a16z: Why Do AI Agents Need a Stablecoin for B2B Payments?

Original Title: Tourists in the Bazaar: Why Agents Will Need B2B Payments—and Why Stablecoins Will Get There First

Original Author: SamBroner, a16z crypto

Original Translation: AididiaoJP, Foresight News

As a tourist walking through the bazaar, you witness a scene: people bustling around, looking at goods, comparing items, trying samples, negotiating with each stallholder, and making payments. It seems like a series of one-off transactions, where each interaction is a small negotiation, trust mediated by cash in hand or exchanged through a bank card for value.

But this isn't how most business is conducted in the bazaar. Take a closer look: most are locals with clear destinations heading towards their favorite merchants. A restaurant owner visits his friends: the butcher, fishmonger, and farmer. A tailor seeks out the mechanic, weaver, and artisan. Both extend credit to each other.

When we think about how smart agents will pay, we usually default to the tourist perspective.

However, smart agents will act like locals. The characteristics that differentiate smart agents from humans—infinite replicability, flexible resource allocation, zero startup costs—mean that a small number of smart agents can win niches in the market. Even as building smart agents becomes easier, relationships, partnerships, and trust still help create winning experiences. Dominant smart agents don't need the tourist payment rails. They require vendor relationships, working capital, and credit.

What will this look like? As smart agents consolidate into commercial-like platforms, agent payments must shift from retail payment rails to pre-negotiated B2B terms and credit, an opportunity current rails can't fully meet. This marks the window of opportunity for the next generation of payment rails, such as stablecoins, if entrepreneurs can build great solutions for the next wave of payment scenarios like smart agents, streaming payments, and high-volume, low-value global business.

This article will explore this viewpoint in three parts: how smart agents differ from humans and how these differences shape winning payment strategies; why current approaches fall short; and what needs to be built for the next generation of payment rails to succeed.

How Smart Agents Differ from Humans

To understand Smart Agent and Payments, we must consider two questions: Will the behavior of the Smart Agent be more like a person or like a business? Will the Smart Agent play the long game or the short game?

The Smart Agent will be more like a business, establishing long-term relationships with its suppliers and partners. The Smart Agent will be an instance of light customization on top of a larger corporate structure—much like a perfect tour guide in a well-established travel agency, or a franchisee adjusting operational manuals to suit local tastes without renegotiating the entire supply chain.

Why Will the Behavior of the Smart Agent Be Like a Business?

First, the best experiences are thoughtfully designed. I don't want an agent who is still wandering around and haggling with suppliers, comparing prices, and negotiating at the checkout. I want an agent who has already done this work—a knowing which suppliers are reliable, have pre-negotiated prices, and can settle immediately. This is a business relationship, not a touristy transaction.

In fact, human agents already exist: travel agencies, literary agents, talent agents, watch dealers, real estate brokers, and more. These agents build crucial, multi-round relationships—with publishers, studios, watch distributors, or mortgage lenders—where each transaction is customized on this foundation.

Second, Smart Agents are infinitely replicable, but scaled businesses (and their advantages) are not. The best Smart Agents will leverage the cost and benefit of scaled enterprises: cheaper computing power, better supplier pricing, deeper integrations, and more certain components. Scale begets scale, where a travel agency booking a million flights a year gets better terms from airlines than one booking just ten times.

We have already seen this. Only ChatGPT has channels to negotiate partnerships with companies like Shopify, Amazon, Expedia, and more. Smaller startups are stuck using browser automation or reverse-engineering APIs while facing a retail fee structure.

That's why Smart Agents will integrate, or at least why most Smart Agents will be built on larger platforms. Building Smart Agents is easy, but economic realities favor only a few Smart Agents per vertical, each with deep supplier relationships and margin to reinvest in a better experience. And with vertically specialized agents possessing profound supplier relationships, they can complement user agents to provide a win-win experience.

Two Payment Relationships

If the Smart Agent acts like a business, then two payment relationships need to be designed: User → Agent and Agent / Agent Platform / "Tour Guide" of the Agent → Supplier.

User Pays Agent – This may be through a subscription, task-based payment, credit line, or granting the agent access to the user's account. The agent then pays the supplier through agreed-upon B2B terms, bulk pricing, net-30 invoices, or through sub-agents. Based on current commercial expenditure, agents occasionally use the retail track to pay suppliers, but even then, this is only a small part of the overall spend.

This is actually how credit cards work today: the issuing bank has a retail relationship with the consumer, takes on risk, creates custom rewards programs, and provides credit. The acquiring bank has a commercial relationship with the merchant, including negotiated terms, batch transfers, and complex operational funding discussions.

Agents and Credit Cards: A McKinsey-Perfect Match

As many have noted, credit cards are actually a rather logical payment product for agent use cases. Credit cards are widely accepted; payments between $20 and $1000 are considered reasonable; and credit cards come with built-in dispute, chargeback, and digital capabilities.

Credit cards also have monthly statements—an essential opportunity for consumers to understand what they have paid for. As smart agents replace kids playing iPad as the primary source of accidental spend, this concept is sure to be iterated upon.

However, there are two problems: First, credit cards are not technically suitable for agents. Second, their fee models have pushed the credit card industry into a classic innovator's dilemma.

Credit Card Technology Is Hard to Upgrade

Nearly all credit card technology presupposes a human-in-the-loop: an approver, a user interface layer, and a traditional payment type (one-time payment, subscription). Stripe Link, Visa 3D, and dozens of other credit card tokenization products—software that lets you save cards on a website for future purchasing or register cards for monthly recurring subscription purchases—are finally running well, but it has taken over 15 years to develop this tech.

The adoption of agents has been too fast, with thousands of PSPs, POS systems, merchants, and client endpoints struggling to slowly upgrade their interfaces, programmability, and fraud detection systems for this new payment flow.

Credit Cards Fall Short in High and Low-Cost Purchases

Imagine an agent streaming payments to a computing provider or facilitating micro-payments for API access. Neither of these payments can run on the credit card rails. Firstly, Visa doesn’t support payments below one cent; secondly, its economic model expects fixed fees of 30 cents. Visa could create the technology to support streaming or micro-payments, but convincing stakeholders to accept lower payment revenues would be much more challenging.

Even more problematic is the credit card's innovator's dilemma. While its user relationship and payment requirements are similar to agent-based payments, agent-based payments often fall outside the range of $20 to $1000. What's worse, many initial scenarios involve payment API fees that are hard to refund or easily resold (fraud).

Even beyond credit cards, traditional rails will still have a place in the future.

Existing Payment Methods Still Play a Role

As smart agents integrate into platform-like businesses, most high-volume spending will shift to pre-negotiated B2B terms: invoices, net-30-day payments, discounts, and credit lines. In that world, the "payment rail" can be anything — often a mundane asynchronous settlement conducted on a traditional rail. Fees are spread across larger transactions, operational funds can be negotiated between two businesses.

But agents won't exist only in that world. Agents are happening, and they are operating in areas where traditional payment methods are suboptimal: initially established relationships, cross-border settlement, simplifying complex reconciliations, new agent-supplier models, instant payments for lower borrowing costs, and microloans.

In these scenarios, stablecoins are a better payment option, and crucially, building the next generation of features on programmable money is easier than on traditional infrastructure. New relationships built using stablecoins will become old relationships still using stablecoins. Over time, as a full stablecoin payment platform comes online, stablecoins (already cheaper, faster, and global) are likely to have a larger share of the payment mix.

Opportunities of New Payment Technologies

To understand what comes next, we should focus on the technologies best suited for constantly expanding use cases.

Stablecoins — backed by highly liquid assets on a 1:1 basis, faster, cheaper, and global currency — represent a new platform to meet the needs of today's underserved business categories, such as international payments and streaming payments. Key stablecoins are programmable, with crucial functionalities such as arbitration, monthly (or hourly) billing, credit, escrow, and conditional payments, that can be flexibly extended to support many new use cases. Unlike bank or card payments, stablecoin payments can be easily integrated into APIs, databases, and agent settlement processes, greatly simplifying reconciliation, approval, and onboarding processes — a huge advantage for entrepreneurs eager to build agent businesses.

At a practical level, stablecoins address the credit card's unit economics problem in extreme cases. With no minimum fee of 30 cents, microtransactions become possible. Without fees swallowing the profits of large transfers, an agent paying a compute provider $0.001 per second and a manufacturer settling a $50,000 supplier invoice can use the same rail. This flexibility is crucial as engineers and entrepreneurs consider the next platform to build.

Building More Stablecoin Infrastructure

The most common criticism of stablecoin usage is the high cost of onboarding and offboarding. For the uninformed "tourist," this is indeed the case, but when users are accompanied and guided by a "guide" — also known as an agent — this issue disappears. The guide can help tourists exchange currencies and facilitate the exact transactions needed while saving on transaction fees.

By adding billing and arbitration features to our stablecoin-supporting guide system, we are closer to the desired system.

Imagine shopping at a department store like Macy's. You browse multiple brands, pick your items, and finally check out at one unified cashier. The store handles the complex transactions to pay each brand. Agents need a similar model: a unified view of items to purchase across multiple vendors, enabling one-click approval for the entire batch. Users see "your agent wants to book a flight, reserve a hotel, and rent a car" — not three separate checkouts. The agent platform manages vendor relationships, while users manage intent. Users can approve, review, or dispute transactions.

Credit cards excel in arbitration, but this layer also needs to be added to the new track. Arbitration is easiest when goods are high-priced or easily returnable. A 24-hour cancellation window for flights, subscriptions not yet started, high-margin luxury goods — vendors can handle reversals. However, early agent scenarios often involve low-margin digital goods, such as compute cycles and API calls, or food delivery.

Summary

Smart agents don't pay like tourists. They pay like locals — through relationships, credit, and repeat transactions. This means real payment volume will flow through pre-negotiated B2B terms rather than credit card swipes. Frankly, pre-negotiated B2B terms do not require a new payment rail. Settlement layers can be anything — wire transfers, ACH, or mundane batch transfers. For established relationships, traditional payment methods work well.

But we are at a crossroads. Agents are emerging, entrepreneurs are building, and they need payment methods that work today — not waiting for years for credit card technology stack upgrades. Credit cards are not ready: too expensive for micropayments, challenging for reconciliation, burdened by technical debt, and fraud decisions requiring human intervention. Stablecoins are ready. They are programmable, global, easy to reconcile with digital services, and seamlessly integrated into APIs and agent checkouts. Even without negotiated merchant agreements or complex B2B terms, they work from day one.

This is the window of opportunity. Entrepreneurs building agents today will seek tools that work well today. Payment is sticky. Ultimately, new relationships built on stablecoins will evolve into old relationships still built on stablecoins. Over the next few years, the ecosystem will mature, onboarding friction will disappear, and gaps in infrastructure — billing, arbitration, credit, batch approval, interoperability — will be filled by a wave of startups building on a stronger foundation.

Original Post Link