The era of AI shopping is approaching, but the payment link is a key obstacle. Companies such as Nekuda are exploring the field of agent payment, focusing on solving the problem of “cardless and unmanned” automated transactions through technologies such as smart agent wallets and contextual authorization mechanisms, building trust infrastructure, and promoting the development of AI e-commerce.
The day for AI to buy things is getting closer and closer.
According to the latest data from Visa, the traffic to retail websites from AI agents has increased by 1,200% year-on-year. From ChatGPT Search’s intelligent recommendations to Amazon’s cross-platform shopping assistant, Agent e-commerce seems to be within reach.
But in this process, there is one link that cannot be bypassed: payment. To put it simply, the existing financial system cannot adapt to the automated transaction of “no card and no one”.
As a result, Agent’s payment link has become a highland for AI startups to compete, and many companies have received financing.
Among them, the company Nekuda is particularly dazzling. Just in May this year, it received an investment of $5 million. Among investors, there are payment giants such as Visa and American Express, as well as Madrona Ventures, a top international venture capital institution.
Nekuda mainly relies on two major technologies, intelligent agent wallet and contextual authorization mechanism, to solve the two major problems of non-probation certificate management, from “passive execution” to “active compliance”.
B-end product manager’s ability model and learning improvement
The first challenge faced by B-end product managers is how to correctly analyze and diagnose business problems. This is also the most difficult part, product design knowledge is basically not helpful for this part of the work, if you want to do a good job in business analysis and diagnosis, you must have a solid …
View details >
Next, let’s take a look at the latest explorations of these AI startups in the field of agent payments.
01 The era of AI buying things by itself has arrived
You only need to say to the AI: “Help me buy a pair of hiking shoes, with a budget of less than 800 yuan”, and the AI will automatically search, compare, place an order, and pay, and the whole process is completely automatic.
This sounds like a science fiction movie scene, but it’s actually happening – major AI giants have laid out the Agent e-commerce field:
- OpenAI’s ChatGPT Search can generate shopping recommendations based on needs, providing product images, reviews, and purchase links.
- The “Perplexity Shop” launched by Perplexity supports problem retrieval and photo recognition, and users can purchase with one click;
- Amazon’s Agent shopping function can display products across platforms, and AI automatically processes payment and order tracking;
While these features are highly anticipated, they also have a key problem that remains unresolved: the AI shopping process often ends abruptly after jumping to the checkout page, and users still need to manually complete the payment.
Agents seem to be able to autonomously complete the entire process of searching, price comparison, and purchase, but agents are still essentially “robots”, and the existing financial system is not designed for the “cardless and unmanned” agent process, so agents will encounter technical, compliance, and authorization obstacles when paying.
So, how can the payment process be truly automated?
There is a company called Nekuda that has specially developed an Agent payment system, which helps developers easily build an Agent transaction system and lower the technical threshold for Agents to enter the field of commercial shopping.
If the traditional payment process is based on the security model of “users actively enter passwords/verification codes”, then agent payment needs to achieve a closed loop of “automatic authorization between systems”.
To achieve this, Nekuda has built authentication, security mechanisms, and compliance frameworks into its SDK to allow agents to complete transactions without having to modify existing payment systems.
It mainly relies on two core designs:
(1) Smart Agent Wallet: No probation certificate management
This technology mainly solves the problem of voucher management. Users can entrust payment vouchers to AI Agent, which automatically stores and fills in payment information at checkout, and complies with payment card security standards (PCI DSS), enabling autonomous transaction completion without human input. The payment voucher is encrypted in transmission and storage, and the user can control the delegation permission (e.g., “Only one secondary card payment is allowed”).
(2) Contextual authorization mechanism: from “passive execution” to “active compliance”
Nekuda’s most important innovation is the Agent Mandates system, which captures various types of information about users’ purchase intentions, including “what agents are allowed to buy”, “under what conditions”, “how much to spend”, etc.
By recording complete context information and encrypting signatures, the system forms a verifiable authorization document, clarifying the responsibilities of all parties in the entire payment chain, so that payment networks, banks, merchants, etc. can be confident that the transaction is truly authorized by the user.
Today, Nekuda has partnered with Visa to study allowing agents to make purchases on behalf of consumers.
02 Seize the trust infrastructure highland of Agent payment
From web to mobile, from social to cloud computing, every technological paradigm change will give rise to new business rules, and AI Agent is reshaping the underlying logic of payments and transactions in a similar trend.
In the exploration route of agentization in the payment industry, some AI companies have already achieved the transition from tools to infrastructure.
In addition to Nekuda, payment giants have taken the lead in deploying agent automation scenarios: PayPal launched the Agentic Toolkit toolkit to automate order and invoice management; Stripe launched the Agent’s API, which allows the Agent to complete financial payments using one-time virtual cards, and the large model calls APIs such as payment, billing, and card issuance.
At the same time, AI startups have also set off a funding boom in this field, with some companies receiving a large investment of $20 million:
It is worth noting that although the technical threshold for Agent construction is gradually lowering (now users can accept AI ghostwritten emails and analyze data), humans are still highly vigilant about “autonomous decision-making” in payment scenarios. When it comes to billing management and authorization permissions, users are more inclined to take control.
This contradiction reveals a key problem: the leap from “tool calling” to “autonomous transaction” essentially needs to break through the bottleneck of trust mechanism and authority system.
For letting AI buy things, only by building trust can agents truly undertake valuable tasks such as purchasing, booking, and trading.
Unlike PayPal and Stripe, which focus on single-function automation, Nekuda not only has more “fine control” in process automation, but also shifts the focus of competition to the construction of “trust infrastructure”.
Its authorization system solves the key problem of “authorization” in agent payment, and also solves the three problems of consumer trust, developer trust, and website trust:
(1) In terms of consumers, form a transparent and controllable instruction closed loop. Users can set clear instructions for the agent, such as “automatic purchase under $500”, “only book Delta 446 flights”, etc., and the agent will not cross the line;
(2) For developers, create a responsibility mechanism for risk isolation. Developers do not need to worry about being misjudged as fraud because of the Agent’s operation, such as “the agent mistakenly ordered 10,000 sneakers”, and the developer will not be held accountable;
(3) In terms of websites, establish a compatible traditional financial system. The payment system gains the underlying ability to identify and process agent transactions.
True AI autonomous transactions will upgrade the payment process from “confirmation instructions” to “intelligent decision-making”, and promote the transformation of agent business companies from “transaction processors” to “business facilitators”.
This process requires not only smarter agents, but also new trust mechanisms, permission systems, and payment infrastructure. The demand for “refined management of permissions” derived from it may be the core incision of the next technological breakpoint.
