- Fintech Brainfood
- Posts
- Fintech 🧠 Food - Moats in Fintech
Fintech 🧠 Food - Moats in Fintech
Plus; Why PSD3 matters, Visa's Pismo acquisition & GenAI field notes by BCV
Hey everyone 👋, thanks for coming back to Brainfood, where I take the week's biggest events and try to get under the skin of what's happening in Fintech. If you're reading this and haven't signed up, join the 31,754 others by clicking below, and to the regular readers, thank you. 🙏
Hey Fintech Nerds 👋
Always a big week in Fintech, with Apple leaving Goldman and Visa acquiring a major payments provider in LATAM.
Landscape shifting. Possibly tectonic.
Yet, if you zoom out, it’s capturing existing trends, not emerging ones.
Occasionally, I find myself ludicrously excited by things that people shrug at.
And I think that’s because the stuff that flips the table upside down often starts out looking technical or insignificant.
Here are a few examples from this week.
The European Commission published PSD3, the UK Law Commission published its report on digital assets, and I saw the founder of Compound.Finance is building a regulated tokenization startup. (All covered in 👀 Things to Know).
Often these small procedural things look not newsworthy in isolation but look like patterns to me.
Europe will have more open payment rails for Fintech companies, digital assets will become legitimate, and tokenization will be the new platform for financial markets and assets.
I don't care if the tokens are issued by a central bank, commercial bank, or a DeFi protocol. Using a token, a long string of numbers to represent assets and cash, will be the default.
And in case you missed it, from iOS17, you’ll be able to side-load apps onto iOS devices. (One of our 💸 4 Fintech companies this week may take advantage).
Regulation and platform shifts are game changers.
If we want to know where the world might be in 5 years, we should study them.
The question is when to move.
And who wins?
Speaking of who wins.
This week's 📣 Rant is about where the natural moats might be in financial services. Can Orchestration be a moat, or is it just another ZIRP (zero interest rate phenomenon)?
PS. Huge news in my day job this week. We announced the launch partners for SardineX, the fraud data-sharing consortium. Partnering with Visa, Airbase, Novo, Blockchain.com, and many more we can't yet announce (trust me, they're enormous). This is meaningful. SardineX is open to all, compliance-first, and ready to go. If you want to learn more or get involved, reply to this email or fill in the form. It would mean the world to me 🙏
Here's this week's Brainfood in summary
📣 Rant: Moats in Fintech Infrastructure
💸 4 Fintech Companies:
Noodle.Shop - The Solopreneur operating system and AI
Centerfin - Family Office Assets for Normies
Vesta - New Mortgage Lending Origination System (LOS)
MayDay - Accounting automation for multi-entity SMEs (UK)
👀 Things to Know:
PSD3 is coming - Full Proposals published by European Commission
🥊 The UK Law Commission released its report on digital assets.
🥊 Blackrock and now Fidelity have filled Crypto ETFs, and DeFi projects are spinning out for registration.
📚 Good Read:
Weekly Rant 📣
Moats in Fintech Infrastructure
Should you orchestrate or own IP?
If it exists in finance, some Fintech company, somewhere, is orchestrating it.
The last 24 months in Fintech have been dominated by orchestration platforms and features. Payments, identity, fraud, compliance, underwriting, accounting, and cross-border payments.
All of these tools are useful to someone, some of the time.
But do they right the right to exist? And if so, what is their right to win?
As the market turned, we've seen BaaS companies either going out of business or being acquired. The unanswered question about Orchestration is whether it can deliver revenue and SaaS margin or is it mostly nice to have? Is it only useful for smaller companies, or can it justify its existence for larger companies too?
Runway gets shorter by the day.
Revenue is vanity; profit is sanity.
Is the only way to unit economics to build down and own some core bit of intellectual property at high margins, or does finance need a platform?
There are two arguments.
Orchestration is not Middleware. It's a platform that adds much more value to users than underlying providers close to the metal.
Orchestration is Middleware. Ultimately, incumbents adopt or acquire the feature because it cannot deliver profitable revenue.
Picking a side in this debate is almost a battle of religious belief and conviction.
I rarely view the world as a set of binary choices.
So let's dive in 🏊♀️
What types of moat exist?
Economies of scale
High switching costs
Network effects
Deep IP/Tech
Brand loyalty
A framework for orchestration vs close to the metal
“Orchestration” can be a moat
Orchestration is more than aggregating some services and providing workflow.
It adds value by fixing underlying infrastructure without opex spend
It can capture value if it has economies of scale; it can create network effects
Orchestrators may struggle with margin and profitability unless they scale or develop deep IP (or both)
The path to winning is becoming a platform, not a tool
Close to the metal can be a moat
Close to the metal in Fintech is anyone touching a rail or regulated activity directly and solely with their IP
Providers add value by giving customers more autonomy and confidence and may pass on economies of scale through pricing.
These providers capture value through long contracts and high switching costs
Close to the metal providers have a cold start problem to overcome
The path to winning is marketing timing and execution
Do current market dynamics change the equation?
Everyone in Fintech is trying to sell to enterprises because they want predictable revenue.
We've moved from everyone being a competitor to how we partner.
Its M&A season.
International is the new battleground.
Non-digital businesses are the new “Every company is a Fintech company.”
Regulation and technology create a platform shift opportunity.
Orchestrate or go deep with a wedge?
They can both work. But they both end up going deep—market dynamics, timing, and ideal customer profile matter most.
1. What types of moat exist?
There are the business school moats and then the less tangible ones.
The least tangible moats are things like product velocity or execution. These likely fit into some mix of Deep IP, brand loyalty, and network effects. But they're also cultural. For example, I often point at Ramp as a company that gets the details right and ships incredibly quickly.
Even Apple and the modern Microsoft exhibit an enterprise version of execution quality. They make good calls and stay consistent. The financial services equivalents might be Capital One or JP Morgan. They have other things (like network effects) but are category leaders due to execution.
But to recap the moats 👇
a) Economies of scale create operating leverage. You can reduce costs by buying in bulk, supporting longer contracts with larger clients, and more complex sales. An enterprise buyer will want to support multiple markets and will want you to have detailed policies and data protection in place. A company also needs to prove it can handle enterprise volume.
💸 Finance example: Large banks like Citi, JP Morgan, and Wells Fargo offer this globally. They use their giant deposit basis and diversified banking offerings to drive profit in almost any market.
b) High switching costs (including regulatory capture). Some systems are mission-critical and touch everything in a business. Changing those can be incredibly complex. Others accommodate the law or regulations and are known quantities. The old saying "nobody got fired for buying IBM" definitely applies to some clients in some ways. If changing a provider can be expensive, career-ending, and isn't guaranteed to succeed, it's hard to choose.
💸 Finance example: Any system of record like a "core" deposit account ledger (often "core banking system") or loan origination system (LOS). Where a system stores the current customer balances, performs interest rate calculations, and reports to the regulator, that cannot fail.
c) Network effects. You have network effects if each additional user adds more value to existing users. Operating systems with more users get more developers building which draws more users.
💸 Finance example: P2P payments apps like Venmo became the verb "Venmo me" as they built out user bases and became the default. Government services like Faster Payments in the UK or Pix in Brazil play similar roles through enforced network effects (i.e., all banks support).
d) Deep IP/tech. Companies like chip designer ARM develop a market position by focusing on pushing the boundaries of chip design. Tesla has vertically integrated much of its supply chain to reduce the number of car parts and build an entirely new market.
💸 Finance example: It's seldom discussed how sophisticated the internal fraud or credit risk algorithms can be inside a major financial institution. You'll also find some of the world's best data scientists working on Wall St, trying to identify new patterns to manage risk or create alpha. These teams don't get credit for the incredible work, but disrupting this moat would be interesting.
e) Brand loyalty. Customers will continue to buy from and use a product they trust. That brand virtue could be stability, fun, or a sense of premium quality. Coca-Cola, Nike, and Louis Vuitton give different brand promises and engender loyalty.
💸 Finance example: This works in B2C as well as B2B. In B2C, the traditional banks are a virtue of stability like Chase or Wells. In B2B, often, the same message works from traditional vendors like FIS or Fiserv. Increasingly Amazon and Google have played to the more innovative message in B2B, and Fintech companies like Square or Current play to consumer segments that traditional banks left behind.
Many of these moats work together. Brands with much customer loyalty often gain economies of scale that drive network effects.
As an incumbent, the question is how to use these strengths to drive new revenue or defend market share. As a disruptor, the question is which can be disrupted, if any, and how?
2. Close to the metal or close to the customer? 🤖
There's another axis to consider in financial services.
Do you move up closer to the customer's problem or down closer to the metal?
Embedders: "Vertical SaaS" These are brands like Shopify, Uber, or non-bank, whose primary go-to-market is unrelated to financial services. They create value by solving more of the customer problem and embedding finance next to that workflow. They capture value by charging higher unit prices or using finance to drive more engagement with their product.
Services: "Consumer and B2B Fintech" Neobanks, accounting software, payment automation, treasury management. A whole layer of companies solves a problem better than traditional banks do. They sit here because they do not own or operate a banking license.
Products: "Banks and other major license holders." Insurance companies, specialized lenders, and banks fit here. They tend to hold the highest level of regulatory burden.
Capabilities: "Providers or Fintech infrastructure," whether incumbent or a startup. This is where everyone from FIS, Fiserv, and Verfin sit through to a company I wrote up on 4 Fintech companies last week. Their "moat" may vary (discussed further below)
Rails: Ways to move value. Cards, P2P, international, government, or private. Visa, Mastercard, AliPay, UPI, Pix, and FedNow all fit in this category. Typically their moat is network effects, but as the world becomes multi-rail, default digital and default global, that's changing.
If you explode this image for Fintech infrastructure, you could split the "capabilities" layer further into two.
Orchestrators
Deep IP (closer to the metal) providers
Orchestrators are closer to their customers (the products, services, and embedders) and solve more of a problem. But. Closer to the metal providers may have more economies of scale, a pricing advantage, or high switching costs.
Which is the correct answer?
It depends.
3. Orchestration can be a moat 🧱
Orchestration is misunderstood.
Some folk write it off as "middleware" or aggregation. A service that collects multiple services together and provides workflow tools in between. They claim many of these companies will have revenue, growth, and margin problems because workflow is best suited to smaller companies that don't have an existing tech estate. These businesses don't tend to own deep IP at the outset, so "how can they have a moat?"
That's not wrong. It's how many of these companies start.
But it's not where the most value is created or captured as they scale.
The key is separating the creation and capturing of value and how that changes over time.
a) What is orchestration? My definition is controversial because many companies get called orchestration, which does several more valuable things. If you combine aggregation, workflow, and automation, you can add new value to customers.
b) How orchestrators create value: Aggregation simply aggregates many other providers. It might be 100s of payment types in payment orchestration, and in compliance orchestration, 100s of providers. This simplifies integration and improves time to market.
Workflow is, as it sounds, starting a process at one provider, moving the data and outputs to another, and making decisions about what happens next. This can remove R&D and operational expenses for users, and for smaller clients, it means entire departments don't have to rely on internal engineering resources.
Automation is much more interesting. On the service, it's running a workflow consistently. One example is building waterfalls. Many Fintech companies that go close to the metal quickly discover at scale that a payment file sent to Wells Fargo breaks if you do X, but if you don't do X, then Citi will reject it. They might try one bank and retry at another (the waterfall), or they might fix the file format before sending. This removes significant Opex from their customers and creates an R&D / Deep IP moat in its own right (but it only works at scale). Crucially by increasing the conversion of payments, payments companies can also increase the revenue for their clients.
Twilio aggregates 190+ Telcos. You could build those integrations yourself and have better unit economics directly. But can you build the waterfall to manage the random errors? Can you deal with the new compliance rules that come along?
With unlimited time and budget, perhaps.
But will that distract from your core mission?
Almost certainly.
c) How orchestrators capture value. Early-stage orchestration tools help smaller companies scale faster and get to market faster. They capture an early-stage startup part of the market that close-to-the-metal providers might be too complex to integrate with or expensive to get to the market.
Often a wide breadth of offerings can create network effects. Where the more services they offer create more client interest, who need more services, generating further client interest. Platforms have a flywheel for going wide and offering breadth.
Later-stage platforms like Twilio and Stripe become entire departments for their clients. They drive economies of scale with their providers and can pass that on; they displace Opex by fixing the infrastructure and handling the changes without clients focusing on it.
You can build it, but can you afford to maintain it?
Some "orchestrators" grow up to become platforms.
But the path to getting there isn't guaranteed.
d) Challenges Orchestrators face. If an orchestrator is mostly re-selling underlying providers unless they hit scale, their margins can be thin. Their profit depends on their economies of scale with their providers or the higher unit pricing they can charge because they're adding such a massive amount of value to users.
e) Path to winning: Becoming a platform, not a tool. We often refer to workflow as a "tool" because it's designed to give the user autonomy over a previously hard-to-manage task. A platform, on the other hand, might be more opinionated. It may still have a workflow, but it has opinions about handling errors in the underlying infrastructure.
A tool might remove Opex and improve time to market. But platforms maintain an R&D advantage and can generate more revenue for some customer segments than that customer could alone.
Providing a feature-complete solution to early or mid-stage customers gives these businesses traction. Their challenge is long-term growth and unit economics. Their moat depends on their product velocity and execution, driving network effects for them over time.
If they can do this, they can build less tangible promises around their brand. Twilio has developer mindshare and is a default choice. Plaid arguably has that brand in Fintech account aggregation. Stripe has it in payment acceptance.
Workflow tools must hit these conditions to become a platform:
Remove Opex and improve time to market for customers
Generate revenue by increasing core metrics for clients
Maintaining an R&D advantage
Eventually, move down the stack and own core IP for better margins
Twilio has done this in SMS, and Stripe has done it in payment acceptance. The jury is out if it can happen in BaaS or compliance areas. That's a good thing.
Before I discuss why, let's look at the alternative path.
4. Close to the metal can be a moat 🏧
The word "core" gets thrown around by people in financial services as a shorthand for something close to the metal that's unbelievably mission-critical and hard to change.
There are certain activities required by regulation, or that touch the rails in financial services that can be incredibly hard to break into as a new provider. The risk of being a provider trying to get close to the metal is you could burn $50m and not drive any revenue. New providers rely on a shift in the external environment to drive growth in an initial customer base.
a) What is close to the metal? I define these providers as anyone touching a rail or regulated activity directly. Natural examples are systems of record (deposit ledgers or loan origination systems), issuer and acquirer processing, identity providers, and transaction monitoring providers.
b) How providers create value: Beyond simply being able to access the rails, closer to the metal, providers bear a significant responsibility to be available and help their clients meet regulatory responsibilities. Their deep IP often does the hard yards to get certification from payment networks. This helps their clients avoid significant Capex and Opex expenditures.
Legacy payment processors sell on their responsiveness and uptime. Loan origination systems sell because they're used today by regulated financial institutions. The confidence they give is that "this works because it has been for decades."
Another way to add value is to provide a regulated function like identifying users or transaction monitoring. These systems and processes must exist and be performed by law. They help their clients meet critical obligations to regulators and governments.
These providers may offer massive economies of scale and pass that on to their users regarding lower fees and breadth of capability. They might support more markets, products, and standards, making the higher-effort integrations worthwhile.
The close-to-the-metal providers can also create more autonomy and control for their clients. The service may be more complex to integrate, but customers can create bespoke functionality or solve more challenging internal use cases in return.
c) How providers capture value: Enterprise buyers typically look for longer contracts when offering a mission-critical service. A core banking platform or issuer-processor might sign 5, 7, or even 10-year contracts. This gives stability to revenue.
Being close to the metal and owning deep intellectual property means these businesses typically don't have to rely on 3rd party providers to deliver their core revenue-generating products. This creates higher margins when offering their main product.
Close to the metal providers are also often mission-critical. Changing a core banking system is like a heart transplant on a live patient - not something that should be done lightly. This creates very high switching costs.
d) Challenges close-to-the-metal providers face: Developing deep intellectual property can be expensive and time-consuming. Connecting to the rails of financial services or building regulated financial activites often requires getting the highest levels of certification. Going direct to Visa, Mastercard, or a payments network switch is not something the operators of those platforms allow without being confident you manage it.
Staying close to the metal is as hard, if not harder, than building v1. Payments platforms change their rules and change with regulations. Transaction Monitoring platforms need to be kept updated as the world outside changes. New people are added to sanctions lists, new crime rings appear, and new payment rails emerge (like FedNow, Crypto, and P2P wallets).
Enterprise clients are demanding af. Selling to a tier 1 bank or a global SaaS platform with 100s of millions of users is a gauntlet. They will need you to support multiple markets, integrate with their customer support, and comply with industry standards most robustly. Getting through the RFP is a big ask if you're a small company starting in this space.
e) Path to winning: Timing and execution. Timing the market is everything for these providers. Marqeta and Galileo appeared as a new generation of consumer Fintech companies scaling rapidly. They had a customer who was easier to sell to than legacy banks who could adopt their product and, over time, battle harden it with volume. Similarly, the various KYC and identity providers appeared around the same time, and many had consumer Neobanks as their first customers.
Building a new issuer processor is non-trivial. Typically founders of these close-to-the-metal providers have some industry background or experience. There are exceptions to this, but those exceptions are those who wrestle with complexity and execute.
Financial services are a multi-faceted industry, and getting closer to the metal offers better unit economics but creates a capex and maintenance challenge.
If they succeed, these providers have a natural moat and become hard to dislodge. But they also enter a space where it's hard to grow without M&A because clients will demand ever higher performance from the core business as they scale their enterprise.
The job becomes keeping the lights on. Leaving space for innovators.
The core banking (system of record) platforms like FIS, Fiserv, and Jack Henry are unlikely candidates for a rip and replace in incumbent banks. The interesting thing to follow is 10x Banking and Thought Machine here. Their flagship clients are rarely "rip and replace" candidates. But digital banks in international markets or incumbent banks often launch a new product. 10x Banking powers the Chase deposits offering in the UK (and likely soon their credit card offering).
5. How this impacts current trends in Fintech 🎮
There are a few topical trends this debate lends itself to.
a) Everyone in Fintech is trying to sell to enterprises because they want predictable revenue. The ZIRP phenomenon of endless new consumer Neobanks to sell to is over. The market cycle has forced companies to focus and do what they're great at.
Incumbent financial institutions won't rip and replace legacy providers for a young company. If they do rip and replace, it will be with someone established.
When I used to work at TSYS, switching issuer processors was literally called a conversion. And it could be a religious conversion. The process before go-live can take 2 or even 3 years to map data structures, accounts, and products from the old system to the new one. If a bank has 100+ credit cards in 10 markets and 6 million customers, each of those has to land in the new system and just work.
An old friend and mentor once told me that as a young company, it's far better to sell a use case to an enterprise and expand from there. Don't expect to rip and replace the incumbents. Find the wedge, find the island, and build a colony there.
Massive SaaS platforms embedding finance do want orchestration, and they want enterprise scale. This is an interesting battleground between close-to-the-metal and orchestration as philosophies. Tech companies can build and want more autonomy. Orchestration can give some of that, but can it give the unit economics these businesses look for?
Trade-offs gonna trade. Off.
b) We've moved from everyone being a competitor to how we partner. This is a dynamic of market timing that could work for orchestrators and close-to-the-metal providers. We're entering the frenemy phase of maturity. Companies hit a certain size, compete, and partner with "competitors."
The thing you used to try and cross-sell might be a distraction. The core IP might work better in partnership.
c) Its M&A season. Those close-to-the-metal providers with strong cash flow and high margins are well-placed to add orchestration as a feature. Those orchestrators doing well could acquire a smaller, struggling, close-to-the-metal provider.
d) International is the new battleground. For example, there isn't a "Middesk for Europe" because Europe is such a complicated space. Companies are building this close-to-the-metal capability, but they're early. There are also plenty of orchestrators.
Now think about this, but for the SaaS companies going global. Will they want orchestration or depth? Or both?
e) Non-digital businesses are the new Fintech companies. It started with fuel cards and SaaS companies in construction, but it's moving to just about everything. B2B providers of commodities, shipping, chemicals, manufacturing, you name it.
Will they want orchestration or depth? Or both?
f) Regulation and technology create a platform shift opportunity. PSD2 arguably created (or mainstreamed) open banking from being "screen scraping" to more legitimate. Mobile arguably made Neobanks and digital onboarding more practical, and Dodd-Frank created an instant revenue model.
PSD3 is coming, GenAI is here and could be a platform, and Apple allows the side-loading of apps onto mobile from iOS17.
The big shift it feels nobody in Fintech is paying attention to? Tokenization. All assets will be tokenized. By a central bank, by commercial banks, and by startups.
6. Orchestrate or go deep with a wedge? 🧀
Most orchestrators start to move down the stack and own deep IP.
Most close-to-the-metal providers buy, build, or partner to cross-sell the same services an orchestrator would offer.
Orchestration's advantage is that it appeals to the long tail, and if that provider can scale with their clients, they can move down the stack and begin to own IP.
Close-to-the-metal providers' advantage is their natural moat and margin advantages once they hit scale. If they can survive and grow long enough to benefit from the moat.
There's no right answer. There's just what works given the market in front of you.
As a close-to-the-metal provider, you must recognize that small, mid-sized, and non-bank clients want orchestration. Orchestration has become a must-have feature, and it's available in the market.
If you start with orchestration, you must recognize the imperative to deliver better unit economics and become a platform. Hitting scale will require developing deeper IP or adding so much value you can charge a significant premium.
There are many paths to the Buddha.
Personally? I prefer to see something that can or ideally does own deep IP, especially if its founders have a superpower in that subject area.
However, orchestration on some level is now a must-have feature if you want to sell to SaaS platforms and start-ups.
Timing when and how it does that is super dependent on market dynamics.
That's why writing about 💸 4 Fintech companies every week is so useful. The cycle time on learning and what's possible is massively improved.
How's that for a segway? 👇
ST.
4 Fintech Companies 💸
1. Noodle.Shop - The Solopreneur operating system and AI
Noodle helps creators, freelancers, and solopreneurs collect sales with payment links, sell digital content, and provides an AI chat sidekick. Users can schedule paid calls, create digital content, and build subscriptions for the user base. It's available as an App but not via the App Store, so payments do not carry the Apple tax.
🤔 From iOS 17, Apple will support the sideloading of apps which is a game-changer. This follows the Digital Markets Act in Europe passed in November 2022 that requires "gatekeeper" companies to open up their platforms to other companies and fees. Noodle is using sideloading to support payments without the Apple Tax. Will others follow, or will Apple's market pressure prove too much? Will they create a "blue bubble" like lock-in by being the default? (FWIW, I'm a Green Bubble guy til I die, sorry, not sorry).
2. Centerfin - Family Office Assets for Normies
Centerfin makes commodities and "alternatives" like private debt and equity available to the mass market. The account is professionally managed by former ex-Wall St bankers and can be used directly or as a platform for investment managers.
🤔 It's a summer of everyone-is-alternatives for low, upfront fees. I completely buy the argument from the person working on Wall St perspective. These fund structures return more than traditional portfolios but require a high minimum because they're often very manual and hard to set up. By aggregating many customers and doing one fund, there's a latent demand for higher returns that could be serviced in theory. My question is, how do you distribute it profitably? Direct as a Robo-advisor has questionable long-term upside (without diversifying). For that reason, Centerfin going via RIAs is interesting.
3. Vesta - New Mortgage Lending Origination System (LOS)
Vesta is a lending system of record that aims to allow lenders to build mortgage journeys that are 100% digital. Lenders can build customized workflows to create more efficient (and more bespoke) lending processes.
🤔 Changing "core" systems is the hardest thing to do in finance. Systems of Record rarely change in financial services; most are 20+ years old. As the name suggests, a system of record is the "core" part of a mortgage, storing critical information like the balance, transaction ledger, interest rate calculation, and reconciling all collected payments. Bankers call it the "golden source" because it represents the truth to the customer and, critically, the regulator. Being so mission-critical means changing the system, and getting it wrong can be career-ending for CEOs. This has two consequences 1) incumbent providers are sticky 2) user journeys are always held back by legacy "core systems."
🤔 Vesta's team is former Blend engineering talent, and their wedge appears to be targetting new, greenfield lenders. Everyone wants to get into lending lately, so their timing could be good.
4. MayDay - Accounting automation for multi-entity SMEs (UK)
MayDay provides a module for Xero that helps accountants run month-end across multiple entities. It helps with use cases like balancing inter-company loans, automating inter-company charges, and invoice matching across the whole group reconciled to bank transactions.
🤔 Accounting is hard. Accounting for multiple entities is stupidly hard. There's a phase of a company before it puts in a full ERP and hires a weapons-grade CFO where they might have multiple entities. The finance teams at these companies are living in a waking nightmare. MayDay has found a super interesting market entry point. I love how they've named their modules like drops "Balancer," "Recharger," and "BRAG." It's a neat way to let the features sell the product.
Things to know 👀
Cohere.io raised $3m in 2021, and customers included Ramp and CopyAI. Cohere told Techcrunch they had been using generative AI and LLMs since "before the advent of ChatGPT" to answer customer support queries. The product was used by companies like Ramp and Deel to generate support knowledge from customer interactions. It could instantly resolve up to 60% of support tickets. Customers also included Rippling, Loom, and Secureframe.
🤔 GenAI is perfect for customer support. Most customer support questions are covered in help docs or FAQs. Chatbots' hard problem was understanding the user question well enough to bring the relevant information back to them. LLMs are perfect for these highly variable queries with multiple possible answers.
🤔 GenAI is Robotic Process Automation (RPA) for Tech companies. The acquisition goal is to use Cohere.io's AI skills to drive customer workflows. Automating as many steps as possible. The fact that Ramp is using LLMs this way is a significant signal that the value add is in workflow automation. As I wrote a few weeks ago, unlike RPA, LLMs are very good at the messy, variable, but low-value tasks humans spend time on. Ramp's product ethos has driven product innovation into those tiny details that make a massive difference. It makes sense they'd lead the charge on this use case.
🤔 The Ramp mafia wins the day. Picture this, a team of ex-Ramp employees leaves, founds a company, sells their product to their old company, and then gets acquired by it 2 years later. That's pretty rapid. Zooming out, though, that sort of makes sense to me. Here's a set of strong relationships, a proven product, and an opportunity to do much more with it if acquired. It looks like a win all around, although I wonder what happens to the non-Ramp customers of Cohere who are now in "maintenance mode?"
🤔 Cohere got a new name for its existing customers. The solution looks to be Cardina, the old Cohere website repackaged under a new name. Which makes this look like an acqui-hire to some extent. There's a way the Ramp team and Cardina branch out on separate paths, but what will drive it forward without the founders?
2. PSD3 is coming - Full Proposals published by European Commission
PSD3 will create rules to ensure Fintech companies have access to payment systems, common rules on open banking interfaces, and protection for consumer data sharing. Another rule will ensure Fintech and financial services companies can use a common IBAN interface to detect and prevent fraud.
🤔 Access to payment systems is a huge win for Fintech companies. Systems access across multiple markets vary wildly, and while there are aggregators, even many of the Open Banking companies have specialized in regions to deliver Open Banking initiated account-to-account payments. Open Banking platforms will only have to re-authenticate users every 180 days instead of 90.
🤔 Europe is a regulatory superpower. Its requirement for consumer protections like disputes and refunds could be a trend replicated across all push payment types globally.
🤔 Will we finally see true open finance? I often complain that Europe has open checking, not open finance. There's a long way to go with estimates placing the final law completed in 2024 and going live in 2025. Time to get your Government and Reg relations hires in Fintech companies 👀
🤔 Europe's Open Banking market is fragmented. A consistent Open Banking framework would move closer to the goal of a single market for consumer data and payments.
🤔 What standards would Europe build? The UK has consistent rules, standards, and APIs, as do some other markets, but it's very uneven. The UK built a common interface through the OBIE (Open Banking Implementation Entity). Might that be a standard Europe chooses to use or diverge from? Since Brexit, the UK hasn't been in the room but will soon have "observer" status in Europe on financial services. This could be helpful since it led much of the financial services activity for 3 decades.
🤔 Fraud data sharing will move from nice to have to required by law. Push payments are more affordable than cards, but they risk users being scammed into "authorized push payment fraud" (APP Fraud). This is a top priority in banks, Fintech companies, and regulatory agencies. Another welcome addition is that all payments companies will be required to use an IBAN-based confirmation of payee to match payer names against account numbers.
🤔 Will there be a pan-European data-sharing consortium? Everyone is launching a data consortium in the US, with Plaid and now Sardine* announcing it in the past week. Pay.UK also announced an effort in the UK just yesterday. The differentiator will be critical mass and effectiveness. Getting people to say they want to join is easy; the hard part is putting in place the relevant legal and regulatory safeguards for full buy-in. And getting the incentive structure right to create a network effect rewarding participants to not create a walled garden.
Quick hits🥊👀
🥊 Visa bought a payments platform focussed on LATAM. 🤔Cloud-native payment platforms have proven themselves in other markets. With clients like Citi, Revolut and Nubank it’s build a tremendous market presence. LATAM is seeing a wave of new banks and Fintech companies who might need this solution. For Visa this increases their likelihood of winning more card programs and growing their network.
🥊 The UK Law Commission released its report on digital assets. The law commission says a token can be three things 1) A thing of value, 2) a register of interests (instead of a separate ledger or database), and 3) a claim on obligations described in the token, like access to a performance or venue. This makes tokens a third category of thing. Not a debt or possession but a novel thing. 🤔 UK common law is considered the most flexible and "fair" globally. It is used in the vast majority of global trade for that reason. The implication is tokens will have a level of legal certainty in UK law at a much more significant level than any international competitor.
🥊 Blackrock and now Fidelity have filled Crypto ETFs, and DeFi projects are spinning out for registration. 🤔 Capital markets will be tokenized from both sides. Incumbents will turn existing assets into tokens on things that look like a database but interoperate externally. DeFi entrepreneurs will build regulated financial products. This will create a platform shift and a possible vendor landscape shift in financial markets.
🥊 Railsr lost its Lithuania license. 🤔 Railsr was once a major Banking-as-a-Service provider but is now fully bankrupt. BaaS in Europe had a significant revenue problem after PSD2 capped interchange on debit to 0.3%. The question in my mind is what comes next. Stripe is moving their BaaS offering to Europe. Will other US players follow, or will the next generation, like Swan and Weave, change where the value is?
Good Reads 📚
AI is heavily used in financial services today (with Fraud detection the #1 use case). It is better than existing methods at question answering, summarization, and information extraction, and crucially, it can generate "new" content. It's also bad at math, don't use it for revenue forecasting, but it is "like sand fills gaps in a glass" GenAI can fill gaps left by traditional AI in finance. GenAI is a new platform because it understands language, no matter how you communicate with it or which device.
I loved this term GenAI will make financial services bots "more emotionally available" to customers. *chef kiss* This technology is at the "peak of inflated expectations." In reality, what we're seeing today is low-hanging fruit, not the creative value-add.
🤔 This note meaningfully carries the Fintech and GenAI conversation forward. Saying GenAI and writing about it is an attention hack. I love that Sarah and the team have created frameworks to consider what to build and how. This is *the* Tech subject. Investors want to hear about it from Port'co's, and Execs want to sound smart on the golf course. Instead of FOMO'ing in, know what it's good at, know your goals, and then execute.
🤔 The core thesis that GenAI is not disruptive and the primary beneficiary is incumbents should be getting more attention. This should be a controversial point, but I agree with it. A few weeks ago, I wrote that it reminds me of Robotic Process Automation (where the biggest customers are enterprises). The difference is it's really good at highly variable tasks.
🤔 Turn that last point on its head. Every growth-stage Fintech infrastructure company is moving up into enterprise (and incumbent) sales. Displacing legacy providers with banks or incumbent financial services companies is nearly impossible. Long contracts and a high risk of failure mean incumbent providers have a massive advantage (and it's a good market for M&A with smaller companies burning runway by the day). The right GenAI feature could be the wedge into this market for new providers to displace legacy. While GenAI itself isn't disruptive, the almost-instant cost saving is. And hey, RPA is a great business. Instabase and UiPath are doing just fine.
🤔 The term "AI" is throwing people. When you think "AI," you think binary, 1s and 0s. True or false. Deterministic. GenAI isn't that. Alex Komoroske always talks about GenAI as "magical duct tape." The metaphor describes Gen AI as an imperfect tool, but it's good enough and way better than leaving something broken.
🤔 Take a bow, Sarah and team; this is stupidly good (and low-key makes me want to take a month off to go write something with a small squad).
Extra credit 🤓
1. Giving GenAI Eyes and Ears by Ethan Mollick (The best writer on GenAI there is, with some kickass use cases).
2. Four Visions on the Future of Money by Oliver Wymann (I love this framing, great mental models, and pictures, I wish I had seen it before last week's Rant)
3. The history of Capital One by Net Interest (Capital One is a God Tier financial services story. If you know, you know.)
4. Defining Bank Sponsorship by Walt Cox (Terminology in finance is confusing, this is the ultimate guide to what is "Bank sponsorship" in a US context)
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 :)
Disclosures: (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 * (3) Any companies mentioned in Rants are top of mind and used for illustrative purposes only. (4) I'm not an expert at everything you read here. Some of it is me thinking out loud and learning as I go; please don't take it as gospel—strong opinions, weakly held.