As artificial intelligence (AI) continues to be a key pillar in reshaping industries, a new threshold is materialising in the form of Agentic AI, a system capable of autonomous decision-making and goal-directed behaviour. When it comes to the ever-evolving fintech space, this technology is proving to be more than just a catchword, as it supports the automation of complex financial processes, scales customer personalisation, and manages risk in real time.
In this article, we will explore Agentic AI’s influence on the financial services sector, focusing on key players and how they leverage the technology, the opportunities it showcases for innovation, how regulatory bodies plan to govern it, and the upcoming trends.
By utilising sophisticated reasoning and iterative planning, Agentic AI can autonomously solve complex, multi-step processes. This technology includes AI agents and machine learning (ML) models that mirror human decision-making to address problems in real time, with each agent performing a specific subtask necessary to achieve an objective. These efforts are synchronised through AI orchestration, which coordinates and manages AI models, systems, and integrations, encompassing the deployment, implementation, integration, and maintenance of the entire AI system, workflow, or app. Compared to traditional models, which operate within predefined limitations and require human intervention, Agentic AI demonstrates independence, objective-driven behaviour, and adaptability.
But what is the exact process behind it? Agentic AI tools can come in several forms, with different frameworks being suitable for specific problems. However, the broad steps that agentic systems take to conduct their operations include:
Perception: Agentic AI gathers data from its environment via sensors, APIs, databases, and user interactions, allowing it to hold up-to-date information to analyse and work upon.
Reasoning: the data is being processed to extract insights. By harnessing natural language processing (NLP), computer vision, and other AI features, the data is interpreted within the broader context.
Goal setting: AI puts in place objectives based on predefined targets or user inputs. Afterwards, it comes up with a strategy to achieve these goals, usually by leveraging decision trees or reinforcement learning.
Execution: after selecting an action, AI interacts with external systems or offers responses to users to carry out the chosen action.
Learning and adaptation: after performing an action, AI assesses its outcome and collects feedback to optimise future decisions.
Orchestration: platforms allow for the automation of AI workflows, tracking of progress for task completion, managing resource usage, monitoring data flow, and handling failure events.
We have not even hit the halfway point of 2025, and several news stories have already broken about how the financial services sector is leveraging AI to power the future of fintech. Some are focusing on GenAI, while others are taking it a step further, diving into Agentic AI to discover its capabilities and use cases for augmenting the ecosystem. This technology has the potential to disrupt the finance sector by optimising how data is processed, how accurate decisions are made, and even how customer interactions are managed.
NVIDIA and Oracle opened the way, teaming up in March 2025 to support enterprises in utilising Agentic AI more efficiently and cost-effectively. Merging NVIDIA’s accelerated computing and interference software with Oracle’s AI infrastructure and GenAI services was set to enable businesses to accelerate and improve the development of Agentic AI applications. Also, the integration between Oracle Cloud Infrastructure (OCI) and the NVIDIA AI Enterprise software platform was set to make over 160 AI tools available through the OCI Console.
Fast forward to the beginning of April 2025, Google Cloud made a series of announcements regarding its partnerships with industry players as it aimed to advance AI globally. The company extended its collaborations with Deloitte, Capgemini, and Accenture, centring on expanding Agentic AI capabilities. More specifically, Google Cloud worked with Deloitte on launching a suite of 100 ready-to-deploy agents, facilitated by its Gemini models and Agentspace, seeking to enhance customer interactions and employee experiences. When it comes to Capgemini, the firm entered a strategic Agentic AI deal with Google Cloud to improve customer experiences and scale value for clients, while Accenture planned to roll out new features to assist organisations in increasing cloud and AI technologies.
Soon after, PayPal made its move, launching its Agent Toolkit to facilitate the integration of its suite of APIs, including those for managing payments, invoices, disputes, shipment tracking, catalogue, subscriptions, reporting, and insights, onto various AI frameworks for developers. Through the PayPal Agent Toolkit, developers were given the ability to build advanced agentic workflows that managed financial operations efficiently. Two weeks later, PayPal announced its plans to equip developers and merchants with tools to create modern shopping experiences. The company allowed developers to support Agentic AI experiences that assist customers in paying, tracking shipments, and managing invoices, driven by PayPal and within an AI agent.
Mastercard took the stage at the end of April 2025, rolling out Agent Pay, its Agentic Payments Program, a solution that aims to offer more secure and personalised payment experiences to consumers, merchants, and issuers. Also, by launching the Agentic Payments Program, Mastercard intended to deliver Mastercard Agentic Tokens, building upon tokenization features able to support global commerce solutions.
Visa made its entrance shortly after by unveiling Visa Intelligent Commerce, a new initiative focusing on allowing secure and simplified transactions in the AI-driven retail landscape. The company designed the platform to integrate its global payment network with the next generation of AI agents, aiming to redefine how consumers shop and pay. When we reached out for a comment on the launch, Mathieu Altwegg, SVP of Products and Services at Visa Europe, highlighted that ‘Agentic AI will transform commerce. AI agents will be able to deliver tailored, personalised recommendations for consumers within seconds, taking the heavy lifting out of organising a family holiday, securing must-have concert tickets, or finding the perfect outfit. This isn’t science fiction: we’re already seeing people use AI to help make purchasing decisions, with 92% of those who do saying it has enhanced their experience. And we have seen traffic moving directly from AI sites to retail sites grow by 1200% in the last two and a half years.
However, AI-enabled commerce without payments isn't commerce: it's just window shopping. And without seamless, secure, trusted payment options, AI commerce cannot reach its true potential. That is why we are taking the power of the Visa network and our decades-long expertise to bring trust and security to AI-driven ecommerce. We’re working with leading AI innovators (including OpenAI, Anthropic, Mistral AI, and many more) to embed AI-powered payments into the agentic experience. Visa Intelligent Commerce will ensure consumers can complete their AI-enhanced transactions seamlessly and safely, with their purchases subject to the same secure infrastructure, standards, and capabilities they currently enjoy with both physical and digital shopping.
Crucially, consumers will be able to set and control the boundaries and conditions, so they always know exactly what their AI agent can and cannot do with their payment credentials, maintaining trust and confidence in both AI commerce and the payments underpinning them.’
With the rise of artificial intelligence in the financial sector, regulatory bodies globally are actively working on developing frameworks to govern the use of both the technology as a whole and its subsets, including Agentic AI. These regulations focus on maintaining an equilibrium between innovation and security, transparency, and accountability, intending to address concerns over systemic risks, consumer safety, and ethical principles.
Since AI started gaining momentum, the European Union has introduced a variety of key regulations to mitigate risks and assist advancement. Among them is the Artificial Intelligence Act (AI Act), effective as of August 2024, with it placing itself as one of the first comprehensive attempts to regulate AI across various industries. In a recent opinion piece, our colleague Irina Ionescu underlined what the act focuses on, how the technology can be leveraged to deceive the public or cause harm to individuals, including specific provisions for the use of AI in content generation, notably in sections such as deepfakes. Additionally, the European Union established the Digital Operational Resilience Act (DORA), mandating that financial entities improve their digital resilience, including risks associated with AI systems.
Furthermore, the UK has chosen to adopt a more risk-based approach, creating the AI Safety Institute (AISI), which centres its efforts on evaluating the safety of advanced AI models and working together with global counterparts to establish worldwide standards for AI security. Moreover, in April 2025, the Financial Conduct Authority (FCA) announced its plans to roll out a live AI testing service to support companies in launching responsible AI models. The new service, which came as an addition to the FCA’s AI Lab, was set to fill the testing gap that posed significant challenges to firms adopting the technology.
When it comes to the US, regulatory bodies decided on a different method, focusing on introducing several AI bills, acts, and guiding principles at the federal and state levels. President Donald Trump’s second term, however, reshaped the industry, with many regulations established during President Joe Biden’s administration being revoked. Currently, the Securities and Exchange Commission’s focal point is the impact of AI on market integrity, while the Federal Reserve and the Office of the Comptroller of the Currency (OCC) are analysing the implications of AI on financial stability and are looking into guidelines specific to the banking sector.
As AI, and more specifically, Agentic AI, is on a growth trajectory, regulators are acclimating in hopes of ensuring that its integration into the financial services sector is as transparent and secure as possible.
With AI expanding beyond traditional automation models, the development of Agentic AI is expected to reshape the fintech sector. Compared to other tools that leverage predefined inputs, Agentic AI adjusts its strategies in real time to augment outcomes across financial systems.
But what can we expect from Agentic AI, and will the industry be ready for this technology? When we reached out for a comment on the topic, Panagiotis Kriaris, Director – Head of Business & Corporate Development at Unzer, underlined that ‘Agentic AI is quickly redefining how financial services are built, accessed, and monetised. Three trends are driving this shift.
First, the interface is disappearing: payments, onboarding, fraud checks - they’re moving into conversations, prompts, and background automation. Agents initiate payments automatically, based on context and permissions - no user action required.
Second, the infrastructure race is on. PayPal launched agent-native APIs for payments, shipping, invoicing, and risk. Mastercard’s Agent Pay enables bots to transact securely with tokenization and biometric SCA. Visa is building modular agent APIs for real-time approvals and spend controls. Stripe is embedding intelligence into payment flows, enabling agents to act on transactions contextually and in real time.
Third, the stack is being standardised. Protocols like MCP and A2A are enabling agents to talk across systems and to each other. The winners won’t just process payments - they’ll orchestrate entire workflows.
What’s at stake? Becoming the new infrastructure.
In the agentic economy, controlling the payment credential means becoming the default wallet in the loop, capturing the transaction, associated data, interchange, and value-added services.
Players who build for this will not only control the new checkout but will also own the customer relationship. Those who don’t risk finding themselves disintermediated by someone else’s agent.’
Agentic AI has the potential to revolutionise finance, but will fintech players keep up while also ensuring that this technology aligns with security standards and customer expectations? It remains to be seen, and we are here to keep you in the loop.
Every day we send out a free e-mail with the most important headlines of the last 24 hours.
Subscribe now