a16z: Five Uncharted Territories of AI Payments and Opportunities for Stablecoins

Author: SamBroner, a16z crypto

Compiled by: Block unicorn

Foreword

Strolling through the market as a tourist, you'll witness a bustling scene: people throng, their eyes glued to the goods, comparing items, sampling them, haggling with each vendor, and exchanging currency. It seems like a one-off transaction—each interaction a mini-negotiation, trust maintained through cash, or value exchanged through bank cards.

But this isn't how most transactions at the market work. Observe closely: most people are locals, and they go to their preferred merchants with a specific purpose. Restaurant owners visit their friends, butchers, fishmongers, and farmers. Tailors go to repairmen, weavers, and craftsmen. They all use credit.

When we discuss how smart agents will make payments, we often unconsciously think from the perspective of tourists.

But intelligent agents behave more like locals. What distinguishes intelligent agents from humans is their ability to replicate infinitely, allocate resources flexibly, and have zero start-up cost—meaning a small number of intelligent agents can dominate a niche market. Even as creating intelligent agents becomes easier, interpersonal relationships, partnerships, and trust still contribute to a successful user experience. Dominant intelligent agents don't need payment channels from tourists; they need supplier relationships, working capital, and credit. Intelligent agents can guide tourists (that is, you).

What does this specifically mean? As smart agents integrate into business platforms, their payment methods must shift from retail payment channels to pre-negotiated B2B terms and credit, a need that current payment channels cannot fully meet. If entrepreneurs can build compelling solutions for next-generation payment scenarios—such as smart agents, streaming payments, and global businesses with high-frequency, low-value transactions—then next-generation payment channels (such as stablecoins) will see significant growth opportunities.

This article will explore this view from three aspects: the differences between intelligent agents and humans and how these differences affect the success of payment strategies; the shortcomings of current methods; and what elements need to be built for the next generation of payment channels to succeed.

The difference between intelligent agents and humans

To understand the relationship between intelligent agents and payments, we must consider two questions: Will intelligent agents behave like humans or businesses? Will intelligent agents focus on long-term or short-term interests?

Intelligent agents will resemble businesses more closely, building long-term relationships with suppliers and partners. They are lightly customized entities built upon large enterprise structures—for example, the perfect tour guide from a well-connected travel agency, or a franchisee who can tailor services to local tastes without renegotiating the supply chain.

Why do intelligent agents behave like enterprises?

First and foremost, the best experience stems from careful design. I don't want a smart agent who's still negotiating with suppliers, comparing prices, and negotiating terms at checkout. I want a smart agent that has already done all that work—one that knows which suppliers are reliable, has pre-negotiated prices, and can process the payment instantly. That's what business relationships are about, not travel transactions.

In fact, human agents have existed for a long time: travel agents are one example, but literary agents, talent agents, watch dealers, real estate agents, and so on are also commonplace. Agents establish crucial multi-layered relationships—with publishers, production companies, watch dealers, or mortgage lenders—and each transaction is tailored on this basis.

Secondly, intelligent agents can be replicated indefinitely, but scalable business (and its advantages) cannot. Excellent intelligent agents fully leverage the costs and benefits of scalability: lower computational costs, more favorable supplier prices, deeper integration, and more deterministic components. Scale leads to even greater scale. A travel agency that books one million tickets annually will obtain more favorable terms from airlines than an agent that books only ten tickets annually.

We've already seen this trend. Only ChatGPT has sufficient channels to negotiate partnerships with companies like Shopify, Amazon, and Expedia. Small startups are limited to using automated browsers or reverse-engineered APIs, while also incurring high retail costs.

This is why smart agents are being consolidated, or at least why most smart agents are built on larger platforms. Agents are easy to build, but economics dictates that the number of agents in each vertical should be kept low—each agent should have a strong partnership with the vendor and sufficient profit margin to reinvest in improving the user experience. Furthermore, vertically dedicated agents with strong vendor relationships can work in tandem with user agents, achieving the best of both worlds.

Two payment relationships

If the smart agent operates in a similar manner to an enterprise, then two payment relationships need to be designed: user → agent, and agent/agent platform/agent wizard → supplier.

Users pay agents—through subscriptions, pay-per-task, credit lines, or authorized access to user accounts. Agents pay suppliers using negotiated B2B terms, bulk pricing, 30-day invoices, or through sub-agents. Based on current business spending, agents occasionally pay suppliers through retail channels, but even then, this constitutes only a small portion of total expenditure.

This is how credit cards actually operate today: Issuing institutions establish retail relationships with consumers, assume risk, develop personalized reward programs, and provide credit lines. Acquiring institutions establish business relationships with merchants, negotiate terms, conduct large-scale transfers, and handle complex working capital matters.

Smart Agents and Credit Cards: A McKinsey-Style Perfect Match

As many have pointed out, credit cards are actually a fairly reasonable payment product for smart agents. Credit cards are widely accepted; payments between $20 and $1,000 are considered reasonable; and credit cards have built-in arbitration, cancellation, and digitalization features.

Credit cards also offer monthly statements—an important way for consumers to understand their spending details, and this concept will certainly be further refined as smart agents replace children playing with iPads as the main cause of unexpected expenses.

However, two problems exist: First, credit cards are technically poorly suited for smart agents. Second, the fee structure has forced the credit card industry into a classic innovator's dilemma.

Credit card technology is difficult to upgrade

Almost all credit card technologies rely on human intervention: they require approvers, user interface layers, and traditional payment methods (one-time payments, subscriptions). Stripe Link, Visa 3D, and dozens of other credit card virtualization products—software that lets you save your card on a website for future purchases or register for monthly subscriptions—are finally working well today, but this technology has taken 15 years to develop.

The adoption of smart agents has been so rapid that thousands of payment service providers (PSPs), POS machines, merchants, and client terminals have been unable to keep up with the pace of upgrading their interfaces, programmability, and fraud detection capabilities to accommodate this new payment process.

The credit card cannot be used for either high-value or low-value transactions.

Imagine a smart agent sending money to a computing service provider or paying a small API access fee. Neither of these payment methods can be implemented through credit card channels. First, Visa doesn't support payments below 1 cent; second, its economic model anticipates a flat fee of 30 cents. Visa may be able to develop streaming or micropayment technologies, but getting stakeholders to accept lower payment revenues will be much more difficult.

More problematic is that credit cards are caught in the innovator's dilemma. While smart agent payments share similar user relationships and needs with credit card payments, the amounts typically range from $20 to $1,000. Worse still, many initial offerings involve paying for APIs that are difficult to refund or easily resold (fraudulent). Credit cards are not unworkable, but the innovator's dilemma has long been undermining existing businesses.

Even setting aside credit cards, traditional payment channels will still have a place in the future.

Existing payment methods will continue to play a role.

As smart agents integrate into business platform-like entities, most large expenditures will shift to pre-negotiated B2B terms: invoices, 30-day net payments, discounts, and credit lines. In that world, the "payment channel" can be anything—typically asynchronous settlements over traditional channels, albeit somewhat dry. Fees are spread across larger transactions, and working capital can be negotiated between the transacting parties.

But the space for smart agents is not limited to this. Smart agents have already emerged and are operating in areas where traditional payment methods are difficult to implement: for example, first-time collaborations, cross-border payments, simplifying complex reconciliation processes, new agent-supplier models, instant payments to reduce borrowing costs, and microloans.

In these scenarios, stablecoins are a superior payment option, and crucially, building next-generation functionalities on programmable money is far easier than on traditional infrastructure. New partnerships established using stablecoins will gradually evolve into existing partnerships that continue to use stablecoins. As stablecoin payment platforms become more widespread, stablecoins (which are already cheaper, faster, and more global) are likely to play an increasingly important role in the payment mix.

New payment technologies hold opportunities

To understand future development trends, we should focus on the technologies that are best suited to the ever-growing application scenarios.

Stablecoins—faster, lower-cost, and globally accepted currencies backed 1:1 by high-quality liquid assets—are a revolutionary platform addressing the needs of currently underserved business sectors, such as international and streaming payments. Crucially, stablecoins are programmable. Key features such as arbitration, monthly (or hourly) settlement, credit, custody, and conditional payments can be flexibly expanded to support numerous new use cases. Unlike bank or credit card payments, stablecoin payments can be easily integrated into APIs, databases, and proxy checkout systems, significantly simplifying reconciliation, approval, and registration processes—a significant advantage for entrepreneurs eager to build proxy businesses.

From a practical perspective, stablecoins solve the problem of unit economics in extreme situations associated with credit cards. They avoid the 30-cent minimum transaction fee, thus mitigating the difficulties of small payments. They also avoid the exchange fees that erode profits from large transfers. A smart agent pays a computing service provider $0.001 per second, while a manufacturer needs to settle a $50,000 supplier invoice; both can use the same payment gateway. This flexibility is crucial for engineers and entrepreneurs when considering their next platform.

Build more stablecoin infrastructure

The most common objection to using stablecoins is the high cost of depositing and withdrawing. This is indeed true for tourists unfamiliar with stablecoins, but this problem can be easily solved if users have a guide or smart agent with them. A guide can help tourists exchange currencies and accurately facilitate necessary transactions while saving on transaction fees.

By adding billing and arbitration features to our stablecoin-supported guided service, we will get closer to the ideal system.

Imagine walking into a department store. You browse multiple merchants, add items, and finally settle a consolidated bill all at once. The platform handles the complex process of allocating funds to each supplier. Smart agents need the same paradigm: a unified view showing purchasing intentions across multiple suppliers, with the ability to approve bulk orders with a single click. Instead of three separate checkout processes, users see "Your smart agent wants to book flights, hotels, and car rentals." The agent platform handles the relationships with suppliers, while the user manages the purchasing intentions. Users can approve, review, or object to transactions.

Credit cards excel at arbitration, but new payment channels need to build upon that foundation. Arbitration is most convenient when goods have high profit margins or are easy to return. For example, flights within the 24-hour cancellation window, subscriptions that haven't yet taken effect, and high-margin luxury goods—suppliers can afford refunds. However, early applications of arbitration typically involve low-margin digital goods, such as computing resources and API calls, or food delivery.

Summarize

Smart agents won't pay like tourists. They'll pay like locals—through relationships, credit lines, and repeat customers. This means real payment traffic will flow through pre-negotiated B2B terms, not credit card swipes. Frankly, pre-negotiated B2B terms don't require new payment channels. The settlement layer can be anything—wire transfer, ACH transfer, or tedious bulk transfers. Traditional payment methods are perfectly adequate for established partnerships.

But we are at a critical inflection point. Smart agents are emerging, entrepreneurs are building their systems, and what they need are payment methods that can work immediately, not payment methods that have taken years of upgrades to credit card technology. Credit cards aren't ready: they're too costly for small payments, too difficult to reconcile, burdened by technological debt, and susceptible to human error in fraud detection. Stablecoins, on the other hand, are mature. They are programmable, globally accepted, easy to reconcile with digital services, and can be easily integrated into APIs and smart agent checkout processes. They work from day one, even without negotiated merchant agreements or complex B2B terms.

This is the crucial moment. Entrepreneurs building smart agents today will choose tools that work immediately. Payments are sticky. Ultimately, new relationships built on stablecoins will evolve into old relationships still based on stablecoins. In the coming years, the ecosystem will mature, barriers to entry will decrease, and infrastructure gaps—such as billing, arbitration, credit, batch approval, and interoperability—will be filled by a wave of startups building on stronger foundations.