventory
- Request quotes from suppliers
- Select the best price
- Complete the payment
- Record the transaction
This reduces the need for manual input while keeping spending within approved limits.
However, companies must define clear policies to prevent mistakes or misuse. Without strong rules, automation could create new risks instead of reducing them.
This is why banks and payment networks are testing AI transactions carefully before allowing them at scale.
Infrastructure Must Adapt to a New Type of User
Visa’s programme highlights an important shift: the payment network must learn to treat software as a participant in the system.
Traditionally, the “customer” in a transaction is always a person or business. In the future, the initiating party may be an AI agent acting on behalf of that customer.
This requires changes in several areas:
- Authentication methods
- Authorization rules
- Transaction monitoring
- Error handling
- Dispute resolution
For example, if an AI agent makes a purchase that the user did not expect, the system must be able to track who gave permission and why the transaction happened.
These requirements mean that payment infrastructure must evolve before agent-based commerce becomes common.
AI Use in Banking Is Expanding Rapidly
Financial institutions are investing heavily in AI to reduce costs and improve efficiency.
Current uses include:
- Fraud detection
- Risk analysis
- Customer service automation
- Data processing
- Compliance monitoring
Many banks are also reorganizing teams to focus more on data and AI strategy.
As AI becomes part of daily operations, allowing software to initiate transactions is a natural next step.
However, regulators are watching closely, especially in areas where AI affects financial decisions.
This means changes will likely happen gradually rather than all at once.
Lessons From the Shift to Online Payments
Visa compares the rise of AI-initiated transactions to the early days of online payments.
When e-commerce first appeared, banks had to redesign systems to support transactions that did not happen in person.
New technologies were needed for:
- Secure authentication
- Fraud prevention
- Encryption
- Online authorization
Today, those systems are standard.
AI-initiated payments may follow a similar path. At first, they will be limited and carefully tested. Over time, they could become part of everyday financial activity.
The current Agentic Ready programme is an early step in that transition.
Testing Comes Before Consumer Rollout
Visa has made it clear that its current work is focused on backend systems, not consumer tools.
The goal is to understand how payment networks should behave when the transaction starts with software.
This includes testing:
- How agents identify themselves
- How banks approve transactions
- How records are stored
- How disputes are resolved
Only after these questions are answered will companies consider large-scale deployment.
This cautious approach is necessary because payment systems must remain reliable and secure at all times.
The Future of Payments May Include Autonomous Transactions
As AI becomes more advanced, the idea of autonomous transactions is becoming realistic.
In the future, users may set goals instead of making individual purchases.
For example:
- “Buy the cheapest flight next month”
- “Reorder supplies when stock is low”
- “Pay bills automatically if under budget”
An AI agent could handle these tasks without manual input, as long as rules are defined in advance.
This would make payments faster and more efficient, but it also requires strong safeguards.
Visa’s current research is helping build those safeguards.
Conclusion
Visa’s Agentic Ready programme shows how the payments industry is preparing for a future where AI agents can initiate transactions. By working with banks such as Commerzbank and DZ Bank, the company is testing how existing infrastructure must change to support software-driven commerce.
The shift raises important questions about identity, authorization, security, and regulation. Payment systems were built for human users, and adapting them for AI requires careful planning.
Early testing suggests that AI-initiated payments could reduce manual work, speed up purchasing, and improve efficiency, especially in enterprise environments. However, strict rules, strong data controls, and reliable authentication will be essential.
Just as online payments transformed banking in the past, agent-based transactions may define the next stage of financial technology. For now, the focus remains on building the foundation that will allow AI to participate safely in the global payment network.
Read Also: For Effective AI, Insurance Companies Must Fix Their Data Foundations First