As this fall’s conference agendas are rolling out, one thing is clear: everyone wants to talk about agentic AI. This newest form of artificial intelligence is just the next step in the recent AI transformation we find ourselves in – not just in financial services, but across every industry.
What is it?
Agentic AI, according to NVIDIA, “uses sophisticated reasoning and iterative planning to autonomously solve complex, multi-step problems.” To oversimplify, agentic AI can think for itself and complete tasks on its own.
How is it different from gen AI?
At face value, it’s easy to look at agentic AI and generative AI and assume they are the same. Gen AI takes a prompt and creates text, images, video, etc., based on the directions it’s given. This is where agentic AI is different: it does not require individual prompts or human oversight. It’s able to use reasoning and make decisions all on its own.
Think of a self-driving car. After giving the initial prompt of the location to drive to, no one has to give it additional prompts to turn left or stop at a red light. It takes in the information around it and makes decisions based on that information in order to safely navigate. That’s agentic AI in action.
Where is agentic AI in financial services?
For consumers, agentic AI in their day-to-day finances might look like a bot that gives advice or helps with financial planning. Essentially, it can serve as a 24/7 assistant for all things banking, investments and more.
For financial institutions, agentic AI can take many forms. It can handle everything from risk assessment and financial forecasting to identity verification and fraud detection. Nearly any routine task can likely be managed by agentic AI. This will have – and is already having – huge impacts on many of the tasks that have been traditionally outsourced, or the manual processes taking up a large portion of financial institutions’ time and effort.
What’s next?
It’s not a matter of if agentic AI transforms financial services, but when. We’re already seeing it at work, especially with larger institutions. This makes AI fluency critical moving forward. Leaders who want to keep up need to ensure that people at all levels of their organizations understand AI if they are going to work with it. Just because it requires little human oversight does not mean agentic AI should run free with no supervision, and people cannot supervise what they don’t understand.
Since AI should never be a replacement for human input entirely, the best applications of agentic AI will marry the efficiency of technology with the quality of human input, and the organizations that strike that balance sooner rather than later will be the ones leading the charge.