Artificial intelligence agents could soon become the primary interface between consumers and many financial products and services, pushing banks into the background in their customer relationships. Such an outcome is no longer hypothetical; in a recent six-month period, AI-driven traffic to US banking sites increased by 1,200%, according to Adobe Analytics.
The credit card business — where some of these dynamics are playing out in real time — illustrates how formidable the threat to banks could be. Agentic AI is poised to become a dominant customer acquisition channel by 2030, turning large language models (LLMs) into key players in what is currently more than 10% of the US consumer banking fee pool, according to our recent analysis.
Taking a wider lens, a shift toward LLMs as the front door to the internet could irrevocably change the dynamics of consumer banking and beyond. The response of individual banks will determine competitive positioning in the AI age.
Why agentic AI is the most important new channel since the Apple App Store
Agentic AI is an adaptive problem solver that can plan, adapt, and coordinate multistep workflows across systems with minimal human input.
The internet experience until now could be defined as a series of sequential, bilateral interactions: Users essentially toggle between websites and apps one at a time. Agentic AI promises a step-change in capability by handling much of the legwork behind the scenes and instantly. Picture a travel concierge that learns your preferences, books your entire trip, and updates plans automatically from a single prompt like: “Book me a family ski trip in February.”
Agentic AI isn’t merely a new technology to improve efficiency — it’s a whole new channel where the manner in which customers search, discover, learn, and buy is different. While still experimental, it offers enormous potential to disrupt consumer experiences.
How agentic AI will affect banking
Agentic AI changes the basis of competition for customer acquisition. For more than 25 years, search engine optimization (SEO) has dominated the process of digital product discovery. But as AI agents become the front door to the internet, a new discipline known as agent or generative engine optimization (AEO or GEO) is emerging.
In traditional SEO, the search engine shows sponsored posts based on businesses that pay for their products and services to be the top results. That model is giving way to a world in which the AI agent converses with the user, accounts for user preferences, provides a direct “answer,” and completes a multistep task.
Credit card acquisition has been migrating to digital marketplaces for a while, with 10% compound annual growth over the past five years, according to our analysis. Chat interactions with AI assistants are a natural fit for product discovery — and it is happening now at scale, led by ChatGPT, Gemini, and Perplexity. This reflects a durable shift in how customers evaluate products, driven by the rise of comparison-led digital journeys.
Let’s assume AI assistants capture all new growth for the digital marketplace channel, and the channel continues to grow at a 10% compound annual rate. Even with flat credit card originations, the market would experience a dramatic realignment.
How agentic AI is reshaping bank strategy and customer acquisition
The implications are profound. Banks that aggressively enter this new agentic AI channel will be share takers, while those that don’t will be share donors. We see this playing out over the next few years across three distinct groups of banks.
Group 1: The banks that embrace early participation
Open AI announced apps for ChatGPT in October 2025. The initial launch included nine partners: Expedia, Booking.com, Tripadvisor, Canva, Coursera, Zillow, Spotify, Figma, and Peloton. Since then, at least 21 additional brands released apps, and it is rumored that there are more than 300 coming soon.
Banks that embrace participation can use these apps to convert chat referrals into new customers by supporting account opening, and in parallel build a strategy for leveraging this new channel across customer segments, use cases, products, and partners. In developing a deliberate participation strategy, banks determine what the app experience is and isn’t. The app could support account opening only, then steer customers to traditional digital channels for daily interactions.
The opportunity for banks is to be an early joiner in an emerging channel that could reshape the world in the same way the internet browser did in the 1990s.
Group 2: The banks that participate through marketplaces
Digital marketplaces such as Intuit’s Credit Karma are already in the ChatGPT app store. Intuit is essentially replicating its marketplace in the agentic channel, capturing credit card, loan, and other leads and matching customers to suppliers (banks).
For some banks, a marketplace like Credit Karma might be the fastest path to market. While marketplaces and ChatGPT will almost certainly seek financial remuneration for providing the channel and directing users to apps, it could be argued that this is no different from existing marketing spending, with the cost to acquire directed at a different channel.
But participating indirectly might cede control. Banks rightfully take great pride in how they show up to customers. (They also have significant regulatory obligations to manage.) Showing up as a faceless chatbot buried in the interface of an LLM might seem unappealing. In credit cards, for example, Credit Karma gets the direct association with the customer and influences what offers the customer sees based on chat interactions. Bank products are visible, but marketing, branding, and customer experience decisions are centralized by the marketplace.
Group 3: The banks that avoid direct participation
There are real tradeoffs to be considered when deciding whether or how to participate in agentic AI. What if the channel never takes off? What if customers don’t trust apps or believe the recommendations provided by LLMs? What if acquisition costs are higher? What if fraudulent account openings skyrocket?
Banks certainly have the option of doubling down on existing channels, waiting to see how agentic plays out and what consumer expectations do and don’t change. But foregoing participation should be a deliberate choice, with efforts to increase consumer engagement in other channels and monitor incoming traffic from LLMs.
The agentic future in banking starts now
Whichever path individual banks take, we expect the dynamics will play out quickly. As agentic AI becomes a dominant acquisition channel across credit cards and beyond, banks that choose not to participate, even for valid reasons, could end up losing out.
This article is part of our Known Unknowns report, highlighting the debates that will shape the future of financial services in the age of AI.