Self-service finance is here, but the technology needs room to grow before consumers will fully embrace it.
In the age of DIY and self-checkout, financial management has entered its own “do it yourself” revolution. Artificial intelligence-powered tools and chatbots now help people budget, invest, and plan for the future — all without waiting for a human advisor.
Large financial institutions like J.P. Morgan and Bank of America see millions using their self-service wealth management tools — like Wealth Plan and Erica, respectively — on a daily basis.
Two years ago, the Consumer Financial Protection Bureau estimated that around 37% of the U.S. population was interacting with their financial institution’s advanced technology tools and chatbots. It also forecasted this number to increase by 13% by 2026.
“Financial institutions have begun experimenting with generative machine learning and other underlying technologies such as neural networks and natural language processing to automatically create chat responses using text and voices,” according to the CFPB report.
The CFPB also cautioned institutions about some risks associated with relying heavily on these kinds of tools, noting issues with compliance, increased consumer distrust with tool interactions and keeping personal information safe.
So while banks and fintech companies are saving significant costs thanks to these advances, the transformation brings new questions about speed, intelligence and user experience, as well as how best to strike the delicate balance between privacy, compliance and technological innovation.
The Evolution From Chatbots to Agentic AI
The journey from basic chatbots to sophisticated agentic AI showcases a dramatic leap in technology. Early “chatbots” were little more than automated frequently-asked-questions portals, handling simple queries like balancing accounts or payment transfers.
Joshua Schechter, Kasisto’s chief product officer, doesn’t particularly like the term “chatbot,” instead referring to these tools as “conversational AI.” Kasisto provides agentic AI agents for financial services through its KAI platform.
He describes this evolution as moving from calling the bank to interacting with self-serve tools. What began as self-service chatbots has progressed to conversational AI and now to agentic AI agents that deliver more personalized and context-aware experiences.
Modern agentic AI understands context, learns user preferences and can anticipate needs — a stark contrast to rigid, menu-driven interfaces that once frustrated users with “I’m sorry, I didn’t understand” responses, Schechter explained.
“What we’re seeing now is that customers of financial institutions and even employees of financial institutions, are demanding a ChatGPT-like experience,” he said. “What does that mean? An experience that’s more intelligent, more personalized, has some memory, has some context about you, and can really predict, be proactive and personalize that experience for the customer or the employee.”
Catering to Consumers
Today’s financial consumers expect speed, convenience, and accuracy, all wrapped in a seamless experience.
Since launching in 2024, J.P. Morgan’s Wealth Plan tool “has been very successful in customer engagement and adoption,” with over 4 million customers setting up Wealth Plan so far, according to Sunsy Hong, managing director and general manager of the planning platform and experience at J.P. Morgan Wealth Management.
One of the popular features is “Goal Simulator,” which allows customers to adjust different goal parameters, such as retirement age, and see how different actions today can impact their future.
“Customers like the ability to play around with different scenarios,” Hong said via email. “They find it very engaging and informative.”
However, not everyone has logged onto these new tools. In early 2025, Deloitte surveyed just over 2,000 bank customers and found nearly 4 out of 10 respondents have yet to interact with AI-powered tools offered by their financial institution.
Unlike the CFPB report from 2023, the good news from Deloitte is that perception of tools like this have shifted, especially among younger generations. The survey found that “Millennials and Gen Zs, raised on digital interfaces, are more likely to report positive interactions, in contrast to baby boomers and Gen Xers, who rate their experiences less favorably.”
That said, a majority — 74% — still want a human agent available when needed. And the chief complaints as to why these tools might not be helpful are the need for better accuracy, more personalization and stronger security.
The best AI-powered systems pass a threshold where they feel “instant and high quality,” no longer leaving users desperate for a human agent or stuck in an endless escalation loop.
In fact, AI can be many things at once, said Darragh Curran, chief technology officer at Intercom, which designed its Fin customer engagement platform for digital-first banking institutions.
“AI can be your support person, your sales person, and it can also talk to millions of people in parallel,” Curran said.
He also observed that while initial bot deployments often led users to ask for a “human” immediately, the newest AI agents increasingly satisfy users, sometimes so convincingly that customers don’t recognize they’re interacting with an AI.
However, not all AI solutions deliver this level of performance. Curran cautions that, despite the promise of universal DIY finance, many AI experiences are still not good enough, and businesses must evaluate not just happy-path outcomes but performance in more complex scenarios.
Speed versus Intelligence
The promise of DIY finance is speed and efficiency, but is it always smarter?
Schechter cites studies showing that up to 86% of typical contact center conversations could be handled via digital menus and automated agents. This means automation has succeeded for the simple and routine, saving banks millions.
Scott Simari, principal at management consultancy Sendero Consulting, draws a parallel to broader digital transformation, sharing that moving from in-person branches to DIY tools is the culmination of two decades of technology investment. But the last mile, automating personalized advice at scale, is an especially hard problem for all but the biggest, most tech-savvy institutions, he said.
Simari quoted MIT researcher Alan Thorogood, who coined the phrase “hug of death,” to describe what happens when banks, fintech partners, and the fintech lose focus on their own priorities in favor of customers’ needs.
“It’s the hug of death because they love you so much they want you to help them shepherd through their internal bureaucracy and move the ship,” Simari said. “It just takes forever.”
Meanwhile, truly complex needs — like nuanced financial planning, investing, transfers between accounts and future expenses — remain difficult to automate meaningfully.
Only a small percentage of AI investments have shown measurable returns, making the leap to agentic AI an expensive gamble for most banks, Simari explained. Most financial institutions are still running pilots in sandboxes (i.e., not yet for public consumption) and “actually getting something to production at scale that is fully governable and auditable is the multi-million dollar question,” he said.
Many solutions today are great for budgeting, basic financial literacy and forecasting, thanks in part to AI. For example, J.P. Morgan’s Wealth Plan leverages generative AI to interpret large amounts of customer data and provide more personalization, which will ultimately improve the customer experience.
However, when customers want tailored investment advice or product recommendations, the role of a human advisor remains irreplaceable.
“Technology has evolved significantly, allowing for an innovative architecture that enables collaboration between clients and advisors in real time,” Hong said. “Wealth Plan was one of the first tools in the industry that achieved this.”
Protecting Data Privacy
Meanwhile, financial data is highly sensitive, and institutions must balance innovation with strict privacy and regulatory requirements.
Regulation hasn’t fully caught up with AI, Curran said. Yet, the principles also haven’t changed: respect for data, treating vulnerable customers well and having audit trails.
AI introduces new risks, especially with non-deterministic outcomes, but progressive companies strive to set a higher bar for AI than for humans, ensuring a consistent and compliant experience, Curran said.
“The way you must treat and respect data in financial services is to demonstrate a consistency and process followed,” he said. “All of these things were common before AI and will be after AI.”
Kasisto has addressed this with its approach to build its own large language models, ensuring that customer data is never sent to third-party models for retraining.
This gives banks greater control over compliance and data hygiene, a crucial differentiator when regulatory scrutiny is high. For regional banks and credit unions that don’t have the resources to build enterprise-scale AI, partnerships with fintech companies and strict orchestration strategies are essential, he explained.
Simari noted that fintechs and smaller banks often must buy and orchestrate AI capabilities, focusing more on interoperability and internal governance than on building massive bespoke solutions from scratch. The key is making sure that different agents, whether they’re handling legal compliance or sales, can communicate and coordinate, rather than operating in silos.
The Future of Self-Service Finance
Despite challenges, the future of self-service finance looks rich with opportunity. For now, DIY finance excels in routine matters and basic planning. But as AI agents grow smarter, institutions will increasingly be able to leverage these tools to drive both customer acquisition and service efficiency.
Schechter points out that agentic AI technology has only been around for about 18 months, so there is still a lot of innovation to be had for both consumers and also employees.
“There’s a push in the market today for financial institutions that it’s not just about servicing their customers, it’s also about servicing their employees,” he said. “It will be about giving employees agents they can leverage to help them become more efficient in their roles and take away the mundane tasks.”
In addition, the human won’t be completely removed, he said. There will also be a human-in-the-loop factor that regularly supervises and reviews what the AI agent is doing. Ultimately, the queue for human interaction will decrease as the AI is able to understand and perform more complex issues.
The best implementations don’t just deflect traffic or save costs; they drive business outcomes and customer satisfaction. Mature organizations, like Intercom, pride themselves not only on innovation but also on reliability, compliance and customer-centric design — a blend essential for success in financial services.
“Fin will do exactly as you expect, but won’t make stuff up or invent things that sound good,” Curran said. “It’ll use all the knowledge, which is what you’d expect of a great human.”
Banks and fintechs that find the right balance between speed, intelligent automation, privacy and personalized service will not only save money but also transform how people interact with their money, democratizing access to financial advice while meeting ever-demanding consumer expectations.
“Everyone’s focused on the authenticated experience, but in the next frontier, we will see personalization levels that go beyond authenticated experiences,” Schechter said. “We’re starting to see banks experiment with this in low-risk areas like instant approval for home equity lines of credit. We will see more education and more proactive and outbound use cases.”
