The first obvious change is the shift from long menus to conversation. Instead of searching through multiple screens, a customer can ask about their balance, last payment, or the reason for a spike in spending in a particular category.
Accenture found that 71% of participants welcome an AI assistant within their primary bank app. The idea is not to create another chatbox, but to link the question to account data so that the answer comes with what is actually happening. This makes small tasks faster and reduces navigation between pages.
It is even better if the user can trace the source of each answer, especially when the question is about a payment date, available balance, or a change in spending.
The next step is the execution of the action itself. The agent may move from a phrase like 'Remind me of the invoice' to 'Pay it if the balance is sufficient' when the user grants the appropriate permission. Mastercard indicates that frequent, low-risk purchases may be among the first tasks users delegate to AI agents.
The real difference here is the transition from suggestion to action, with control remaining with the account holder.
Monthly bills are well-suited for an agent because they are recurring and predictable. The system can collect their due dates, detect a change in a usual amount, and then alert the user before the deduction.
Practically, the task can include:
Sorting bills by due date;
Showing the expected amount for each payment;
Alerting the user when a recurring value changes;
Suggesting a suitable payment day based on the available balance.
These functions do not require full autonomy. Sometimes it is enough for the agent to prepare the steps and request one confirmation at the end.
Continuous monitoring can be more useful than a long monthly report. A banking agent can compare usual flows with new movements and highlight what deserves attention.
For example, it may notice that several small payments appeared in one day, or that a subscription was repeated with a different amount, or that the balance is approaching a level set by the user in advance. The important thing is that the signal is specific and understandable, not just a general warning.
Accenture expects that banking experiences will become more context-dependent and conversational. This makes monitoring useful when it explains the reason, not when it adds more notifications.
And this monitoring becomes more useful when the customer himself determines what deserves an alert, such as a balance limit or an unusual increase in a specific category.
Personal analysis becomes stronger when it does not just display a graph. The agent can link spending changes to clear behavior, and then answer a question like: Why did this week's expenses increase?
Function
What the Agent Sees
Useful Outcome
Spending Analysis
Recurring Transactions and Categories
Explanation of Spending Change
Balance Monitoring
Incoming Income and Payments
Early Alert
Period Comparison
Previous Weeks or Months
Revealing Habit Changes
Intent Reading
User's Question and Context
Suggesting an Appropriate Step
These insights do not mean making a financial decision instead of the customer. Their value lies in transforming account data into a short, quickly reviewable explanation.
Banking does not always happen on a single screen. A user may start a question in the bank app, then complete the task from a digital wallet or another chat interface.
Accenture says that 65% of participants are open to using a financial assistant similar to GPT tools through an AI platform or digital wallet. Therefore, context continuity becomes important: the user should not have to re-explain the task from the beginning every time they switch between channels.
This puts banks in front of a practical challenge. The system must know who the user is, what they have requested, what has been executed, and what is still pending confirmation, without the experience becoming confusing.
Switching between channels also requires a clear record of previous steps, so the customer knows what has been done and what needs new approval.
What makes agentic AI different is not the smarter speech, but the ability to transition from understanding to execution within clear boundaries. Mastercard expects the use of agents in purchasing and payment to expand, while Accenture believes that banking will become more conversational and context-adaptive. For the customer, the real value will appear when the steps in daily tasks decrease, the information becomes clearer, and the final decision remains under their control.
