Are banks becoming irrelevant in the AI era? With Theodora Lau

Why this conversation matters now: #

For most of the last decade, the question for digital banks was about feature parity and operational efficiency. Better app, cleaner onboarding, lower fees. In this episode of the Fintech Garden Podcast, Theodora Lau — founder of Unconventional Ventures and author of Banking on (Artificial) Intelligence — reframes the question entirely. The competitive frontier is no longer between digital and incumbent banks. It is between the entire banking industry and the AI tools that customers are now using as their first financial advisor. Per a recent JD Power report, 53% of US consumers have asked AI for financial advice in the past three months. Most of those conversations happen without a bank in the loop at all.

We’re in the iPhone moment for AI, but most operators don’t recognize it: #

Theo opens with a useful parallel. She was working for a mobile carrier in 2006-2007 when she first saw the iPhone, lined up to buy one, and came back from an overseas trip a week later with $300 in roaming charges because she could not stop checking the weather app. At the time, that obsession felt trivial. In retrospect, it was the signal — a new form factor that would reshape how people related to information, money, and one another. AI is in the equivalent moment now. The technology has been embedded in financial systems for years. What is changing is who controls the user’s relationship to it, and how that relationship is shaped.

Banks know mechanics. AI requires they understand intent: #

A persistent gap in financial product design is that banks know transaction mechanics — what was paid, when, where, in which currency — but not transaction intent. Was the user buying dinner or restocking inventory? Saving for a holiday or covering an emergency? Pursuing a long-term goal or reacting to a short-term shock? AI’s value in financial services is not primarily in automating workflows. It is in orchestrating the data banks already hold around the user’s actual life — life stage, goals, decision context — and surfacing it back as something useful at the moment of need. The product mindset has to flip: not “what can I sell you” but “what do you need, and how do I meet you there.”

What banks should do in the next two to three years: #

Theo’s answer is operational, not aspirational. The first priority is data foundations. When a customer’s information lives across 20 or 30 systems, no AI strategy will produce useful output, because the underlying retrieval will fail before the model gets a chance to reason. The second priority is updating systems for real-time service, because customer expectations on speed are now non-negotiable. The third, and most underdiscussed, is organizational. Most banks talk about leveraging AI in the future workforce. Fewer have started the actual rescaling, upskilling, and structural reorganization that comes with it. The cultural and training work runs in parallel with the technical work, and it cannot be sequenced after.

Transparency is the foundation, with or without AI: #

A recurring theme in the conversation is that financial services is a regulated industry, and that constraint is the asset most often misread as a burden. Whether a lending decision is made by a human or an algorithm, it requires audit trails, explainability, and an expert-in-the-loop at the final checkpoint. Theo’s highway analogy lands clearly: drivers need traffic rules, speed limits, and signals before they will adopt a road at scale. The same is true for financial products. AI deployments that cannot be defended in front of a regulator are products that cannot scale. AI deployments that can be explained back to the customer are products that earn trust.

The AI-native bank is not about the interface: #

Asked what an AI-native bank should look like in five to ten years, Theo declines to commit to a form factor. It might still be a mobile app. It might be a chatbot embedded in another surface. It might be multi-channel. The defining characteristic is not the interface but the depth of understanding the bank has of the customer — anticipating needs rather than reacting to them, knowing what the customer’s financial life looks like rather than offering a generic product catalogue. The mistake operators make is treating “AI-native” as a UX direction. It is a relationship and data direction. The interface will follow whatever the customer’s primary AI happens to use.

Where AI should not replace humans: #

The conversation is explicit on the limits. Wealth management decisions should keep humans in the loop. Credit decisioning needs an expert at the final checkpoint. Anything that has to be defensible to a regulator needs a human signature, not just an algorithmic one. And — a less obvious point — many people call their credit union not because they have a question, but because they want to talk to a human. Theo makes the case directly, with a respectful pushback against the framing that bots are now indistinguishable from humans. “Would you like to replace yourself with a bot and have the bot interview me?” The point is not nostalgic. Some interactions exist precisely because the human element is the value, not the inefficiency.

The harder question: banks are no longer in the conversation: #

The most consequential shift Theo identifies is not about how customers interact with their bank. It is about whether they interact with the bank at all. A US consumer asking for financial advice today is increasingly uploading a CSV of their finances to ChatGPT, Claude, or Perplexity. The conversation runs end-to-end between the user and the frontier AI model. The bank may surface in the response, or it may not. If it does not, the customer’s decision is made without it. The strategic question is no longer how to position the bank against competitors. It is how to position the bank to show up at all.

From SEO to GEO — discovery is being rewritten: #

The acquisition model that fintech operators have built over the last decade depends on the customer encountering a message — a search result, a notification, an ad, an in-app prompt. AI agents do not see messages. They synthesize answers. The industry is starting to talk about Generative Engine Optimization (GEO) as the successor to SEO, but the rulebook is mostly empty. There is no clear advertising market, no settled mechanism for showing up in answers, no agreed methodology for measuring presence. It is wild-west territory. The companies that figure it out first will own a layer of customer access that most banks are not even tracking.

The closing reflection: who gets left behind? #

Theo’s final point is not technological. The industry conversation about AI is overwhelmingly about efficiency gains, streamlining, doing more with less. Layoffs at the largest banks are running into the tens of thousands. The smaller institutions are not far behind. Theo’s question is whether the design conversation is balanced — whether the same energy directed at automation is also being directed at the populations who do not want to interact with AI, the workers whose roles are being eliminated, and the communities that may simply not benefit from the model the industry is building toward. The implicit warning is that shareholder value and sustainable business models are real concerns, but they are not the only concerns, and an industry that forgets the second category eventually loses the first.

Why listen: #

This episode is the clearest articulation of where the strategic frontier in banking has actually moved. For founders, product leaders, and bank executives, it offers a structured map: where to focus over the next two to three years, where to leave humans in the loop, and how to think about competing in a world where the customer’s first financial advisor is no longer a person or a brand, but an AI model. Theo’s framing is grounded — fifteen years of fintech analysis, three books, and a consistent focus on the populations the industry tends to overlook. The episode is short on hype and long on the questions operators are not yet asking themselves.