LONDON: Agentic banking AI is moving beyond chatbots as banks and payment networks test systems capable of executing financial tasks under predefined customer permissions.
Santander and Mastercard completed a live, end-to-end AI-agent payment in Europe in March 2026, according to the source provided. The test used predefined limits and permissions.
BBVA and Visa followed with an agent-initiated payment test in July. The transaction used real card credentials, tokenisation, fraud monitoring and biometric authentication.
DBS has also moved its corporate virtual assistant toward authenticated banking tasks. ING, meanwhile, tested an agentic mortgage assistant while keeping the final lending decision with a human employee.
Read: ChatGPT Work Launches as OpenAI Adds Office Agent
The UK Financial Conduct Authority’s July 2026 Mills Review examined how autonomous agents could manage finances and accelerate switching between providers.
Read: Meta Acquires Moltbook AI Network In Push Toward Agentic AI
Its research found 20% of consumers would consider AI acting autonomously within preset goals. HM Treasury has also opened a consultation on payment rules for agentic AI.
The review considers authentication, consent and liability when software initiates or executes transactions on users’ behalf.
Read: US Financial Firms Step Up Cyber Monitoring as Iran War Raises Attack Fears
Continuous AI-to-AI negotiation over savings rates, fees, or loan pricing is not yet mainstream. The supplied material presents that model as an emerging next step rather than an established banking market.
