AI Breaks Every Moat in Fintech

Welcome to Fintech Brainfood, the weekly deep dive into Fintech news, events, and analysis. You can subscribe by hitting the button below, and you can get in touch by hitting reply to the email (or subscribing then replying)

Come meet me, and your favorite Fintech leaders, operators, and founders in Miami at Fintech Nerdcon 👇

Hey Fintech Nerds 👋

Chime filed its S-1. Nik Milanovic describes it as “GTA6 for Fintech Nerds.” We got GTA6 for Fintech Nerds before we got GTA6 😭. I dug deeper into the numbers to find the story you’ll find in. Things to Know 👀

Nubank grew revenue by 40% YoY, has 118m users (up 19%) but shares are down. Markets are weird, but they’re also worried about capped upside. More in Things to Know 👀

News hit that Klarna “is hiring human agents again to ensure quality.” The cynics loved this story. Klarna says, AI Agents now do the work of 800 people; they’re hiring another 100 to layer quality over that. Finding that “bite point” for AI is nontrivial. Which leads me neatly to this week’s Weekly Rant 📣 AI will eat every moat in finance.

This is a bumper edition. Grab two coffees and a snack. Click “Read Online.” It’s Brainfood time.

Here's this week's Brainfood in summary

📣 Rant: AI Will eat every moat in Finance

💸 4 Fintech Companies:

1. Peek Money - If PFM wasn’t cringe and used better AI

2. Firmly.ai - Agentic Commerce for LLMs and Merchants

3. Duna - KYB Orchestration (AKA Middesk for Europe)

4. Mintago - Employee benefits and wellbeing platform

👀 Things to Know:

1. Chime files its S-1 breaking the IPO glass ceiling? 

2. Nubank crushes earnings and there’s lots to learn

📚 Good Read: Pre-Approved vs Pre-Qualified

Weekly Rant 📣

AI kills every moat in finance.

If it takes weeks to build products that would have taken months or years, building a “finance app” with features won’t differentiate you. This destroys moats, and AI can make delivery happen 100x faster, but it can’t replace taste, experience, the need for security, and the human touch. 

• Fintech App Peek Money was built in less than a month. It now has 5k paying customers and claims 7x higher retention than most PFMs because the vibes are immaculate.

• Klarna announced this week they’re re-hiring humans, after famously saying AI did the job of 700 people. They’re now hiring again to improve quality. Klarna told me their AI now handles the volume of 800 staff, but the new customer support teams do more product-focused work.

• The same week saw Jony Ive on stage at Stripe sessions rebuking the Silicon Valley obsession with OKRs, metrics, and “engagement.” 

Experience is your differentiator. Print out these lessons and pin them on your wall 👇

This raises all kinds of interesting questions and implications. 

• What happens when you someone can build your competitor in a weekend? AI is making it cheaper than ever to build a competitor to a billion dollar business 

• What happens when all commerce happens through a chat or voice interface? Agent-first commerce is remaking merchant and the shopping brand presence moat

• What happens when all finance happens through chat or voice instead of mobile? Agent-first could remake customer interaction from default mobile to, default chatbot?

• You have four remaining moats. Taste, context, security, and the human touch. Understand how each works with AI, not against it,

• Dismiss it at your peril, because this isn’t if, it’s when. Today’s tiny, niche, not very complete products are tomorrows giants.

We can build Neobanks and PFM in a weekend.

A couple of weeks ago, Sherry Jiang tweeted about building Peek Money in 3 weeks using Claude, Copilot, and GPT-4. Not a prototype. A full stack with web and mobile apps, core banking integration, a website, and happy customers.

This isn't an outlier.

YC interns are shipping functional apps in hours. 

College kids are launching AI products between classes. The person who would have been your toughest technical hire last year is now your toughest competition this year.

This is the birth of the vibe-coding founder. There are classes on how to create your business over the weekend.

This changes the economics of the industry. Again.

(Here’s the o3 rationale for those unit economics, which track with my overall experience)

Branch-first banks have higher costs but more revenue options.

• Spend billions in CAPEX (and OPEX), 

• Cost to acquire (CAC) is higher because of the branch infrastructure. 

• Cost to serve is higher too with legacy tech and a larger staff. 

• But they balance this with more profitable customer segments, and using customer funds to sell complex lending products, mortgages and more.

Mobile-first Neobanks changed the cost structure

• CAPEX and OPEX $100m’s per year not billions

• Cost to acquire (CAC) is lower because it is usually digital-only

• Cost to serve is lower with cloud native infrastructure and automation

• So they often serve other segments and need to expand top lines

AI-first becomes the great unknown

• CAPEX and OPEX could be 10x or 100x less than mobile-first

• But CAC, CTS and revenue are simply unknown

• We’re in an era where its easy to sign up for a new subscription product, but we don’t know the long-term churn rates, or even if this is a direct competitor to incumbents.

The move to AI-first Fintech could be another changing of the guard. 

Just as Mobile first Fintech has slowly started to eat market share of incumbent financial services, AI-first may do the same, possibly faster.

And, while AI demolishes barriers to entry, it's simultaneously rebuilding how we interact with every financial product. Take commerce, for example.

Commerce is becoming Agent-first

It’s now possible to buy items directly from your LLM (like Perplexity). 

1. You can buy items directly from Perplexity searches with virtual cards. Stripe merchants can recognize a user from Perplexity and issue an AI Agent with a virtual card. (Explained more in this previous Brainfood).  At Stripe sessions, they also added their “Order Intents API” giving merchants tools manage the order lifecycle for AI Agents as customers.

2. Merchants can also create mirror sites for Perplexity if they don’t use Stripe. This is a really interesting model. Firmly uses a card-on-file to make transactions work between any LLM and any merchant. They create a mirror site for the merchant, partnering with platforms like Perplexity, social media platforms, or anyone who sells ads and routes that via their mirror site + card-on-file setup to complete a purchase. That’s how you can “shop the ad” or “shop the LLM.”

Any LLM (or platform) is a shopping assistant. 

These models are early, but they will quickly become mainstream. Visa and Mastercard have both announced plans to make a Tokenization standard for AI, and you can see how this follows a pattern.

Agentic commerce is following a well-worn path of payments tech shifts. In the early days of mobile payments. NFC wasn’t available at the physical point of sale, checkouts weren’t optimized for mobile. Apple, POS manufacturers and companies like Square, Stripe, Toast, Adyen closed that gap, and eventually it became mainstream.

Stripe, Firmly, the payments industry, and Visa/MC will eventually close that gap for agentic commerce.

And when they do.

Shopping, banking, and finance will happen in conversation, not your app or website.

Consumers will have their own Agents

Maybe we finally get a private banker in our pocket?

While some corporate treasuries and high net worth individuals have full-time, professional staff getting the most from their bank, that isn’t available to the mass-market.

The agent doesn't want your cross sell into a checking account. It wants to fulfill its user's financial needs. 

When your customer primarily interacts with an AI agent that manages their money, what happens to your careful brand positioning?

The most interesting net new revenue opportunity and category creation opportunity. 

• Will a Robinhood or Neobank offer this?

• Will it comes from the big LLMs like ChatGPT?

• Will it be new subscription products like Granola or Fyxer?

In the UK, the regulator is looking at how their policy guidance on financial advice might need to shift in light of AI Agents. In the US, where regulatory innovation is more open, might we see the same? In Europe, we have “regulatory simplification.” Perhaps even there, we might see movement (if GDPR doesn’t kill it first).

It’s never been easier to launch something. 

But the question remains. Is that something good?

Your Four Moats in the Age of AI

Will it last the test of time? To answer that we need to consider

1. The importance of taste

2. The importance of context

3. The importance of security

4. The importance of the human touch

1. The importance of taste.

It’s almost unfashionable to talk about “taste” at the same time as “corporate treasury management.” 

It’s like those two things don’t fit.

And yet, taste is everything in finance. 

The difference between Stripe and Worldline is in the decisions Stripe makes about their product, even when no OKR can justify it. These "invisible details" that Ive referenced define quality relationships, and they're hardest to program. 

Klarna is actually a model to learn from. When Klarna realized cost-cutting had gone too far, they acknowledged the same thing Jony Ive was pointing to: joy and care matter, even when they don't show up in KPIs. 

The opportunity in 2015 was to create new brands that felt like they were actually there to help you, unlike the banks embroiled in the financial crisis. To create a new form of trust.

The opportunity in 2025 is for anyone to create one of these brands.

The early winners in the AI-First era like Cursor, Loveable, Granola et al, are scaling to $10m, $100m + faster than any category or tech boom in history. And I’d put it to you what they all have in common is taste. 

An LLM can listen to and replicate every jazz solo ever. But it can’t feel the silence or understand the intentional mistake.

That’s the role of taste. That’s your role.

2. The importance of context.

Finance that knows you before you even ask is the dream.

If AI had access to all of your accounts, tax data, emails, it could start moving money for you, managing your savings, and paying bills. We’d get self-driving money, and self-driving admin around the money (which is a much bigger schlep)

To do this AI needs context about you.

Think of AI as an alien that knows everything about the universe, but nothing about what you wish to achieve. The skill of using AI today is being able to give your AI enough context to be capable of performing a task well on your behalf. A lot of this gets wrapped into really great products for a subscription.

This is changing.

OpenAI has launched its memory feature, in which ChatGPT will recall your preferences from previous conversations with the goal of becoming more helpful over time. You can also ask it to store or remember specific details like times, locations, or even your address. This is how AI gets context, and how we begin to unlock self-driving everything.

There’s just one problem.

Giving AI the keys to your digital life could also be a privacy and security nightmare.

What happens when your vibe-coded financial advisor is hacked and gives away all of your company data? What happens when the context your LLM has built up about your entire life becomes a privacy nightmare?

3. The importance of security.

Security means things to people like

• That my money is safe

• And that if it's stolen, I’ll be made whole

• Someone, somewhere, is working realy hard to keep things that way

As we saw with Banking-as-a-Service and Synapse/Evolve, that’s not always a guarantee. Plenty of fintech companies appeared as safe as their bank counterparts with logos from the FDIC displayed prominently. 

But when Synapse went bankrupt, customers of Yotta, Copper and other Fintech programs suddenly couldn’t access their life savings. That involved an FDIC-insured bank and a well-funded fintech company. 

What happens when it’s a much smaller team with less investment?

The open question about vibe-coded apps today is security. PCI/DSS, SOC2, and all of the things mature companies worry about are a much bigger problem as you scale to millions of users and start moving money.

4. The importance of the human touch / in-person.

Have you noticed? In-person is back. 

Maybe it’s a reaction to the pandemic, but people are hungry for in-person connection, creativity and problem solving. This has several implications.

• Customer support is a creative function: A lot of CS work has historically been reporting, feedback loops, and script following. Great CS teams solve problems and even drive the product roadmap. Klarna has leaned this way.

• Branches need to be problem solvers: Chase is opening branches and that’s still driving growth for them. But what's the point if all branches do is a worse version of online banking? Solve complex problems, help vulnerable populations, and give the staff the power to do that (just like CS).

• In B2B finance, deals still close over dinner: Term sheets get signed because two people connect over shared backgrounds or mutual passions. The capital markets still run on relationship managers taking clients to Yankees games.

• Events create shared experiences that digital can't replicate. Conference side conversations create partnerships that would never happen on Zoom. Charisma still moves mountains in ways algorithms can't measure.

If you’re hyper-online and following AI news its easy to forget that we’re still early. Most people don’t use LLMs or AI in their everyday lives, and if they do, it’s very basic.

Don’t dismiss AI

It’s too easy to dismiss something as small and a fad.

• It’s a bubble: perhaps, but then what happens? After the bubble comes the steady rise of new Giants. The “Mag 7” were tiny (or not yet founded) in 2001.

• It has no taste - what if we add that in? What stood out to me about Peek Money most wasn’t just that it was made in 30 days. It’s that the vibes were immaculate.

• It has no context - That’s changing. Open AI’s memory feature will become standard. It will get wild if Google ever figures out how to connect your digital life.

• It has no security - That’s something we can work on. The new thing is never viewed as secure. Mobile, Cloud, “big data” are now enterprise defaults that used to be “not secure.”

• It has no human touch - That’s what we’re here for. Coming of age is about finding your voice, your likes, dislikes, and unique takes. Use it.

As Jony Ive might say, “Joy and humour has been missing” from technology. 

When we don’t have to spend as much effort on the engineering, when ever part of the value chain of building is 10x faster, we get to build better things. With more taste.

The opportunity is to use AI to increase capacity for the human elements that resist automation. Greatness is in the details. That’s perhaps why I view Klarna's moves differently to others. It is a progression into a world where technology amplifies humanity rather than replacing it.

We don’t have to bland wash everything to be “pro.” When they could be delightful.

We are the AI-First Generation

More accurately, you are.

With every generation comes new companies, new technologies and new ways of working. It’s easy to dismiss anything as a fad, as a passing trend, and in doing so miss the shift.

As Sam Altman explains here:

AI is already proving it can improve education and therapy outcomes; it can even discover novel ideas. 

It hasn’t yet proven it can help consumers and businesses reach better financial outcomes. 

That to me feels like our opportunity as an industry.

As Chime goes public with a mission focused on better outcomes, I’m wondering, who’s building the next Chime? What will Chime itself build with AI?

Chime will build products that suit its mission, possibly faster and cheaper. So will their competitors, so will new upstarts. The time to market and cost to market for new ideas will shrink. There will be compliance and security issues in the process, but we’ll also solve them.

So that in aggregate, we can build a much better financial system.

What will you build?

That’s uniquely you?

ST.

4 Fintech Companies 💸

1. Peek Money - If PFM wasn’t cringe and used better AI

Peek Money tracks spending, vibe checks decisions and suggestions, and builds to-dos and notifications to head towards a better financial life. It is positioned as a “friend that understands money without judgment.” Users connect their bank accounts, tracks your money, provides a weekly “vibe check,” and helps you “slay goals that don’t give you the ick”

🧠 This app was written by a first-time vibe-coder in a weekend. The dividend of new developer tools is massively increasing the scope of who can be a founding engineer. That means apps can serve new audiences, and the cost of creating goes to near $0. As a solo founder business, if this PFM hit a few $m run rate, it might never need VC. It speaks to an entirely new class of financial services experiences, that will raise the bar on marketing and UI. I’ve generally been sour on PFM as a category, but when the cost of creating is so low, that picture changes. It noticed too, that the average reviewer age is 23 to 26. New audience unlocked.

2. Firmly.ai - The Agentic Commerce Platform

Firmly “makes content shopable” by working with Merchants to understand their real-time inventory and pricing, and make that available to chatbots, social platforms, and ad-tech solutions. So you’d be viewing a social story, and see an overlay of the dress, and have the option to buy for example. Merchants get to sell in any chanel, distribution partners get a new revenue source. 

🧠Agentic Commerce makes all purchases embedded everywhere all at once. Instagram and TikTok have done a good job of social commerce, but the network effects have been limited. The real story now, is the incredible growth of direct purchasing within the chatbot experience. Perplexity is a Firmly partner (and Stripe). So you can research a new gift for fathers day in the chatbot, compare options, and buy them, without leaving the chatbot experience. That’s very different to todays model and changes the entire funnel from marketing to payment to loyalty. AEO (AI engine Optimization) is growing exponentially fast. What happens to merchants and commerce when their website isn’t what they’re optimizing any more? What happens to fraud? Loyalty? Woof. Questions!

3. Duna - KYB Orchestration (AKA Middesk for Europe)

Duna 10x’s business onboarding speed, and claims a 37% increase in conversion of business customers for companies like Qonto, Adyen and Plaid. They have “20 KYB modules” from KYC to AML across 210 global registries and local tax authorities. They provide ongoing monitoring, re-KYC, policy management and legal agreement management.

🧠Big name customers, gorgeous website, kudos. KYB simply isn’t solved across Europe. I love how elegantly they’re describing some of the solutions like “Re-KYC.” Duna is largely competing with manual work or in house build. The wave of business focussed neobanks are marketplaces in Europe had to do a lot of this orchestration stuff themselves. As a principle, generally I find orchestration has a churn, revenue upside risk (most of the economics are passed through the 3rd parties). KYB could be an exception. It’s such a big cost center, that if you can move the dial on the onboarding and unit economics, you’re in great shape. 

4. Mintago - Employee benefits and wellbeing platform

Mintago helps employees in the UK save on childcare, groceries, mobile subscriptions and technology through salary sacrifice. This tax advantaged way of buying every day items can be offered by employers to attract and retain talent. It also helps with CFO firedrills like employees with emergency costs and building a more resilient workforce. 

🧠These platforms are great, but I feel like they need to be baked into payroll, ERP or banking platforms. HR budgets will only ever go so far, but if you move them into the CFOs office they become a different business case. Mintago does partner to offer “pension hunting” to help users consolidate and compound their retirement plans, but if I were them, I’d be building channel partnerships aggressively with Monzo, Revolut, Payhawk, Intuit, Xero etc. 

Things to know 👀

1. Chime files its S-1

The most well-known Neobank in the US, Chime, has filed its S-1 with the SEC to go public. Key numbers, 8.6m members (82% growth of actives since Q1 ‘22). $121bn purchase volume, $251 ARPAM (revenue per active member), 67% are primary account holders. They ended 2024 with $1.67bn revenue and an EBITA of 4% (no adjustments!)

🧠 We might get our 2025 IPO yet. Klarna and Circle paused theirs. Chime is being more brave here. 

🧠 Chime has had to grow thick skin. As the consumer-facing fintech brand, they often take the heat for the whole sector. “Not very profitable,” “all built on Durbin interchange,” “not primary accounts.” Well, the S-1 is showing otherwise.

🧠 They are the primary account, and product expansion is showing promising signs. This is a critical push-back against banks that saw it as a secondary account. Monthly attach rates are 69% for savings, 54% for pay anyone, and 37% for credit builder.

🧠 The active customers are profitable. I couldn’t find the exact amount, but they list a radical cost-to-serve as a competitive advantage. They have no branches, built their own core (ChimeCore), and have reduced fraud rates by 29% since 2022. 

🧠 They’re serving a growing segment. As a cost-of-living crisis hollows out the middle class, the mass market is increasingly becoming a Chime customer. The window of “profitable” for big banks is shrinking as the FICO score they need before customers “fit their window” has inflated steadily for decades.

🧠Chime has intentionally avoided going heavy on high-interest credit because of its mission and segment. They believe payment products better suit customers who make less than $100,000. These customers account for 35% of deposits but 75% of transaction volume

🧠If the Durbin Exemption falls, Chime will struggle. There’s political pressure to remove the Durbin exemption that benefits companies like Chime. But that would be a sad day. (Is it a coincidence that Senator Durbin has announced his retirement?)

🧠SOFI is probably an interesting alternative model. With 10.9m members, they did $771m in Q1 2025 and 35% growth in revenue from lending. They have a different mission and serve higher FICO scores, but its fascinating to see so many at-scale companies battling it out. The US doesn’t have a mega dominant player like Nubank. Speaking of…

2. Nubank crushes earnings 

The Stats should impress, but markets were concerned by a dip in earnings that Nubank says is due to “higher credit loss allowances and increased interest expenses in Brazil, reflecting the rise in the SELIC rate”

• Revenue: ↑40% to $3.4bn

• Profit margins: 40% (slight dip due to regulatory requirements)

• Return on Equity: 27% (still exceptional)

• Customer base: 118 million (with 83% monthly active)

• Revenue per active user: $11.20

• Efficiency ratio: 27% (category-leading, but Brazil is a cheat code)

• Cost to serve: $0.70 per customer

🧠 Why Are Shares Down Despite 40% Revenue Growth?

Markets expected even more. This slight earnings miss triggered a selloff despite numbers that would make most banks drool. The worry is there will be more pressure on earnings. But this is nearly a RULE OF 80 COMPANY. 

🧠 That cost advantage is almost unfair. At 27% that's incredible 

• JP Morgan is at 52%

• BofA at 53% 

• Santander at 42%. 

It’s going to take a sustained, multi-year commitment to catch Nubank. And I don’t know that it’s possible with branches. The bigger threat might be bottom-up, AI-first challengers.

🧠 The Ceiling Problem

• Nubank dominates Brazil (96M customers) with a proven cross-sell engine. 

• Their expansion is progressing in Mexico (11M) and Colombia (3M), but then what?

Nubank’s biggest risk might be a capped upside. Where does future growth come from when you've conquered your home market?

🧠 North America: The Next Frontier?

Could Nubank's approach work in the US and Canada? They operate differently than SOFI, Chime, and Cash App, functioning more like a hyper-efficient digital bank (that somehow also sells eSIMs and crypto).

I'd argue their biggest competition in North America might actually be Robinhood - they're getting aggressive on both product development and market expansion.

Good Reads 📚

1. Pre-Approved vs Pre-Qualified

These two words sound the same, but are meaningfully different. The one you pick depends on your target customer experience and opportunity size. Underwriters often “pre qualify” 75% to 80% of businesses, vs “pre approve” closer to 60%. 

Pre-qualification misses several steps like credit checks, and can lead to disappointing a large proportion of businesses. However, companies “pre-approved” have a ~90% chance of getting credit, meaning less disappointed customers. People often assume pre-qualified is better because it hits a bigger audience, but on net, often leads to less loans overall. 

🧠Honestly, I had no idea in the difference in these two terms. Luke built the credit programs at Intuit and Square. If anyone knows B2B credit, Luke is the guy. And he’s going to run a masterclass at Fintech Nerdcon you won’t want to miss!

Tweets of the week 🕊

Brale is becoming the on-ramp and off-ramp between off-chain and on-chain for the banks. Ben gets it.

There’s also another Fintech IPO happening.

That's all, folks. 👋

Remember, if you're enjoying this content, please do tell all your fintech friends to check it out and hit the subscribe button :)

Want more? I also run the Tokenized podcast and newsletter.

(1) All content and views expressed here are the authors' personal opinions and do not reflect the views of any of their employers or employees.

(2) All companies or assets mentioned by the author in which the author has a personal and/or financial interest are denoted with a *. None of the above constitutes investment advice, and you should seek independent advice before making any investment decisions.

(3) Any companies mentioned are top of mind and used for illustrative purposes only.

(4) A team of researchers has not rigorously fact-checked this. Please don't take it as gospel—strong opinions weakly held

(5) Citations may be missing, and I’ve done my best to cite, but I will always aim to update and correct the live version where possible. If I cited you and got the referencing wrong, please reach out