Agentic AI allows financial systems to act on a user’s behalf, making payments based on permissions and goals agreed in advance. In this article, Sofia Khatsernova, Legal Function Owner at xpate, discusses what happens to consent, control and accountability when people no longer approve each transaction themselves.

When a payment goes wrong, responsibility can become difficult to trace between the user’s instructions, the information the system relied on and the decision it made. Sofia explains why financial institutions need to address these questions when designing autonomous services. Spending limits, specific permissions and clear transaction records are among the safeguards she proposes, alongside human oversight. The focus is on keeping users informed and in control as financial decisions increasingly happen in the background.