In Pixar’s award-winning 2008 sci-fi film “WALL-E,” it is very clear what use cases the AI-powered robots, WALL-E and EVE, were designed for. WALL-E, a little robot left behind on an inhospitable, abandoned Earth, is designed for garbage collection. EVE, a sleeker new robot model, is designed to search and scan the planet for any signs of new life that may indicate Earth could become habitable once more. Agentic AI — a type of AI that can autonomously make decisions and execute on them — has entered the financial services industry, and more use cases are on the horizon.What Agentic AI Can DoRather than relying on traditional AI models that need human oversight to create content, agentic AI uses large language models (LLMs) to go one step further and perform real-world actions online. The agents can perform tasks with relative independence when they are given pre-defined goals and can even adapt their approach or decisions based on new experiences or knowledge. IBM notes that “this approach allows agents to ‘think’ and ‘do’ in a more human-like fashion.”That doesn’t mean, however, that humans are completely out of the loop. Ian Glasner, group head of emerging technology at HSBC, likens agentic AI to an intern “helping you get all of the more simplistic tasks done, but the human is still there to oversee and take the final decision.” Capability assessments conducted by bank leaders corroborate this approach: Between 95% and 92% of financial institution executives surveyed deploy agentic AI to advise and assist their human counterparts, rather than replace.How Financial Institutions Are Using Agentic AIAccording to a survey of 250 banking executives by MIT Technology Review Insights, 70% of respondents say their institution is already using agentic AI. Successful use cases include improving fraud detection (56%), enhancing security (51%), reducing cost and increasing efficiency (41%), and improving customer experience (41%).In addition to these use cases, community financial institutions (CFIs) can take advantage of even more ways that agentic AI can improve efficiency with the right training data. Since most CFIs will likely turn to fintechs and other vendors with developed agentic AI solutions that institutions can plug and play, it’s important to know what types of human behaviors should be mimicked and incorporated into the agentic AI’s training model in order to make it as accurate and efficient as possible.The following use cases are the most promising for CFIs to incorporate in their operating models:
- Contact center agents. Agentic AI within contact centers can provide top-tier customer service. Not to be mistaken for a chatbot, these agents are able to operate more independently and make complex decisions. To make the tool most feasible, CFIs should build or buy solutions that mimic how human staff build rapport and strengthen relationships with customers, modulate their communication style to fit a particular customer, choose which problem-solving approach is most effective for the situation in a way that builds loyalty, and how they spot opportunities to add value.
- Processing ACH returns. Agentic AI can easily handle ACH returns because they adhere to a fairly fixed set of rules. The agent can resolve straightforward items and flag those requiring human intervention in real-time, often scanning, validating, and making corrections within minutes.
- Loan underwriting. Underwriting is time-consuming, manual, and prone to human error, so it is a good candidate for agentic AI to step in and automate. Agentic AI can assist with making credit decisions based on a CFI’s credit policies and historical data. To increase accuracy, agentic AI should also be trained on how a CFI’s underwriters balance quantitative metrics with qualitative factors, determine which data points to prioritize during different market conditions, and which workarounds to use when clear answers are not provided. Of course, the more complex decisions should have that extra human touch that CFIs are known for along with the agentic AI.
- Money laundering monitoring. Agentic AI can support anti-money laundering (AML) compliance through credit risk assessment and fraud detection on transactions, then flag suspicious transactions in real time. To increase efficiency and decrease the amount of false positives, these agents should be trained on how real compliance experts prioritize alerts based on contextual factors, how to determine which data combinations trigger deeper investigation, and how to adapt the approach based on transaction patterns.
Where Agentic AI Could GoWith a learning, autonomous model like agentic AI, more use cases are constantly being developed and finetuned. The following are additional use cases that, with the proper training, CFIs could implement in the future:
- Automatic CD transfers to those with higher yields, based on a customer’s designated parameters.
- Treasury management tools that balance competing priorities and make strategic decisions under uncertainty.
- Wealth management tools that analyze portfolios, understand unspoken client concerns, and provide personalized guidance based on life circumstances.
- Commercial relationship management based on how bankers actually structure deals, negotiate terms, and identify cross-selling opportunities.
Agentic AI is emerging in the financial services world. There are already common use cases that your CFI can investigate and implement, when and if you’re ready. With the rapid adoption of agentic AI tools, it’s wise to start imagining how your financial institution can best leverage these agents for certain functions to increase back-office efficiency, improve accuracy, and free your staff for more consultative services. However, agentic AI is still new and isn’t a replacement for your trained, experienced staff, so starting small and with low-impact tasks is advisable, if your CFI wants to give agentic AI a try.
