What is agentic AI, and why should banks or customers care?

2 Likes 0 Be the first to comment

What is agentic AI, and why should banks or customers care?

Contributed

This content is contributed or sourced from third parties but has been subject to Finextra editorial review.

You’ll soon be hearing a lot about agentic AI, if you haven’t already.

The World Economic Forum defines and explains agentic AI and its impact in a recent article on the topic, noting how this discipline/approach/tool “will transform” the financial services industry.

“Agentic artificial intelligence (AI) goes beyond generative AI (GenAI) by enabling autonomous decision-making, collaboration, and learning to revolutionize financial services.”

Before we go further into actual agentic AI examples, it’s important to include a bit of history. Financial institutions have used AI for many years - in some cases multiple decades - to fight fraud by identifying risky or questionable transactions and patterns, to detect suspicious payments, connections, or trends that might indicate possible money laundering activity, or to feed machine learning algorithms with any or all of these data points. These ‘back office’ applications have proven successful at building longer-term ‘knowledge’ of transaction, behavioural, and other data and patterns to help the financial institutions using them to augment human interpretations and assist them in performance of their duties.

AI’s been in larger institutions for decades, with agentic variants now being tested

Beyond traditional uses of AI in financial services, in the real world, generative AI’s training and usage from basic to sophisticated (could) mean that if a bank customer consults a chatbot to ask, for example, “How can I change my payment due date?” on a credit card account, the agent/bot might respond with instructions on how to do this (if possible or as allowed) and even better, it might provide a link to the page where the change could be executed by the client. Or, if it’s not a very effective AI agent, it might say: “I’m sorry, I don’t understand, I’m limited in the questions I can answer, please consult our FAQ’s.” Or something worse, like, “Please wait for one of our staff members to contact you.” And then, the frustrated customer hopes they actually do, sooner rather than later or before the customer has to try other avenues of service resolution.

How agentic AI stands out from more traditional models

Technically, how agentic AI differs from other forms of AI (like ‘traditional’ AI – as defined and used by financial institutions up ‘til now), and its cousins' machine learning and even standard, ‘vanilla’ generative AI, or Gen AI, is something anyone can look up and understand. Most people have at least heard of, if not used ChatGPT, Claude.ai, Perplexity, Copilot, and/or any number of other Gen AI platforms and providers scrambling to establish and cement their positions in this nascent industry. That’s where generative AI starts, while agentic AI is one of the first places it can go from there.

Typical queries placed into the simple, text-based ‘windows’ in mass market Gen AI systems might mimic internet searches, like “List the top five ways AI is used in financial services companies worldwide.” Or, they might be significantly more complex, like “List the names of the top five financial services companies using AI worldwide, and place them in a table with five columns including name, address, asset size, ranking in current year, and ranking in prior year.”

Input these prompts into any of the popular Gen AI platforms, and you’ll likely get what you asked for, and sometimes more than you expected. Hopefully, the answer it provides is accurate, and the system doesn’t ‘make up’ or incorrectly juxtapose details in what is being called a ‘hallucination.” But, going beyond just what that system knows, perhaps to providing additional useful information or further recommendations is where agentic AI might come in to the picture. And that reality is not far off, according to most industry sources.

Big questions surround agentic AI’s potential within banking

Beyond its definition, and in the quarters of many larger companies, including the top financial services firms right now, sit questions of what agentic AI might really mean for them and their clients. Well, nearer still if you bank with one of the largest institutions, and definitely not too far down the road beyond late 2025 if your money is placed with many of the regional or local banks or credit unions.

In fact, some large US financial institutions and companies in other industries are now testing all sorts of agentic (or “agentive”, another term for these AI agents) use cases for validity and tangible results.

First, they say, they’ll likely perfect internal examples, like proposing talking points for upcoming meetings based on agenda description and documents, writing user manuals from product and client service data, or identifying sales opportunities from analysis of client relationship management (CRM) records. Of course, these are just a sampling of many internal applications for agentic AI described in recent months through industry panels, presentations, and webinars.

Once deployed, agentic AI use cases expected to multiply quickly in financial services industry

External use cases for AI agents will be slower to hit the US marketplace, in the view of most industry experts, especially within the highly regulated, and typically risk averse, cautious banking sector. That said, when these new AI agents do emerge, don’t be surprised if they bring ‘friends’ along to the party. Industry pundits and AI program managers inside and outside financial services are already talking about designing multi-level AI agents to work together to accomplish many tasks or make several recommendations at once.

So, to use our previous example, not only will the financial institution’s client service bot quickly explain how to change a customer’s due date upon request, then take them to the page to do it, it might wait for them to make the change, then send a notice to a program database or service staffer (or another bot) to follow up, and perhaps put the customer on a list for outreach from another part of the institution with a potential new cross-selling opportunity – the ‘perfect product’ to help the customer meet their time and money management needs.

Comments: (0)

/ai Long Reads

Sehrish Alikhan

Sehrish Alikhan Reporter at Finextra

How can AI become more sustainable?

/ai

Scott Hamilton

Scott Hamilton Contributing Editor at Finextra Research

What is agentic AI, and why should banks or customers care?

/ai

Dominique Dierks

Dominique Dierks Content Manager at Finextra

What is the global AI legislative outlook?

/ai

Hamish Monk

Hamish Monk Senior Reporter at Finextra

How AI can be embraced by financial services

/ai

Editorial

This content has been selected, created and edited by the Finextra editorial team based upon its relevance and interest to our community.