Alicia Ngomo Fernandez
VP Head of Visa Consulting & Analytics, UKI, Visa
Today, generative AI is already transforming the way we live and do business through a multitude of use cases, from human-like chatbots to improving productivity. Next, we are moving to AI-powered agents that make decisions and even transact autonomously on our behalf.
Over the past few years, AI has become a focal point for investors, businesses, regulators and the general public, thanks to the rapid adoption and mainstream application of transformer-based models like ChatGPT. As more powerful models and applications emerge, AI is set to revolutionise commerce by enabling agentic transactions, where AI autonomously makes decisions and transacts on our behalf.
Currently, AI apps and agents support the user to the recommendation or listing stage when looking to buy goods or services through their interface. We are witnessing the first steps into enabling autonomous checkouts. Visa Intelligence Commerce is at the forefront of this transformation, providing developers with tools to create personalised and secure shopping experiences, including the checkout journey.
How AI agents can reshape commerce
Most agents are currently deployed to execute tasks related to internal operations. However, the trend is shifting towards using agentic AI in customer-facing applications, especially in commerce. While purchase recommendations are common, autonomous checkouts are still rare. Existing payment infrastructure, standards, rules, compliance and customer journeys will need to adapt to enable these new autonomous commerce and payment experiences, as most current experiences and processes are designed as human-centric systems.
Agents will require set limits at the workflow or agent level, explicit permissioning and a robust identity framework. The level of autonomy of these systems in purchasing on our behalf will evolve alongside customer trust, orchestrating models’ intelligence and ecosystem enablement.
Maintaining the right balance
between AI automation and
human-in-the-loop will be key for a
secure AI-powered commerce experience.
Phases of agent-powered commerce
At Visa Consulting and Analytics, we foresee three potential distinct phases in the adoption of agentic commerce. In the first phase, ‘AI Initiates,’ AI interacts with ecommerce sites and narrows on relevant personalised recommendations with the potential to initiate a checkout process. However, purchases need to be actively confirmed by the user and payments authorised on a one-by-one basis.
In the second phase, ‘AI Transacts,’ AI agents will be able to purchase on our behalf and execute payments for low-risk transactions. Payers will grant limited autonomy to these systems by setting limits on transaction value, count and vetted merchants and categories, plus apps.
In the final phase, ‘AI Orchestrates,’ advanced agents will manage complex workflows with little human intervention. These agents will execute payments with mid to high levels of autonomy under continuous automated monitoring and audit mechanisms. AI agents will require an assigned budget; access to payment credentials; and an identification system. We are currently in the ‘AI initiates’ phase.
Architecting the future of AI-powered commerce
We strongly believe that maintaining the right balance between AI automation and human-in-the-loop will be key for a secure AI-powered commerce experience. User control, even in an autonomous environment, is key.
We are working to architect the future securely with Visa Intelligent Commerce. We are leveraging our 30-year experience in AI to fight fraud and risk, as well as being at the forefront of secure ecommerce through tokenisation.
Visa Consulting and Analytics collaborates with partners (issuers, acquirers, PSPs and merchants) to envision, plan and implement innovative experiences for the next wave of commerce. Examples include a marketing agent that monitors keyword pricing and purchases ad campaigns autonomously; a logistics agent that autonomously schedules maintenance and arranges payments to the auto-shop ensuring minimum downtime; an AI-powered 3D design software that buys 3D assets by itself to achieve a set goal; or a concierge agent that plans complete birthday parties based on themes and family preferences.
The future of AI is exciting, transformational, AI-powered, but human-centric — and is already here.
