The Agentic AI shift – intelligence that acts

Artificial intelligence has already revolutionised how we search for information, generate content, and automate processes. But we’re now stepping into something far more transformative: a world where AI doesn’t just assist, it acts. This is the era of Agentic AI, where intelligent systems not only understand complex tasks but also carry them out on our behalf. Instead of stopping at suggestions or summaries, these agents take the next step: booking the trip, filing the report, reordering the shoes, all while you can do something completely different. They operate independently, securely, and with a genuine awareness of context, marking a significant leap in how we interact with technology. 

At Money20/20, executives from Mastercard, Microsoft, ING, BridgeWise, and Airwallex provided a detailed look at how Agentic AI is transitioning from a cutting-edge concept to a business-critical reality. Their collective insights signal a fundamental reimagining of how consumers interact with commerce, how financial services are delivered, and how trust will be maintained in a world where decisions are increasingly machine-driven. 

From assistant to agent: what Agentic AI means 

While the term “agentic” is often used interchangeably with AI in general, Bahadir Yilmaz, Chief Technology Officer at ING, emphasised the need to clarify what qualifies as Agentic AI. As he explained, true Agentic AI goes beyond chatbots or productivity enhancers. It is about giving large language models (LLMs) access to tools, APIs, and user context so they can complete a task autonomously, whether that task involves answering investment questions, booking travel, or submitting a loan application. 

In this framework, an agent doesn’t just respond to input; it takes initiative, draws conclusions, and interacts with systems on behalf of the user. This marks a significant departure from earlier AI models, which operated within narrow boundaries and left the final step to the human user. 

The technologies enabling autonomous AI 

According to Greg Ulrich, Chief AI and Data Officer at Mastercard, the rapid emergence of Agentic AI is powered by the convergence of three critical innovations: advanced reasoning models, multimodal capability, and user-level interface control. 

Reasoning models are increasingly able to mimic human-like logic, breaking down problems step-by-step and acknowledging limitations. This makes them not only more accurate but also more trustworthy.  

Multimodal capability allows agents to combine text, images, video, and voice in a single context window, a leap forward in comprehension. If a user is shopping online, an AI agent doesn’t just parse a product description; it also recognises the photo of a blue blouse, understands what that means, and factors it into its output. 

The final breakthrough – what Mastercard calls “computer use” – refers to models that can interact with software the same way a human would, using a cursor or making selections. This enables agents to perform tasks across applications, like editing a document, filling out a form, or scheduling an appointment. 

Together, these capabilities give rise to truly autonomous agents that do more than answer; they act. According to Martin Moeller, Head of Artificial Intelligence & Generative AI for Financial Services at Microsoft, the rise of Agentic AI is not just a technological evolution but a behavioural shift. Users are already embracing AI systems as companions, seeking not just functional support but also life advice, task execution, and emotional insight. In this environment, businesses must fundamentally rethink how they design services. 

“If that’s the way we live our daily lives,” said Martin Moeller, “then of course, it changes a lot of the design principles for any company that wants to interact with us.” Drawing parallels to the way digitalisation replaced physical media, he noted that consumers no longer want to shop for individual products; they want to shop for outcomes, journeys, and experiences. Agentic AI is becoming the primary interface through which those experiences are delivered.  

Real-world applications across financial services 

The theory behind Agentic AI is compelling, but its practical implementation is already underway in various corners of the financial ecosystem. 

At ING, several key applications are in development. In mortgages, the bank is exploring how AI can simplify applications, offer personalised recommendations, and reduce processing time. In investment advisory, the bank is working on scalable solutions to deliver contextual, individualised answers, even if they don’t amount to regulated advice. “What you want to know about an ETF can be very different from what someone else wants to know,” Bahadir Yilmaz explained. “If we can understand that and respond in a personalised way, that’s real value.” 

Voice agents are also gaining ground. In countries like Spain and Germany, ING is testing voice-based AI systems that can handle customer queries 24/7. These aren’t simple IVRs; they are adaptive, context-aware interfaces capable of solving complex problems in real time. Finally, ING is applying Agentic AI to its internal processes, particularly in software engineering, where AI helps teams ship products faster and with fewer errors. 

BridgeWise, meanwhile, is tackling the intelligence gap between institutional and retail investors. Gaby Diamant, the CEO, explained that while professionals rely on deep analysis, retail investors often make decisions based on instinct, social media, or news headlines, resulting in nominal financial losses over time. BridgeWise uses proprietary models and LLMs to deliver institutional-grade insights to retail users, complete with explanations and timing cues. With the addition of agent-driven signals, the system doesn’t just answer questions – it proactively alerts users to significant market events. “Instead of the client approaching us, now we detect something in the market and serve it to them,” said Gabriel. “That’s what makes it truly agentic.” 

While front-end experiences capture most of the excitement, the infrastructure layer is just as critical, especially when agents start initiating financial transactions. Airwallex has positioned itself as the operating system for Embedded Ffinance in an agentic world. Their global stack enables businesses to issue cards, hold funds locally, and execute real-time payments – all accessible via API. 

As AI companies scale, they increasingly require infrastructure that matches their agility. “Agents don’t fill out paper forms,” Ciaran O'Malley, Enterprise Commercial Director at Airwallex said. “They connect via APIs. That’s where we come in.” This infrastructure becomes essential when AI agents act across geographies, currencies, and compliance regimes. By offering regulated, API-first solutions, Airwallex ensures that agents can perform their tasks without friction while remaining compliant and auditable. 

The challenge of trust and liability in machine-led transactions 

As agents become active participants in commerce, the need for trust, identity, and accountability becomes paramount. One proposed solution to this challenge involves giving AI agents verified digital identities that can be recognised across the payments ecosystem, a step toward making their actions traceable, trusted, and secure. 

But identity is only part of the equation. Liability must also be clearly defined. “If the agent buys the wrong product and it can’t be returned, whose fault is that?” Greg Ulrich from Mastercard asked. “Right now, it’s the consumer, but that’s not going to work in the long term.” The company is working to define new rules for dispute management, potentially recognising the agent as a responsible party in financial interactions. This would mark a shift in how commercial and legal systems treat machine decision-making. 

Ultimately, the success of Agentic AI hinges on trust, not just technical trust, but emotional and institutional trust. Consumers must feel confident that agents act in their best interest. Merchants must trust that transactions initiated by agents are legitimate. Issuers must believe that they are part of a transparent, stable value chain. “This is a very pro-consumer technology,” Greg Ulrich concluded. “If we get this right, with the right trust and the right standards, it could become even better than the shopping experience today.” 

Rethinking business models for the agentic economy 

While most discussions around Agentic AI focus on what it can do, Martin Moeller from Microsoft warned against a more subtle but serious risk: misusing the technology by applying it to outdated processes. “We saw this during digital transformation,” he said. “Companies sprinkled digital onto broken workflows and called it innovation, but it didn’t deliver the full value.” 

To realise the full potential of Agentic AI, businesses must do more than plug it into existing systems. They must redesign their offerings around autonomous decision-making and outcome-based experiences. Martin Moeller described this as “starting with a blank sheet of paper” – a clean-slate approach that considers the end-to-end experience from both the agent’s and the user’s perspectives. 

This redesign extends to the workplace itself. AI is no longer just about boosting productivity; it’s about changing the very nature of work. “Nobody joins a company to copy-paste information all day,” Martin Moeller said. “But that’s what fragmented systems force us to do. With agents, we can finally eliminate that friction.”  

Conclusion: a paradigm shift in action 

Agentic AI isn’t just another tech wave; it represents a full-scale reset of how systems operate, how services are delivered, and how people interact with technology. By shifting from passive assistance to autonomous action, it narrows the gap between intent and outcome, turning insight into execution in real-time. This evolution has the potential to simplify user experiences, unlock new forms of value, and rewire the infrastructure of finance, commerce, and digital life itself. 

But progress won’t be automatic. The rise of Agentic AI raises urgent questions around accountability, regulation, identity, and trust. It will require robust frameworks for verifying agents, managing liability, and ensuring transparency at scale. Legacy systems will need rethinking. Business models will need to be rebuilt for a future where the end user may no longer be human, but the human still holds the risk. 

Agentic AI isn’t about replacing people; it’s about designing smarter systems around them. And in doing so, it could reshape not only how we work and shop, but how we trust, decide, and live. 

About Claudia Pincovski