Agentic AI and the Future of Payments

By Alex Taylor, Managing Director UK, Mangopay

The evolution of AI in commerce is moving beyond insight generation and into direct action. Agentic AI will increasingly act on users’ behalf, moving from discovering and recommending products to initiating transactions. As such, we’ll begin to see commerce shift from user-driven clicks to agent-triggered actions. When agents transact, they coordinate value flows across multiple parties involved in the transaction lifecycle. Businesses like Alibaba and Walmart have already started to integrate AI agents into their commerce functions that find products, make recommendations, and, in some cases, support in-app purchasing. As the payment journey matures, payments must support automated fund splitting, distribution, and settlement across merchants, platforms, and service providers. For platforms, agentic AI has the most immediate impact at the pay-in stage. This is where agents can interpret user intent, select payment methods, apply incentives, and initiate transactions.

Our research highlights the rising dominance of in‑app wallets – with 43% of platform users now paying through them – showing wallet infrastructures must evolve alongside new commerce models.

Agentic commerce changes transaction economics

Agentic AI will change how transactions are initiated, shifting compensation models towards outcomes and performance rather than clicks, impressions or sessions. This has direct implications for platform economics.

A single agent-triggered transaction may need to distribute funds to multiple parties, including merchants, platforms, recommender agents, logistics, and service providers. Platforms and marketplaces must reconsider revenue sharing, acquisition costs, and pricing models to accommodate agent incentives.

Agentic AI can influence incentives and transaction logic, but the execution of value splitting, conditional payouts, and settlement is best handled by programmable, wallet-first payment infrastructures. These capabilities already exist and provide the control and predictability required for multi-party payment flows; and they’re becoming core requirements.

Trust and control in in-app wallets

As agents begin acting autonomously, wallets and platforms must ensure decisions remain aligned with user intent and protected against misuse. Multi-party payments raise the risk profile, as funds may be split, held or released automatically without human intervention.

Control mechanisms such as permissions, authorisation boundaries and agent verification processes will be crucial. Trust will become a key differentiator as agent-triggered payments scale. Platforms that cannot explain or control how value moves will struggle to maintain user confidence.

Payment infrastructure does not need rebuilding. Inputs do

Existing payment rails remain fit for purpose. The key change is in the data that feeds into those rails.

Agents require machine-readable product catalogues, accurate pricing and availability data, and clear business rules and constraints. They also require explicit multi-party payment logic defining fund allocation, settlement timing and conditional release.

The shift to agent-triggered actions demands stronger data discipline. Without structured and reliable inputs, agents may make incorrect assumptions, transactions may fail or execute incorrectly, and user experience and trust will suffer.

Operational reality still matters

While agentic AI introduces new capabilities, the main pain points in payments remain operational. Regulatory compliance, onboarding, reconciliation and dispute management continue to dominate platform complexity. Agentic AI can improve user experience, but its adoption should be assessed carefully against whether it addresses real priorities rather than adding novelty without operational value.

What FinTech leaders should prepare for now

Agentic AI introduces a permanent new stakeholder in commerce and payments. Transaction economics, commercial models and trust frameworks will evolve to accommodate outcome-driven incentives and agent-triggered payments.

Leaders should focus on practical preparation. This includes making catalogues, pricing and business rules agent-readable and reliably updated; defining governance boundaries, authorisation flows and recovery steps; deploying real-time fraud detection and monitoring, such as Mangopay’s AI fraud product; and creating cross-functional teams across product, payments, legal and compliance.

Engaging with emerging protocols and standards from payment networks will also be essential to ensure agent-initiated payments can be authorised, executed, and reconciled across merchants and payment rails.

Preparing for the future

Agentic AI is driving a rapid shift in how value flows through digital commerce, redefining payments, wallet infrastructures and platform economics. Payments move from single-destination transfers to orchestrated, programmable value distribution.

PSPs with a wallet-first approach, flexible multi-party flow control and strong platform risk tools are well positioned to support this transition. Platforms that act now can pave the way into the agentic era, ensuring smooth, secure and responsible commerce from 2026 onwards.