Today’s AI agents and Gen AI tools have the potential to remake the financial industry, but only if leadership addresses the human element in AI transformation. This isn’t simply a feel-good strategy, it’s critical for ensuring ROI on AI investments. Although agentic AI could generate up to $450 billion in value by 2028 through revenue uplift and cost savings, Gartner predicts that more than 40% of agentic AI initiatives will be shut down by then because of “escalating costs, unclear business value or inadequate risk controls.”
Agentic AI and Gen AI investment offers the banking and finance sector a high-risk, high-reward scenario. AI success starts with understanding that even the most powerful AI models today need to be trained as if they are new employees who require time to learn and develop habits. In tandem, employees need new skills for working with and managing AI agents and processes, as well as new pathways to channel their innovation.
New roles for AI and people in banking
Agentic AI is the latest automation solution adopted by the financial services industry. That process began with rules-based robotic process automation (RPA), progressed to simple AI models that could leverage unstructured data, then evolved to Gen AI that can create new content, and now to AI agents that can orchestrate complex end-to-end processes with maximum autonomy. Unlike past automation tools, agentic AI acts like a team member, and like any new member it changes the team dynamics. All the people on the team will need to know how to work with the agent, including initiating, controlling, and validating the agent’s work.
Agentic AI is ideal for financial services because many tasks, like wealth management strategy development, are personalized for each client. AI can also automate and orchestrate repetitive and complex processes that currently require lots of manual work, such as know-your-customer (KYC) checks for new customer onboarding and compliance. Deploying AI for these use cases — and others such as hyper-personalized marketing in retail banking — can let institutions accomplish much more with the same number of people.
If AI agents are handling, let us say, 50% to 85% of repetitive processes, the workers completing the rest of those tasks will also need new skills to manage the agents. For example, bank IT departments are always overwhelmed with requests. With AI automation and agents, an IT team can complete more requests from the business side:
- Gen AI can generate a large proportion of needed code automatically.
- Agentic AI can analyze and resolve support tickets.
- Human IT team members can work on higher-value projects and oversee the AI — if they’re given the training they need to do so.
Readiness for AI implementation in banking
Despite its potential, there are few AI agents in production in the banking sector now, although there are many in the pilot stage. Trust and compliance are probably the biggest hurdles to full deployment, due to several challenges.
One is model training requirements.Many AI agents can achieve about 85% accuracy soon after deployment, but getting the rest of the way to 99% or 100% takes time and training by employees with proper skills. Successful AI model training also requires a strong data foundation, which may take time to build.
The second challenge is model risk management, including cybersecurity and governance.Agentic AI systems must comply with the organization’s ethics and with data privacy regulations. This requires the development of guardrails and transparent model validation processes, including documentation and prerequisites. Compliance-by-design principles can ensure that agentic AI or Gen AI-based systems are designed and built to facilitate validation by the model risk management team.
The third and potentially most overlooked challenge is change management. You cannot deploy agentic AI or Gen AI systems without onboarding all your teams, because otherwise adoption can be a problem. Although nearly 70% of workers say they welcome AI automation that gives them more time for more important work, 45% have “doubts about the accuracy and reliability of AI systems,” according to a Stanford University study. Those doubts, if not addressed through proper model training, model validation, and employee training, have the potential to undermine adoption and ROI.
Where could AI take banking in the next few years?
Banks that successfully implement agentic AI and Gen AI can expect major changes in several areas, such as these.
New brand engagement strategies
It’s easy to imagine agents handling virtually all payments for banking customers, so that they become invisible in the way Uber payments are now. For example, instead of paying separately for your hotel, car, and airfare when you book a trip, your card issuer’s AI agent might handle it all for you. If customers no longer need to engage with their financial services providers for day-to-day experiences, banks will need to find new ways to maintain brand awareness and loyalty.
Compliance and risk management improvements
When KYC and other processes are highly automated and agents can orchestrate vast streams of data for more accurate risk forecasting, banks can manage risks more effectively and maintain compliance more easily. This can help institutions avoid severe financial losses and weather whatever economic shifts the future may hold.
More focus on managerial skills and innovation
Employees will need skills for training, managing, and monitoring AI agents, as well as for whatever iterations of AI and automation come next. They’ll also need the opportunity to do more innovative and higher-value tasks in order to work to their full potential.
We can’t be sure how the future plays out, but a lot will depend on how financial institutions strategize and implement their AI and employee training initiatives during the next couple of years. Approaching these projects with an eye on training, validation, and change management can help institutions succeed in realizing the value that today’s AI offers.
The post How Will Agentic AI and Gen AI Transform Banking? appeared first on PaymentsJournal.
