The payments infrastructure that has powered global commerce for decades was meticulously designed for human interaction. Now, as artificial intelligence agents mature and begin to autonomously execute transactions, route funds, and manage complex multi-party financial settlements, this established infrastructure is revealing significant limitations. Two recent protocol developments, one from payments giant Stripe and another from fintech startup iWallet, highlight this growing chasm and signal the urgent need for a paradigm shift in how digital transactions are conceived and executed.
The advent of AI agents capable of independent action presents a fundamental challenge to systems built on the assumption of human oversight. These agents are no longer confined to merely authorizing actions that a human will later approve; they are increasingly expected to complete purchases, manage intricate payment flows across numerous stakeholders, and settle diverse financial obligations without any manual intervention. The infrastructure that Stripe pioneered in 2010, aimed at simplifying online payments for human-driven e-commerce, was not architected for this new era of autonomous economic activity, and this disconnect is fast becoming a critical bottleneck.
The tension is palpable, underscored by two distinct but related initiatives. Tempo, in collaboration with Stripe, has introduced the Machine Payments Protocol (MPP). This protocol aims to establish a standardized method for programmatic transactions between AI agents and the services they interact with. Concurrently, iWallet, a fintech company serving the home services sector, has outlined its Autonomous Settlement Protocol (ASP). ASP is designed to automate multi-party financial settlements triggered by verified physical events, such as the completion of an HVAC repair. While these efforts originate from different teams and address different layers of the problem, they collectively illuminate the profound transformations required within the payments ecosystem.
Payments: Architected for Human Interaction
The bedrock of contemporary payment systems is the implicit assumption of human involvement at every stage. Establishing an account necessitates identity verification. Selecting a service plan requires navigating a user interface. Inputting billing details involves completing a form. Even the most sophisticated APIs for payment processing are, at their core, designed for configuration by human engineers and authorization by human cardholders.
AI agents fundamentally disrupt these assumptions. An agent tasked with autonomously scheduling cloud computing resources, making a call to a paid API, and then allocating the associated costs to the appropriate departmental budget cannot afford to be interrupted by human verification steps. Similarly, an agent orchestrating a complex travel itinerary, booking flights and accommodations, cannot be expected to pause and enter corporate card details into a checkout form. Any mandatory human touchpoint halts the very autonomy that makes these agents valuable.
This challenge intensifies as agentic systems grow more sophisticated. While simple per-token API calls are already common, the more complex scenarios involve agents making purchasing decisions, initiating real-world actions, and managing ongoing financial relationships on behalf of individuals or organizations. Without a new framework, every such transaction risks becoming a cumbersome support ticket.
Layer One: The Machine Payments Protocol (MPP)
The Machine Payments Protocol, developed by Tempo and Stripe, directly addresses the transactional layer. It establishes a standardized protocol for AI agents to request resources, authorize payments, and execute transactions programmatically. Businesses integrating with MPP can leverage existing Stripe infrastructure, supporting a range of payment methods including credit cards, buy-now-pay-later options, and stablecoins, while retaining their current reporting, fraud protection, and settlement processes.
Early applications of MPP demonstrate its potential. Agents are already using it to pay for individual API calls, purchase on-demand services, and trigger real-world actions like sending physical mail or placing food orders. These use cases, while automated on the buyer’s side, still represent relatively straightforward, discrete, single-party transactions from the merchant’s perspective, often configured by humans.
However, MPP’s current scope does not extend to the settlement layer—the complex process of distributing funds when multiple parties are entitled to portions of a single transaction. This is where the home services industry provides an unexpected yet instructive case study.
Layer Two: Multi-Party Settlement in the Physical World
The home services industry in the United States alone generates over a trillion dollars annually. A single installation, such as an HVAC system, can involve a chain of participants including the manufacturer, a regional distributor, a licensed contractor, a financing company, and potentially one or more utility rebate programs. Each of these entities expects to receive a portion of the total payment. Currently, managing these diverse financial relationships relies on manual invoicing, spreadsheets, and reconciliation processes that can drag on for days or even weeks after the service is rendered.
Jim Kolchin, CEO of iWallet, frames this challenge not as a payments problem, but as a settlement problem. "Payments move money from the customer," Kolchin explains. "Settlement determines where that money ultimately goes. Our goal is to automate that entire process for the service economy." This distinction is critical, as it highlights the need for different solutions at different layers of the financial stack.
iWallet, which already processes hundreds of millions of dollars in payment volume for contractors and service companies, is developing iWallet 4.0. This next iteration will introduce what Kolchin describes as a programmable settlement layer. Once a payment is captured, a predefined set of rules will automatically direct funds to every participant in the job’s supply chain.
A key innovation within the ASP design is machine verification. The environments in which home services are performed are increasingly data-rich. Modern HVAC systems transmit performance telemetry, installers photograph completed work, and equipment carries serialized identifiers that can be cross-referenced with manufacturer records. Energy efficiency programs collect installation data to determine rebate eligibility.
The ASP leverages AI agents to analyze these various signals—job documentation, equipment data, installation images, and IoT device readings—to confirm that a service event has been completed correctly. This verification event then automatically triggers the settlement sequence, eliminating the need for manual paperwork submission or waiting for rebate program processing. The completion of the physical event itself closes the financial loop.
The Technical Hurdles to Autonomous Economic Loops
Both the MPP and ASP approaches are moving in the right direction, but both encounter significant engineering challenges.
Identity and Trust in Machine-to-Machine Payments: Authorizing payments between machines requires establishing robust identity and trust mechanisms that current systems do not cleanly support. When an AI agent presents credentials to complete a purchase, determining accountability becomes complex. How are spending limits enforced? What recourse exists when an agent makes an error or is manipulated into a fraudulent transaction? Existing risk systems and fraud detection pipelines are trained on human behavioral patterns, meaning agent-initiated transactions may appear anomalous by default.
Standardization in Multi-Party Settlement: The settlement layer introduces another layer of complexity. Programmable fund distribution necessitates formalized agreement structures among parties that currently lack standardization. A contractor, a manufacturer, and a utility rebate program operate with distinct systems, timelines, and payment expectations. Achieving agreement on a machine-readable settlement specification before a job commences is a coordination challenge that extends far beyond software development.
Reliability of Machine Verification: The accuracy of machine verification systems raises its own set of concerns. Computer vision systems confirming that an installation photograph meets expected standards, or that sensor telemetry indicates a successful repair, must perform with a high degree of reliability. False positives could trigger payments for incomplete work, while false negatives could delay legitimate contractors. The overhead generated by disputed outcomes must not outweigh the efficiencies gained.
Regulatory Ambiguity: Underlying these technical advancements is a regulatory landscape that has not yet caught pace with technological evolution. Automated fund disbursement across multiple parties begins to resemble money transmission, a highly regulated activity. The regulatory framework governing agent-authorized transactions and liability for errors remains largely undefined.
The Evolving Payments Stack
For autonomous economic loops to close reliably, a new integrated stack is required:
-
Transactional Layer: Protocols like MPP need to mature beyond their initial use cases to handle authorization, identity, and error recovery in a manner that financial systems deem trustworthy. The inclusion of stablecoins in MPP is particularly noteworthy. Stablecoins can bypass some of the friction inherent in legacy card networks and may emerge as the native currency for agent-to-agent commerce due to their programmability.
-
Settlement Layer: The challenge here is more about standardization than pure technology. The software for conditional fund routing already exists. What is lacking is a universal specification language that multiple parties—including equipment suppliers, financing companies, warranty administrators, and rebate programs—can agree to use for encoding their payment terms. iWallet is effectively attempting to create such a specification for a specific vertical. The broader question is whether this approach can generalize across industries.
-
Verification Layer: This layer demands AI systems capable of confirming real-world service events with sufficient accuracy to justify triggering financial consequences. This is where the integration of physical IoT systems and agentic AI represents a frontier of novel engineering. The necessary signals—equipment telemetry, image documentation, and serialized component tracking—are becoming available. However, integrating them into a pipeline that settlement logic can reliably trust requires infrastructure investments that many service industries have yet to make.
Current Landscape and Future Trajectory
As of this writing, MPP is a live, production-ready protocol facilitating discrete agent-to-service transactions. iWallet’s existing platform is also operational and processes significant payment volumes. The Autonomous Settlement Protocol (ASP) represents a forward-looking design rather than a currently shipping product.
The gap between MPP’s current capabilities and ASP’s vision serves as a useful roadmap for the work ahead. While paying for a single API call is largely a solved problem, automatically distributing the proceeds of that call across ten counterparties, based on machine-verified physical outcomes, is several years of infrastructure development away from becoming routine.
The AI agents being developed today are, at best, authorized to initiate human-reviewed payments. However, the trajectory of development suggests that within three years, agents will be expected to close economic loops autonomously. The necessary infrastructure for these advanced agents is still in its nascent stages, and the teams building it are just beginning to coordinate their efforts.
The fintech revolution of the 2010s democratized the ability for humans to collect money online. The next decade’s innovation will likely focus on enabling software to manage money on behalf of humans, and eventually, between machines, with the reliability that underpins all financial relationships. The work undertaken by Tempo/Stripe and iWallet represents early bets on the architecture of this future. The fundamental engineering challenges remain wide open, offering significant opportunities for innovation in the coming years.
