AI agents are evolving from generative models into autonomous systems capable of planning and executing multi-step trade transactions with minimal human intervention.
The shift towards continuous, granular transactions may require trade finance to move away from traditional document-based systems towards real-time, programmable data models.
South-East Asia is positioned as a primary testbed for these innovations due to its integrated supply chains, active fintech ecosystems, and regional digital economy agreements.

The first few months of 2026 have seen strides towards agentic commerce and agentic payments.

Sea Ltd announced its collaboration with Google to advance agentic commerce and agentic payments. Sea is the operator of Shopee, the largest e-commerce platform in Southeast Asia, and of Monee, the digital financial services platform that includes Maribank in Singapore and SeaBank in Indonesia and the Philippines. 

A few days later, Alibaba released Qwen 3.5 in China for trials, which it said could, from a single text or voice request, order food, complete in-chat payments, plan and book travel, amongst other things.

These come after trials in 2025: about a year ago, Amazon announced the testing of a beta feature for its US customers to use an AI agent in the Amazon Shopping app to discover items, order, checkout, and pay for products from various sellers. At about the same time, Visa and Mastercard both launched their AI agent trials (Visa Intelligent Commerce and Mastercard Agent Pay, respectively). These trials began in the US, but their expansion to Asian markets like India, Singapore, and Malaysia is nascent.

Agentic artificial intelligence (AI) builds on generative AI, using the latter as its reasoning and language engine to interpret instructions and act thereon – AI agents are software systems built on large language models (LLMs). AI agents can plan, decide, and execute multi-step tasks autonomously, and interact with other systems to complete actions with real-world effects. 

In agentic commerce, AI agents would act on behalf of buyers and sellers to execute transactions. This means the AI agents would autonomously manage the procure-to-pay and order-to-cash cycles, with minimal or no human intervention. 

The publicly known pilots or trials of agentic commerce are largely in the business-to-consumer (B2C) space, with the technology being offered by businesses (the e-commerce platform, the payment technology provider) to the consumer. These are arguably lower-friction environments compared to the business-to-business (B2B) space, which, whilst not the low-hanging target, will undoubtedly be a frontier for agentic commerce.

Agentic commerce in B2B payments and trade: MRO, inventory management

One example pointing in the direction of B2B agentic commerce is the growth of maintenance, repair, and operations (MRO) supply chain management vendors such as Eezee, a Singapore-headquartered platform active across multiple South-East Asian countries. They act as procurement agents for many large corporates, managing the long tail of low-value, high-volume MRO purchases, thereby freeing these corporates to focus on their core business’ procurement requirements. 

This is an extant model that illustrates how intermediaries perform agent-like functions in B2B procurement, a role that agentic AI could potentially be used for by the corporates or by the procurement agents themselves.

Agentic commerce could significantly transform inventory management via automated replenishment of stock levels monitored by AI agents. More than just stock taking: they would also act based on demand signals and supply conditions and options, on a continuous basis. If realised, agentic commerce creates more robust management of supply chains and more frequent actions.

If executed by AI agents, the velocity of commercial transactions would increase dramatically. Machines can search, compare, decide and transact in seconds rather than hours or days, and perform 24/7. 

Human handling tends to aggregate demand into larger batches because procurement, negotiation, and approval processes are slower and costly; AI, by contrast, can execute granular transactions continuously, shifting purchases from periodic bulk orders toward more real-time, automated inventory management.

Such a shift to agentic commerce has implications for payments. If transactions become faster and more granular, would payment flows also follow suit, moving from periodic batch settlements to more frequent and smaller transfers? The bigger question would be the need for or use of programmable settlements, with authorisation, limit amounts, identity of payer and payee controls for safe execution.

How would trade finance work for agentic commerce? 

If programmable settlements mean that payments are automatically made once conditions are met, the funds for the payment must be available in anticipation of when the payment must be made. This goes counter to the practice of supplier credit, where the supplier extends credit terms (of, say, 30/ 60/ 90 days) to the buyer. Would supplier credit make sense if the payment is prefunded?

Supplier credit is often required by buyers. It fulfils the buyers’ working capital needs, but also gives time for goods receipt, inspection, and dispute resolution. It is a design challenge for programmable payment logic to accommodate such commercial considerations and enforce settlements. If payment is dependent on the availability of funding at the time it is needed, the supplier takes counterparty risk on the buyer.

It is when there is a mismatch between the timing of delivery (including billing) and payment that financing requirements arise. A supplier requires financing when it extends credit to the buyer, and a buyer may need financing to pay the supplier if it needs payment terms longer than those provided by the supplier.

If purchases are executed continuously and in small increments, traditional financing arrangements built on larger orders and invoices may need to adapt to more dynamic financing models executed potentially on data rather than documents. Trade and supply chain finance may need redesigns to become embedded in the commerce and payment workflows. Embedded finance is a delivery model that operates in real-time, using platform data for transactional underwriting and disbursements. 

Banks’ traditional trade finance booking systems, usually designed around bulkier transactions, manual approvals, and document-based financing, may prove sub-optimal in supporting such higher frequency granular financing. Even supply chain finance systems, which are designed to process voluminous invoicing, may need redesign to manage financing agentic commerce transactions.

Could the financing itself be agentic? 

It is conceivable that financing could employ AI to trigger funding decisions within programmed parameters, making trade and supply chain finance a programmable service to support commercial transactions. 

AI agents representing buyers, sellers, logistics providers, and financiers could communicate with one another machine-to-machine, negotiating the conditions to trigger orders, payments, and financing.

The weighty questions of risk management, accountability, and governance would need to be addressed. The initiatives on agentic payment are at a nascent stage, and agentic financing is even more incipient. 

It is conceivable that agentic financing will be deployed in due time. Getting there may require a period where AI agents perform certain operational tasks with humans-in-the-loop, until such time when sufficient safeguards, reliability, and regulations have been developed to allow agentic AI to perform financing autonomously.

—

South-East Asia could prove to be a natural testbed for these developments. The region benefits from fast-growing e-commerce platforms, high small and medium-sized enterprise (SME) participation and active fintech ecosystems, alongside supply chains that are deeply integrated into global trade. 

Governments and regulators in numerous South-East Asian economies are champions of digital trade and finance. The Association of Southeast Asian Nations (ASEAN) Digital Economy Framework Agreement (DEFA), expected to be signed in 2026, holds the distinction of being the world’s first major region-wide legally binding instrument to accelerate a region’s transformation into a leading digital economy with digital integration across member states. 

This combination, along with recognition of SME financing gaps, creates conducive conditions where agentic commerce, agentic payments, and agentic financing could be trialled and refined. South-East Asia could well be one of the first regions where agentic trade finance is introduced and scaled.