Payment leaders gathered on Thursday, October 9, 2025 for the Payments Exchange Roundtable, an intimate discussion designed to delve into the transformative impact of generative AI and agentic AI on the payment ecosystem in Canada.

Co-hosted by Cognizant and The Payments Canada SUMMIT, the roundtable built on the success of last year’s discussion on real-time payments. The evening, held at the Canoe Restaurant in Toronto, began with insights from one of the world’s leading minds in artificial intelligence, Babak Hodjat, Chief AI Officer Cognizant. Mr. Hodjat opened the discussion with insights drawn from his deep experience in leading AI-driven transformations, including his foundational work as the primary inventor of the agent-oriented technology behind Apple’s Siri.

The discussion provided an invaluable forum for payment leaders to explore how these advanced AI systems will shape efficiency and security across financial services. The practical application of agentic AI within payments specifically is an area with a limited number of formally established use cases and much of the discussion was dedicated to exploring that potential. The consensus of the evening was a clear affirmation that organizations embracing agentic AI will gain a significant advantage in an evolving market.

There’s no escaping it: AI dominates today’s business conversations, and the payments industry is no exception. Generative AI and agentic AI are moving from buzzwords to strategic imperatives for growth and efficiency.

To separate reality from hype, Cognizant and The Payments Canada SUMMIT gathered a cohort of payment leaders for an exchange on the promise of AI and the practical implications of adopting AI-enabled solutions to solve today’s complex payment challenges.

Over the course of the evening, the conversation examined the potential of agentic AI to deliver autonomous and embedded payments, and its ability to deliver hyper-personalized engagement experiences for direct-to-consumer applications. Of particular interest were the topics of trust and human engagement in payments – and the potential of agentic AI to irretrievably alter both – for the better, if done right; else not.

The evening concluded with the assembled group aligned on three key views:

• Multi-agent systems have the power to significantly improve payment solutions by introducing autonomy, scalability, and intelligence into payment workflows
• That power needs to be harnessed and effectively channeled via responsible governance to deliver trust and foster adoption
• The potential for agent-based payments systems to altogether eliminate human involvement/intervention needs to be handled with sensitivity

On behalf of Cognizant, we thank The Payments Canada SUMMIT for partnering with us to plan and execute this session, and to the assembled payments leaders for their diverse and enriching viewpoints. The recap below distills the key themes, real-world use cases and benefits, and the impact agentic AI will have on the future of payments for those who embrace change and are ready to innovate securely, responsibly, and in alignment with regulatory expectations.

Artificial intelligence is rapidly transforming the payment landscape and multi-agent AI systems are at the forefront of this revolution. These sophisticated systems, which involve multiple AI agents collaborating to achieve complex tasks, are proving to be game-changers in driving efficiency.

At its core, an AI agent is a large language model (LLM) paired with code that enables it to perform actions. While LLMs are powerful at processing and generating data, the added code allows agents to translate intent into action through APIs. A multi-agent system takes this a step further, creating a "team" of these agents, each with a specific scope, to tackle intricate problems that would be challenging or impossible for a single agent or even a human team to handle.

Cognizant, a leader in AI innovation, has been instrumental in developing and implementing these advanced systems. Mr. Hodjat and his team also champion AI for Good initiatives, aligning their work with the UN's Sustainable Development Goals.

One compelling example of their multi-agent system in action was its use in evaluating 32,000 hackathon submissions. A task that would have taken a year and nine full-time employees was completed in just 24 hours by a multi-agent system.

Real-world applications and benefits

The applications of agentic AI extend across various use cases:

• Network monitoring: Agents deployed on network nodes can proactively monitor and manage severe incidents, ensuring quick response times.
• Consumer decision systems: Agent networks can make decisions on behalf of consumers, interacting with various merchant agents to optimize outcomes.
• Financial systems: Cognizant's open-source multi-agent accelerator, Neuro SAN is designed for scalable, interoperable financial systems, and enables the creation of intelligent, purpose-built agent networks where multiple AI agents collaborate to solve complex tasks. These systems are particularly applicable in payment systems for complex, uncertain scenarios where human judgment or autonomy is currently relied upon.

These systems offer significant benefits, including:

• Lower costs: Automating complex tasks reduces operational expenses.
• Increased efficiency: Tasks are completed much faster than with traditional methods.
• Improved quality: Multi-agent systems can often achieve higher levels of accuracy and thoroughness.
• Interoperability and future-proofing: Agents can communicate seamlessly, even if provisioned independently, allowing for expanded capabilities without re-engineering existing components.

Addressing challenges and ensuring reliability

While the potential is immense, challenges emerge in ensuring system reliability, particularly in heavily regulated industries such as payments. Solutions include:

• Semantic density: A method to assess an agent's confidence in its responses.
• Guardrails: Approaches like confidence thresholds for sensitive tasks, custom-built rules, and continuous fine-tuning of LLMs to correct errors.
• Fraud protection and security: Code-level authorization and authentication prevent unauthorized actions, and LLMs can identify unreasonable or fraudulent requests.
• Balancing efficiency and security: Multi-agent systems can improve both by restricting agent domains and allowing agents to check each other's work.

Impact on the future of payments

As Mr. Hodjat emphasized, agentic AI is poised to permeate every aspect of business operations. Organizations that embrace agentic AI to drive down costs and address inefficiencies in internal processes, risk management and governance will gain a significant advantage. The market is evolving and smarter, faster and more efficient systems will undoubtedly prevail as agents collaborate seamlessly to deliver increasingly complex services.

The Payments Canada SUMMIT would like to thank Cognizant for their partnership in developing and co-hosting the roundtable discussion series. Interested in joining future roundtable discussions? Email conference@payments.ca for more information.