AI helpers taking over money tasks – but who’s really in charge?

At a big tech fair in Amsterdam last month, three big money companies showed off a new kind of digital helper. Instead of just answering questions, this helper actually paid for concert tickets after picking the best one within a set price and date. The shopper only had to say yes – everything else happened automatically. Fintech fans say these helpers, called agentic AI, are the next big step. For years, new digital banks and old-school banks fought over customers. Now many are working together to use this software that acts like a super-smart intern – doing research, comparing options, and finishing tasks without constant checking. A study by university researchers asked over 600 companies worldwide about their AI plans. Almost a quarter use these helpers now, but by 2030 that could jump to four out of five. The catch? Most places aren’t ready to watch over these helpers properly. The rules and safety checks haven’t caught up with the speed of change. Some companies already let these helpers trade shares for them. One investing app scans politicians’ posts and market news, then buys or sells stocks in users’ accounts within seconds when something important happens. The company says it uses ten times more AI code than six months ago – but still needs humans to set the limits and make the final call.Customer service is another hot spot. A Swedish “buy now, pay later” firm built an AI assistant that handles as many chats as 700 human workers. The CEO admits they cut jobs to save money, then had to hire some back because customers complained about poorer service. He warns that while AI might replace some tasks, customer-facing roles like sales and legal work will likely stay human-heavy. Old banks are also jumping on board. One major Dutch bank went from 500 branches to just 26 in ten years, and plans to cut thousands more jobs by 2028. Their CEO says 85% of staff now use AI at work, but insists humans stay in charge. “Bad processes stay bad even with AI, ” she says. “We need guardrails, not handcuffs. ” Not everyone’s convinced. A research firm predicts over 40% of AI projects might get scrapped by 2027 because they cost too much or don’t deliver real value. Another report suggests companies need clear rules for when to let AI decide versus when humans must step in. The big question remains: how much control should we give to machines that follow instructions but can’t understand real consequences?