What if an AI assistant could do your shopping? This is agentic AI: software that doesn’t just answer queries but carries out complex tasks autonomously.
Last week, at the Money 20/20 Europe event in Amsterdam, Mastercard, Dutch bank ING and payment services company Worldline announced they had completed “Europe’s first live end-to-end agentic payment.” A shopper told an AI assistant to look for concert tickets in a certain place on a given date, within a defined budget; the assistant found options, and after the shopper selected one, it paid for them, with human approval.
Agentic AI was one of the hot topics at the conference — billed as the largest annual gathering of the financial technology, or fintech, industry. For many years, fintech and traditional banks were seen as rivals, but many are now partnering to adopt these technologies.
Scarlett Sieber, the conference’s chief strategy and growth officer, told CNN that AI in finance used to be a buzzword, but now real adoption has gone beyond startups, and is “happening across the board.”
The deployment of AI agents across the financial industry is expected to jump from 24% today to 81% by 2030, according to a 2026 report led by the University of Cambridge that surveyed over 600 firms and regulators worldwide. However, it cautioned that the rapid technological change “currently outpaces the supervisory frameworks and technical capacities required to oversee them.”
The Israeli multinational eToro is known for its investing app that lets users buy shares and replicate other traders’ moves. CEO Yoni Assia told CNN the app’s AI assistant had been recently upgraded, from providing financial advice on a user’s portfolio to acting on their behalf under preset limits.
One striking example is an app on the eToro platform called POTU$, which scans Donald Trump’s social media and news about him. When the US president posts something that may move markets, it can place a trade in a user’s account within seconds.
Across the company, Assia claimed, AI use had grown roughly tenfold in six months, and 95% of its new code was now AI-written, up from none two years earlier. Still, he added, AI “is useless without humans steering it.”
Last month, Klarna, the Swedish company behind the “buy now, pay later” buttons at many online checkouts, became the launched a shopping search app in ChatGPT. In 2024, Klarna built an AI assistant with with OpenAI to handle customer service — claiming that it did the work of 700 full-time human agents.
New technology such as AI “allows us to do more with less,” Klarna CEO Sebastian Siemiatkowski told CNN at the event, adding that the company’s workforce had fallen from 6,000 people to fewer than 3,000 in recent years, while revenue per employee rose.
Last year, Bloomberg reported Siemiatkowski saying that cost cutting had led to “lower quality” and that Klarna had begun rehiring human agents, “investing in the quality of the human support.”
Speaking to CNN, Siemiatkowsk conceded that AI could lead to job losses across industries. He said customer-facing jobs, from sales to lawyers, will “fare very, very well,” but there “might be short-term negative implications in specific job areas.” He declined to say what share of Klarna’s jobs had been replaced by AI.
For traditional institutions like ABN AMRO — the third-largest Dutch bank — AI is part of a broader digital shift, as it went from 500 physical branches in 2010 to 26 today, CEO Marguerite Bérard told CNN. It also plans to cut 5,200 jobs by 2028, from 2024 levels.
“Eighty-five percent of our colleagues in the bank, myself included, use AI in our daily work,” she said, adding that customers hold millions of conversations with its AI bot “Ana” and that its AI bot “Lenny” is streamlining credit requests.
Beyond banking, the growing adoption of agentic AI across industries has raised concerns from some quarters. Research company Gartner last year predicted that more than 40% of agentic AI projects will be canceled by the end of 2027, “due to escalating costs, unclear business value or inadequate risk controls.”
A recent report from professional services company Accenture and Wharton business school noted that with increased automation from agentic AI, “leaders must determine which decisions to delegate, where human judgment must remain central, and how governance, accountability and trust are designed into the system.”
For Bérard, human oversight is key. “If you put AI on a bad process, you still have a bad process,” she said. Her rule is “guardrails but no handcuffs,” with “always a human on top and in the loop.”
Leave a comment