In recent years, artificial intelligence (AI) has begun to transform not only the automation of business processes but also the way online purchases, payments, and commercial interactions take place. Emerging in this context is the concept of agentic commerce, a model in which AI agents can initiate, manage, or complete transactions on behalf of users. For Italian digital platforms and software-as-a-service (SaaS) companies, this change requires a rethinking of payment infrastructure, checkout flow management, and billing logic.

Preparing for agentic payments means designing a payment system capable of handling transactions initiated by intelligent software, while ensuring authentication, explicit consent, security, and regulatory compliance. This article will explore what agentic commerce is, how AI-powered workflows function, and what the operational implications are for Italian businesses.

Key takeaways

• Agentic commerce allows AI agents to make purchases and perform operations on behalf of users.
  
• Italian companies that intend to adopt agentic commerce must adapt their payment systems to automated workflows and AI-driven decisions.
  
• Checkouts initiated by agents require strong authentication, verifiable consent, and advanced security controls.
  
• Payment management must support complex models, such as dynamic subscriptions, usage-based billing, and multilevel approvals.
  
• Regulatory compliance remains a key priority, specifically with regard to PSD2, data protection, and transaction traceability.
  
• Stripe offers tools for building scalable payment and billing infrastructures that are ready for agentic commerce.
  

What agentic commerce is and how it works

Agentic commerce is a business model in which AI agents can perform operations on behalf of users or companies. Unlike traditional automation systems based on static rules, these AI agents are capable of interpreting objectives, evaluating options, and initiating operations autonomously within defined parameters.

In ecommerce, this means that an AI agent can search for products or services, compare prices and terms, choose the most suitable payment methods, and complete the checkout process autonomously, even handling renewals or recurring purchases.

For instance, business software could automatically monitor cloud resource consumption and acquire additional capacity when needed. Similarly, an AI assistant integrated into a SaaS platform could renew licences, purchase credits, or upgrade pricing plans without requiring constant manual intervention.

This proposition radically changes the role of payment systems. Platforms must no longer exclusively support human users interacting directly with an interface, but also software agents that perform operations autonomously, whether programmatically or contextually.

Differences between traditional automation and AI agents

Many companies already use automated systems for managing payments such as recurring charges or automatic renewals. However, in agentic commerce, AI agents can make real-time decisions based on dynamic data. For instance, a SaaS system could:

• Automatically modify a customer’s subscription plan
  
• Increase usage limits
  
• Distribute payments among multiple suppliers
  
• Approve expenses within predefined thresholds
  

Accordingly, the payment system must be designed to handle more complex payment flows than those typically seen in traditional online payments.

What agentic commerce means in practice for Italian companies

For Italian businesses, agentic commerce is not merely a technological advancement, but a tangible shift in how purchases, digital services, and customer relationships are managed. Companies must establish the necessary infrastructure, policies, and processes to ensure that AI agents can interact securely and effectively with company payment systems.

New expectations for platforms and SaaS

With the proliferation of AI-driven workflows, users will begin to expect increasingly automated experiences. Digital platforms and SaaS companies will need to enable AI agents to manage recurring purchases, perform automatic upgrades, and optimise costs and usage, while keeping payments within limits set by the company or user.

This means that payment management can no longer be viewed merely as the final step of the checkout process, but must become an integral part of the entire digital experience.

More flexible infrastructure

To support agentic commerce, a payment system must be:

• API-centric
  
• Scalable
  
• Programmable
  
• Auditable
  
• Compatible with AI-powered workflows
  

Many legacy systems used by Italian companies are not designed to support this level of automation. As a result, businesses might need to modernise their payment infrastructure by integrating more flexible tools capable of processing automated decisions in real time.

Risk management and oversight

Even though an AI agent can initiate transactions on its own, companies must maintain clear controls and adequate levels of oversight. In many cases, it will therefore be necessary to establish approval thresholds, spending limits, authorisation workflows, and verification tools.

For instance, a company might allow an AI agent to automatically purchase services up to a certain monetary amount, while requiring human approval for operations exceeding that amount.

How agent-initiated checkout flows work

One of the most important aspects of agentic commerce involves checkouts initiated by AI agents. In this scenario, the agent acts as an intermediary between the user, the platform, and the payment system, performing authorised operations based on predefined rules, preferences, or objectives.

The checkout flow

A typical AI-powered checkout process might follow these steps:

1. The user defines preferences, budgets, or permissions.
   
2. The AI agent identifies a product or service that meets those criteria.
   
3. The agent initiates checkout.
   
4. The payment system verifies authorisations, identity, and transaction conditions.
   
5. The transaction is approved or subjected to further review.
   
6. The platform logs the transaction and generates the necessary documentation.
   

Payment management must therefore be able to distinguish between transactions carried out directly by users and transactions initiated by authorised agents, taking into account the level of autonomy assigned to the agent and any verifications required.

Tokenisation and secure credentials

To prevent AI agents from directly accessing sensitive card data, many payment systems use tokenisation, whereby the payment system retains the actual credentials and provides the agent with temporary or limited-use tokens.

This approach enhances security, reduces the exposure of financial data, and allows for more precise control over the permissions granted to AI agents.

Invisible, contextual checkout

In agentic commerce, the checkout process can become virtually invisible to the end user. For instance, software can automatically purchase cloud resources, an AI assistant can renew SaaS tools, or a platform can allocate advertising budgets in real time. Payment processing must therefore operate in the background without compromising transparency, control, or regulatory compliance.

How can companies enable agent-driven checkout?

To enable AI agent-driven checkouts, companies must set up payment flows that allow intelligent software to complete transactions within rules defined by the user or the company. This means enabling AI agents to initiate automatic purchases, renewals, or upgrades while maintaining control, security, and traceability.

To support these payment flows, platforms typically use payment APIs, configurable authorisations, and billing systems capable of handling subscriptions, variable usage, and automatic approvals. Solutions such as Stripe Payments and Stripe Billing help companies develop more automated checkout experiences that are compatible with agentic commerce.

Authentication, user consent, and security

In the context of agentic payments, authentication and consent are very important. Companies must ensure that every transaction initiated by AI is authorised, traceable, and compliant with European regulations.

Strong authentication and PSD2

In Europe, the revised Payment Services Directive (PSD2) requires Strong Customer Authentication (SCA) for many electronic transactions. This requirement applies even when the payment is initiated by an AI agent. The payment system must be able to:

• Verify user identity
  
• Confirm consent
  
• Apply any exemptions provided for by law
  

Platforms must therefore design workflows that allow users to pre-authorise specific agent actions.

Granular consent

In agentic commerce, consent is no longer simply a matter of accepting or rejecting. Users might want to authorise only specific amounts, particular product categories, select providers, or recurring transactions within predefined time frames. For this reason, payment management systems must support granular, configurable policies.

Security and fraud prevention

The rise in automated transactions introduces new risks: malicious actors could attempt to manipulate AI agents or exploit overly broad permissions. Therefore, payment systems must integrate:

• Anomaly monitoring
  
• Fraud detection
  
• Behavioural verification
  
• Real-time risk monitoring
  
• Systems for the immediate revocation of authorisations
  

SaaS platforms must also maintain detailed logs of the actions performed by AI agents to ensure auditability (i.e., the ability to reconstruct and validate the operations performed) and traceability.

Billing logic for agent-driven transactions

One of the most complex aspects of agentic commerce involves billing logic. Companies must manage dynamic pricing models, variable consumption, and automated approvals.

Dynamic subscriptions

Many SaaS services already use subscription models. However, in agentic commerce, AI agents could automatically modify plans, add features, increase usage limits, or purchase additional credits based on user needs or detected usage levels.

Pay-as-you-go models

More and more platforms are adopting pay-as-you-go (consumption-based) models. In such a case, an AI agent could monitor usage levels and automatically provision new resources, such as server capacity, API calls, storage space, or additional software features.

Approval workflows

Not all operations can be fully automated. Some companies might want to implement multilevel approvals for more sensitive or higher-risk transactions.

Accordingly, payment processing must be able to combine AI automation with human oversight.

International invoicing and value-added tax (VAT)

Italian companies operating online must also consider:

• Italian VAT
  
• One Stop Shop (OSS) regime
  
• Electronic invoicing
  
• Multicurrency management
  
• International tax compliance
  

When an AI agent initiates a cross-border transaction, the payment system must automatically apply the correct tax rules.

Does agentic commerce comply with payment regulations?

Yes, agentic commerce can comply with payment regulations, provided that companies design their authorisation, authentication, and control workflows in a secure and transparent manner. Even when a transaction is initiated by an AI agent, regulatory requirements such as SCA under PSD2, European data protection rules, and security standards for payment information continue to apply.

To ensure compliance, companies must therefore pay particular attention to three key aspects: the traceability of operations performed by AI agents, the protection of personal and financial data, and the establishment of clear accountability and governance mechanisms for automated activities. Let’s look at these in detail.

Traceability of operations

Every AI-powered operation should generate a complete log of the activities performed, including:

• User identity
  
• AI agent involved
  
• Authorisation parameters
  
• Exact date and time of the operation
  
• Decision logs (i.e., records that explain which rules, data, or conditions led the AI agent to make a specific decision)
  
• Authentication status
  

This information helps companies prove regulatory compliance, facilitates audits, and allows for the accurate tracing of each transaction.

Data protection

AI-powered applications can process large amounts of personal and financial data. Payment processing must comply with:

• General Data Protection Regulation (GDPR)
  
• Payment Card Industry Data Security Standard (PCI DSS)
  
• European data protection regulations
  
• Secure information retention policies
  

Platforms must also restrict AI agents’ access to only the data strictly necessary to complete the transaction.

Accountability and governance

One of the most sensitive issues concerns accountability for automated operations. If an AI agent makes an incorrect or unauthorised payment, companies must be able to accurately reconstruct which agent initiated the transaction, what authorisations were granted at that time, and what operational or spending limits applied to the transaction.

For this reason, it’s important to define:

• Spending limits
  
• Approval rules
  
• Oversight policy
  
• Dispute procedures
  
• Systems for the immediate revocation of authorisations
  

In practice, the payment system must support clear mechanisms for the control, governance, and oversight of AI-powered activities.

How Stripe supports an agent-ready payment and billing infrastructure

To prepare for agentic commerce, Italian companies need a flexible, scalable, and programmable infrastructure. With tools such as Stripe Payments and Stripe Billing, digital platforms and SaaS companies can build payment and billing systems ready for AI-powered workflows.

Scheduled payments for AI-powered workflows

Stripe’s APIs enable you to integrate advanced payment features into platforms, SaaS companies, marketplaces, and AI-driven applications. This allows for the automation of complex processes while maintaining compliance with company policies.

For instance, companies can:

• Automate checkouts initiated by AI agents
  
• Manage dynamic authorisations
  
• Process recurring payments
  
• Monitor transactions in real time
  
• Create custom workflows for transaction oversight
  

Thanks to an API-centric infrastructure, platforms can adapt their payment systems to increasingly automated flows.

Recurring billing and pay-as-you-go models with Stripe Billing

In agentic commerce, AI agents can automatically modify subscriptions, increase usage limits, or activate additional services. For this reason, billing management becomes a central component.

With Stripe Billing, companies can manage:

• Subscriptions
  
• Flexible pricing models
  
• Automatic upgrades
  
• Custom renewals and authorisations
  

This helps SaaS platforms create billing logic compatible with agentic commerce and pricing models based on actual service usage.

Learn more about Stripe Billing or get started today.