Banking journeys change all the time. How things happen now is a far cry from how they happened one or two years ago, never mind the dark days of branch visits and queueing.
What hasn’t changed, though, is that those banking journeys are still repetitive, data-heavy, and often spread across multiple systems, some of which are older than the employees using them.
Agentic AI has promise, and that promise could untangle things like onboarding a customer, investigating a disputed transaction, collecting compliance documents, or starting account applications.
But these are the same journeys where a wrong action can move money, expose personal data, deny access to a service or create a regulatory problem.
The question is, how much authority should a bank give AI?
What makes AI agentic?
A chatbot can answer a question, collect information and follow a predefined workflow. Many good banking journeys can and should be solved that way.
An AI agent is different because it doesn’t answer; it acts. It can decide which steps are needed, select approved tools, act on what it discovers, adapt when the situation changes and continue until it reaches a defined outcome or a boundary that requires human judgement.
Take an SME account application for example. A chatbot can provide a document list and collect uploads. A proper agentic system could identify the type of legal entity, build a case-specific plan, inspect the submitted documents, query approved identity and company-information services, spot inconsistencies, ask a tailored follow-up question and prepare the case for review. The final decision can still belong to a human who’s authorised to make the call.
The market is not on autopilot (yet)
The latest official local benchmark, published by the FSCA and PA in November 2025 using survey data collected in late 2024, found that 52% of responding banking institutions were already using AI in some way, shape or form.
Globally, adoption has moved beyond conventional AI. The University of Cambridge’s 2026 Global AI in Financial Services Report found that 81% of surveyed financial services firms were adopting AI at some level, while 52% were already piloting or deploying agentic AI. However, only 23% had progressed to the scaling or transformation stage. Among traditional financial institutions specifically, agentic AI adoption stood at 45%, compared with 57% among FinTechs. The same report highlighted privacy, bias, discrimination, explainability, governance, reputational harm and systemic vulnerabilities. In other words, adoption is growing at the same time as the consequences are becoming harder to ignore.
Five reasons why banks should use agentic AI
- It can manage an outcome across systems
A chatbot can tell a customer how to replace a lost card or guide them through a predefined sequence of questions. An AI agent can go further. Within its approved permissions (this is an important point), it could verify the customer, retrieve recent transactions, identify which payments need to be queried, request confirmation before blocking the card, initiate the replacement order, for example.
- It can adapt inside the case
Banking cases are usually more complicated than they appear to be. Maybe there’s a document that’s missing or the ‘customer’ is a trust rather than an individual. An agent can adjust its next step within policy instead of sending every deviation into the same manual queue.
- It can do prep work for human agents
An agent can gather records, compare evidence, retrieve the relevant policy, flag discrepancies and produce a traceable case summary. Fraud, compliance and credit teams would spend less time assembling the file and more time applying judgement.
- It keeps work moving after the chat ends
An agent can monitor an approved process, check whether a requested document has arrived, update the case and alert the right person when a condition is met.
- It can give skilled teams their judgement time back
The point is to remove the work that keeps skilled employees away from exceptional cases, relationships and decision-making. Agentic AI can give specialists a more complete case and customers a faster route to someone who can help.
