Scenes, Contradictions, and the Endgame from the Perspective of 16-Year AI Payment Practitioners

Writing by: Ivy & Hazel

AI payments are no longer just a concept. x402, MPP, Tempo, AP2—over the past year, Coinbase, Stripe, Google, Visa have built protocol frameworks at different levels. Real on-chain data, real merchant integrations, and genuine model misreads are beginning to surface one after another.

Last Saturday, the organization 支无不言 held a closed-door Agent Payment meeting, with 16 guests from payment infrastructure, wallet services, large corporate payment businesses, investment institutions, and more. Nearly three hours were spent answering four questions: Where exactly is AI payment happening? How to make AI spend money safely? How does this business make money? And what will be the future of the game between big companies and startups?

Below are the core judgments emerging from this discussion:

The most mature scenario for Agent Payment is API calls, with small amounts of $0.01 supported by high frequency;

There is a fundamental conflict between the uncertainty of AI outputs and the certainty required by the financial industry—this is the underlying technical contradiction of Agent Payment;

The security framework for Agent Payment is shifting from identity verification to intent verification;

Chargeback mechanisms fail in Agent scenarios, and a three-layer arbitration will become the new paradigm for payment security;

Big companies’ design philosophy is distrust of Agents, only trusting the transaction;

The real bottleneck of Agent Payment is not the payment itself, but the upstream transaction processes that have yet to be rebuilt for Agents;

Startups’ role is as component suppliers for big companies, not as direct C-end service providers.

Hazel Hu

Host of the podcast 《支无不言》, core contributor to the Chinese Public Goods Fund GCC, X: withhazelhu; also known as a casual “Yue Yue” on Jike.

Ivy Zeng

Host of the podcast 《支无不言》, exploring practical use cases of Agentic Payment, focusing on Fintech growth, previously involved in VC post-investment, and responsible for 2C product regional growth at a new-type bank. X: IvyLeanIn.

Thomas Zheng

Head of Capital Markets at 支无不言, with over 6 years of experience as a primary market financing advisor, serving multiple top projects in the industry, helping to connect and foster win-win collaborations.

Insight 01

Real-world scenario—Agent Payment is already happening, but in forms different from expectations

API calls are currently the most mature on-chain scenario for Agent Payment

Analysis of on-chain data from ClawRouter (, an application using USDC payments to pay for LLM API ), shows that the API call scenario features high frequency and small amounts: as of early April 2026, about 1,400 unique addresses generated 530k transactions totaling approximately $28k. Considering the platform also offers free models, actual usage might be underestimated—the free tier alone accounts for about 1 million API calls per month.

[Image: ClawRouter official website]

Data from a payment infrastructure startup also indicates that since deploying native Agentic Payment layers last September, API call volume accounts for about half.

Quota authorization is the foundational authorization mode for Agent Payment

A2A ( Agent 2 Agent ) The unexpected success of red envelope campaigns has driven innovation and adoption of authorization mechanisms. This mode centers on quotas rather than approvals: users pre-authorize a quota to AI, which can then operate within that range independently—no need for per-transaction confirmation. “Within this scope, AI can move your money without your confirmation.”

Offline consumption has yet to take off; what’s missing isn’t payment but experience

Explorations in online and offline settlement have covered 50 million real merchants, including scenarios like booking flights, topping up mobile phones, and buying gift cards. But C-end consumer scenarios still face dual challenges: cultivating user habits and leapfrogging experience.

Experts and KOLs have distilled that Agent has mature business models

Successful cases have validated this path: famous doctors, KOLs, and others distill their expertise and content into Agents. When users can’t meet real people, they can first use an Agent. For example, a media practitioner distilled past content into an app costing 199 RMB per month, with excellent sales—whereas a 15-minute call with the person costs thousands or even tens of thousands RMB, but the Agent version costs only dozens to hundreds of RMB.

[Image: Media professional distilling past content into an app]

Transaction Agents find PMF faster than Payment Agents

Data from the crypto space shows that transaction scenarios are currently the most concentrated user demand, with business models inherently featuring take rates. Comparing to early blockchain development, those who preemptively built merchant and stablecoin scenarios during high gas fee periods, like Tron, find it hard for users to migrate even after gas prices rise.

C-end consumption scenarios have yet to be validated by real demand

The phenomenon of over a billion users using Qianwen for milk tea during the Spring Festival sparked discussion: are users engaging because of better experience, or because of a 25 RMB subsidy per order? The dialogue format limits information density; future C-to-B scenarios may require seamless dialogue via smart glasses, demanding a leap in experience.

Participants listed scenarios better addressing user pain points:

Procurement: with strict budget controls, requiring comparisons among multiple suppliers (e.g., Alibaba’s AI e-commerce Agent - Accio)

Complex tasks: wedding planning, travel bookings, and other multi-step coordination scenarios

Ticket grabbing: high-urgency needs like concert tickets

[Image: Alibaba’s AI e-commerce Agent - Accio]

Agent Payment as a new traffic entry point

From a traffic acquisition perspective, Agent Payment is akin to early SEO and short videos—representing new traffic opportunities. Those who first studied SEO, though starting small, have continuously found ways to attract early traffic. The “Jin Guyuan Dumpling House” event might be comparable to buying pizza with Bitcoin—something remembered long after.

Background story of Jin Guyuan Dumpling House skill: “On April 7, 2026, amid the popularity of OpenClaw, the owner of the dumpling house created an AI capability module called ‘Jin Guyuan Dumpling House·SKILL’. This AI skill is designed for AI Agents, not directly for humans. After installation, the AI assistant can autonomously query dish info, business hours, queue rules, and even handle online number-taking. During the winter solstice of 2025, due to excessive queues, the delivery platform’s server mistakenly flagged the store’s interface as abnormal and banned it. The owner hopes to optimize future queue experiences through AI.”

[Image: Jin Guyuan Dumpling House Meituan queueing skill]

The real start of Agent Payment has not yet arrived

From a macro perspective, discussing true Agentic Payment now might be premature. It’s like a child’s growth—from ages 1 to 5, income comes from parents, and the available quota is parentally authorized. The child has no independent intent yet.

Currently, Agent Payment is concentrated in productivity scenarios

The consensus is that real Agent Payment is currently focused on productivity:

1. API calls: for enhancing productivity via large models or API purchases

2. Enterprise scenarios: procurement and finance teams’ Agent functions

3. Vibe Coding: rapid development of demos or products

Insight 02

Identity and Authorization—AI Uncertainty vs Financial Certainty

Agent Payment security requires a four-layer framework: identity, risk control, compliance, arbitration

Payment security can be broken down into identity, risk control, and compliance. For AI payments, this framework should be followed, with arbitration added as a fourth layer for final assurance.

1. Identity layer: shifting from identity verification to intent verification

Issuing IDs for Agents, establishing credit scoring (based on Agent’s expertise, adoption, effectiveness, token price, etc.), completing identity verification. Blockchain-based traceable, verifiable decentralized DID identity systems. On this basis, traditional identity verification is transitioning to intent verification in Agent scenarios. Intent verification considers whether the Agent’s payment is reasonable, whether the behavior meets needs, aligns with the final intent, and complies with regulations.

2. Risk control layer: fundamental conflict between AI uncertainty and financial certainty

There’s an inherent contradiction: AI outputs are uncertain, conflicting with the high certainty and trial-and-error costs in finance. Real scenarios show:

• Exposure to amount recognition errors (e.g., 0.01 USDC read as 10k USDC)

• Susceptibility to manipulation (e.g., delivery descriptions claiming “cure all diseases,” leading models to order)

[Image: AI misreading 0.1 USDC as 10,000 USDC]

Simultaneously, supply chain poisoning in R&D is a new risk control challenge. Since the rise of OpenAI, poisoning in npm packages, dependencies, and other layers can occur without direct use—users depend on packages that may be poisoned. Risk control must cover identity authorization (anti-money laundering), model drift and hallucination, and execution chain attacks.

Tech giants’ philosophy is to treat all Agents as malicious by default. They aim for “verifiable transaction chains,” not necessarily “verifiable Agents.” By introducing authorization protocols (Mandate), breaking down tasks, setting constraints, and cross-checking, fraud prevention involves layered data zero-knowledge proofs, zero-trust principles, and self-verification mechanisms.

3. Compliance layer: semi-decentralized lightning networks as a good solution for micro-payments

Traditional finance and blockchain face bottlenecks under high concurrency. For Agent design, micro-payments are key. The security of micro-payments can be achieved through a balance of decentralization and centralization—lightning networks, with their high TPS, may see a renaissance in the Agentic Payment era.

4. Arbitration layer: layered arbitration replacing traditional chargebacks

The credit card chargeback mechanism in Visa networks is hard to implement in Agentic Payment. A new layered arbitration system is needed:

1. First layer: AI automatically arbitrates clear disputes (duplicate charges, incorrect amounts, service not delivered)

2. Second layer: AI arbitration teams handle judgment-required issues (service quality, authorization boundaries)

3. Third layer: human involvement for complex disputes

Insight 03

Business models—Seizing niches, re-pricing AI, risk control, and authorization

Startups are currently “powering love” to seize niches

Before the business model matures, honest startup founders say “powering love, occupying positions, waiting for the wind”—as one API platform founder described the current stage.

Transaction scenarios inherently feature take rates

Similar to early blockchain, those who preemptively built merchant and stablecoin scenarios during high gas fees, like Tron, find it hard for users to migrate after costs rise. The crypto industry’s trading scenarios naturally have take rates.

Bill aggregation is key to solving small payment inefficiencies

For card payments, transactions under $10 may be unprofitable for merchants. In Agentic Payment scenarios, small payments are frequent; the solution is bill aggregation to increase per-transaction settlement amounts.

Pay-per-result pricing only works for quantifiable, piece-rate work

Users may call one API, but results vary greatly. How to price AI services? Participants believe pay-per-result only works in simple, piece-rate tasks (e.g., customer service agents resolving tickets). In uncertain scenarios (e.g., sales agents qualifying leads), it’s highly subjective. Pay-per-result is limited to a few straightforward tasks; mainstream scenarios will likely stick to call-based or subscription models until verifiability breakthroughs occur.

Lessons from 400+ companies and 50 unicorns on pricing AI products | Madhavan Ramanujam

Vibe Coding’s commercialization hinges on subscription and usage conversion

The goal is to enable new AI companies or individual developers to quickly commercialize products built with Vibe Coding. Many independent developers can create demos easily, but turning these into sustainable business models is harder. The key is converting each large model usage into a monthly plan or a subscription plus credits.

Insight 04

Competitive landscape—Big companies’ offensive and startups’ strategies

Stablecoins are lowering the barrier for traditional card organizations

Before Stripe acquired stablecoin firm Bridge, its valuation dropped from a peak of $92 billion to below $70 billion. After the acquisition, valuation quickly rebounded to around $90 billion, with the latest funding round valuing it at $159.1 billion. Its stablecoin settlement service charges 1.5%, much lower than the 2.8–3% average fee of traditional card networks, and could even drop to 1%. In contrast, traditional payment companies’ business models are fragile (e.g., Visa relies heavily on transaction fees), and PayPal, wary of impacting its core business, hesitated in stablecoin deployment, missing scale.

Startups will become component suppliers for big companies

For a long time, the business model will likely involve big companies integrating and reselling these tools rather than individual C-end users directly invoking them. Big companies may become clients, startups as suppliers, assembling and selling tools at higher prices. This trend will inevitably increase industry centralization.

AI tax is an inevitable form of high-frequency small payments within 3–5 years

Some believe AI taxation will serve as a source of UBI and unemployment benefits, with high-frequency small AI payments becoming infrastructure. Possible tax methods include:

1. Introducing “AI penetration rate” concepts, with progressive levies based on penetration levels

2. Taxing token call volumes, akin to VAT

The real bottleneck isn’t payments but the upstream—transaction layers haven’t been rebuilt for Agents

Protocols and user wallets might solve payment issues, but the bigger problem is that transactions can’t happen yet. All payments require a prior transaction—scenarios like e-commerce or flight bookings can’t be completed via Agents. No transaction means no subsequent payment.

C-end breakout: the importance of ground promotion and startup boundaries

Why did OpenClaw suddenly become popular? It was driven domestically by ground promotion—big industry players selling cloud services and doing ground campaigns. Like early mobile payments, a key reason even seniors can use it is because of subsidies: “Install the app, I’ll teach you how, and I’ll give you 50 RMB.”

But for startups, many needs take a long time to realize. A founder of AI payment infrastructure said that after assessing this, they decided not to target user scenarios directly. They believe user education shouldn’t be borne by a few startups but by the entire industry. If the industry isn’t viable, it’s pointless; if it is, big companies should share the costs and enjoy the benefits. Conversely, they focus on abstraction—removing all industry-specific accounts, wallets, bridges, chains, and payment networks so users don’t need to understand them. Once this is clear, they know where their small team’s advantage lies and which costs to avoid.

This may be the key question all Agent Payment participants face today: not “Will Agent Payment succeed?” but “Before it succeeds, which layer are you prepared to stand on?” Protocol, wallet, identity, authorization, transaction, settlement—each layer has bets and waiting players.

Big companies are preparing to take over the entire chain; startups are preparing to be integrated into it. Those who survive are likely those who neither overestimate their ability to sustain an independent track nor underestimate their value at any particular layer.