- Fintech Brainfood
- Posts
- FinCEN reports $212bn identity related suspicious activity
FinCEN reports $212bn identity related suspicious activity
Plus; The AI-First Fintech Company, Wise crushes earnings *again* & From the Big 5 Megabanks to the Big 1.
Hey everyone 👋, welcome to Brainfood, the weekly read to go deeper into Fintech news, events, and analysis. Join the 37,542 others by clicking below, and to the regular readers, thank you. 🙏
Hey Fintech Nerds 👋
FinCEN reported $212bn of identity-related suspicious activity in a new report. KYC is not a checkbox process but has been treated that way for far too long.
There aren't a big 5 banks in the US anymore. There's a big 1. When Citi's revenue was down 3%, JP Morgan had its best-ever quarter.
Meanwhile, Fintech keeps crushing. Wise revenue is up 23% YoY.
This week's Rant is an almost pure stream of consciousness from conversations with a few founders who remarked, "We're amazed by how small the teams we're selling to are, even at unicorn valuations."
API-first changed the world, but what about AI-first? This week's Rant 📣 AI's all the way down
That's your Fintech Brainfood this week 🧠
Brainfood in summary
📣 Rant: AI's all the way down
💸 4 Fintech Companies:
Tennis Finance - The BaaS Partner Marketing Compliance AI
Mercoa - Embedded AP Automation
Architect Capital - Debt Funding for X-border Fintech
Wealt - Open Finance for Wealth (UK)
👀 Things to Know:
📚 Good Read: FICO could be losing its grip
Want to see more Brainfood? Gmail spoils all the fun. Click below for more.
Weekly Rant 📣
AI's all the way down
Finance teams average 70 people per $1bn in revenue managed, but that is often 10x less at digital companies.
APIs unlocked a productivity revolution.
It's common for a Series A company to have one person in their finance department and as few as 3 in Series B.
APIs dramatically reduced the number of people needed to run any part of an organization.
AI will do that again.
Departments you hire as an API
Older companies accept “good enough,” making them inefficient
Younger companies defaulted to more efficient tools
API-First enables billion-dollar scale with 10% of the headcount
AI could be the team to staff the department you hire as an API
By creating value when it needs to get done but can’t be fully automated
Most won’t unlock the value because Generative AI is a skill, not just a product
It won’t kill incumbents or jobs immediately but it will accelerate the broader trends of growth going to younger companies
You owe it to yourself to figure out how to get 10x more productive with the tools
1. API-first businesses became departments you hire as an API.
As Packy McCormack put it beautifully in this picture, if it needs to get done but it's not worth your time doing it, that's the sweet spot.
Cloud and SaaS are how companies like WhatsApp get acquired for billions, with a headcount of 18.
APIs are how Shopify generates most of its revenue from financial services without that being a core focus (via its Stripe partnership).
We're trending towards insane productivity per headcount, but rather than making humans obsolete, it moved the battleground. Young companies that adopted API-first tools compete with each other for customers. Incumbents are forced to consider API-first and cloud to compete on cost.
How will AI-first change that?
A similar but distinct pattern. Where SaaS outsources much of the department, GenAI can automate much of the work that the department still has to do.
The toil, the schlep work, the messy complexity between the APIs. That's work humans do, and they're drowning in it instead of solving the big problems.
Some basic examples from Fintech of problems not solved.
8% to 10% of low-income consumers are behind on debt repayments
~60% of consumers abandon carts at checkouts
Fraud is growing at 31% YoY (that's faster than the S&P or NASDAQ)
Financial services is the world's largest profit pool. Finance is the mission-critical part of any business: to get paid, pay staff, and do business.
Even in the age of APIs, 9 out of 10 tasks in admin functions are still highly manual. There is no one API to rule them all, meaning data collection, approvals, and status updates are still a fact of life.
The next wave could be staff you hire as an AI.
AI that can do the toil and enable a new generation of leaner businesses to be built.
Those businesses will be 10x more productive.
How?
First we need to over come our own biases and inefficiencies.
2. Old companies are inefficient
Inefficiency comes from several directions.
That's the way we've always done it (functional fixedness). The simplest answer is often the most compelling. Most people might not know another approach exists; if they do, they might lack the authority (or capacity) to change that in their organization.
High switching costs. Changing a process or vendor becomes complex and high risk at the billion-dollar scale. The older the organization, the more approvals are required to change anything. Changing an internal tool becomes a potential distraction from growing revenue, and there might be quicker wins elsewhere in the short term.
Lack of garbage collection in processes. The sheer effort to change means companies rarely upgrade their processes and tools. They add to them.
Incumbent industries suffer from this problem the most. Instead of fixing their process and tech debt, they outsource the manual work to the lowest-cost provider.
Young companies and startups suffer this the least. They can build whatever process they want, select any API, and have no biases about how it should be done.
That's changing.
Company creation is cheaper and faster than at any time in history.
Old companies are dying faster over time.
Combining these two forces puts us in a cycle of creative destruction, pushing companies to be ever more efficient. Survival of the fittest is how markets should work. But what's not apparent is that this trend is accelerating.
That doesn't mean all companies will die.
It means we need a step change in productivity.
Just like happened with APIs.
3. Younger companies are default efficient.
Financial services, until recently, was an industry of good enough.
You had to accept that some consumers would fail to repay debt, some fraudsters would get through the net, and payments would sometimes randomly fail.
Neobanks, embedded finance, and API-first infrastructure companies all materially improved this. We get digital onboarding KYC (Know Your Customer), cash flow, payroll data-based underwriting, and expense cards that work. More than anything, businesses have dashboards that become a true operating system for departments of their business.
API-first providers have reshaped the Fintech value chain.
4. API-first unlocks billion-dollar scale at 10% of the headcount
The first buyer of a new expense card, payments API, or AP automation platform was often another new, young company. With very few exceptions, the inside of most modern companies is a patchwork of APIs and 3rd parties.
The value chain and the tech stack look like fractals. Zoom in on any department, and you'll find 10 SaaS platforms and APIs running the show.
It's this that has enabled companies who are D2C or D2B to focus relentlessly on their customer and massively improve the UX or ROI. The provider ecosystem evolved with the D2C and D2B ecosystem for API-first.
This is how billion-dollar scale companies can be formed, grow quickly, and solve much more of their customer's problems.
5. AI scales human work inside a given department
If 90% of work is still manual in operational jobs, we have a long way to go.
The smaller the company, the more efficient it needs to be. What if they could hire AI agents to perform much of the manual work?
Let's talk use cases. (Or, more accurately, personas).
The AI-First Loan Officer: For SMB lending, an application process involves data collection, back-and-forth document submission, clarification, and data entry. Even the most modern KYB tools and APIs can only automate some of the collection because it's context-specific. GenAI is already in numerous pilots doing exactly this. Today. In production. Reducing ~90% of the work up until the underwriting step. That means more capacity to look at more applications and approve more loans.
Fraud and Compliance Co-pilot: Analysts must raise a suspicious activity report (SAR) if a bank system generates an alert. That includes a detailed narrative of what is suspicious and why so that law enforcement can investigate. The volume of these is increasing by 15% YoY. SAR narrative generation is one simple summarization use case. I also see disputing a chargeback, data collection, and even brainstorming what a data set could mean to go live.
The Accounting Assistant (CFA): Filling taxes, understanding tax advantages, and managing invoices for home improvement. This stuff is the long tail of the long tail. It always disappointed me that Gmail didn't do more with where email becomes automation because so much of the unstructured data is right there.
You could imagine 100s more.
But there's a problem.
APIs were always as good as your ability to feed them.
AIs are as good as your ability to partner with them.
You cannot trust the output of a Generative AI to be accurate. It makes you more productive, but you must hold agency. You have to error correct and course correct.
Today's SaaS platforms are adopting AI, and new companies are being born AI-first to close this gap.
AIs can work with messy, unstructured documents and squish them onto any screen a human could. The glue, the duct tape, and the connectivity between things that just don't connect right are still there no matter how hard you try. That's the use case.
And that's 90% of some jobs.
6. How does Generative AI actually add value?
The answer is simple yet complicated.
Modifying Packy's previous image.
The core insight of API-first businesses was that things could be better if you focus on just that thing at scale. Stripe delivers higher conversion at checkout because it sees billions of payments and engineers around the underlying complexity.
The core insight for AI-first products and businesses is automating most things that machines couldn’t do historically.
Either because it’s too messy or there is no clear process.
These messy tasks are often the most manual and time-consuming. They’re most likely to require a large headcount or be outsourced to humans, and the work is often boring and not valuable.
Companies have had to increase staff to deal with more volume in messy tasks, but now the same staff can do more. This productivity boom can be real if you know how to use the technology.
One example 👇
Take Tennis Finance as a case study (covered in 4 Fintech companies this week). They offer compliance checks for customer-facing copy for regulations like UDAAP (Unfair or Deceptive Acts and Practices). They know this is a "pre-check" and not a replacement for a full review.
But here's the most interesting part.
They also offer a fractional Chief Compliance Officer service.
Now, imagine you're a nonbank, embedding finance, and you need to get a compliance check. You could go to a consulting firm (and there are many good ones) or use an AI pre-check plus their fractional CCO.
Every senior leader's bank role will be fractional, supported by AI in the future.
Do this for every job title.
Now, imagine the productivity boost.
7. Counterarguments and challenges.
Nothing is ever simple.
Most banks deployed Generative AI in the wild, which is awful. Jason Mikula recently gave an example of a banking app that deployed a chat service that did not return a correct response to a single query. Running in production. In 2024.
Most humans are so bad at prompting that they're getting less than 10% of the value from AI. Tomorrow's fractional CCO or CMO needs to be really really skilled at AI.
Most large banks have spent a massive amount on process automation, and there's little short-term value from GenAI (if they can get over their own internal inertia or fears about data leakage in the first place).
We're clearly in the mother of all hype cycles where some (not all) VCs and entrepreneurs are taking any tiny task and building an AI company around it. Most of this will fail.
But that's venture. That's progress.
GenAI requires skill and people willing to give it a shot. There's no lack of that right now.
Getting 10x productivity around a senior job title is difficult to do well. Yet some are starting to do it. The beneficiaries are those small enough to adapt to it. Which, by the way, was true of Stripe and its APIs in 2010.
In the age of the dinosaurs, mammals were insignificant, tiny little things running around under their feet, generally trying to survive. The same is true of the fractional job title supported by AI.
We're still waiting to see how the dust settles with who wins and loses in Generative AI and its broader implications for copyright.
But in the interim, one thing is clear.
Productivity is coming.
8. Are incumbents screwed?
No.
Do you know who the most productive company in Fintech is on a revenue-per-headcount basis? The answer might surprise you.
(Chart by Jevgenijs Kazanins)
Incumbents with great moats or network business models are best placed. But even that will erode over time as the cycle of company creative destruction gets ever faster.
Incumbents rarely die quickly.
It's more a malaise of job cuts and market exits.
In the US, there isn't really a "big 5 banks" anymore. There's JP Morgan and everyone else. JP Morgan spends more on tech than its competitors by a substantial margin and is the only bank on record that fears Fintech.
Which is smart.
Productive, ultra-low costs to operate businesses are a threat.
It seems there's only one dino alive to that. Jamie Dino.
9. Getting 10x more productive
Every founder, CEO, and operator's obsession is figuring out how to do more with less.
We live in an age where that's possible, but the catch is you must make time to play. You must endure the pain of switching costs and build new habits as a person and an organization.
Things you can do right now:
Make time in your week to play with the tools. Go into your diary and allocate a slot.
Whatever your task is, involve AI next time.
Based on where customer pain is, figure out how hiring 10x staff to do the manual work would make life better. Now, productize that.
The prize for doing so is some level of liberation from the crap.
That makes me hopeful.
ST.
(PS. Yes, I'm drowning in admin, and yes, I'm also training various AI-agents to help)
4 Fintech Companies 💸
1. Tennis Finance - The BaaS Partner Marketing Compliance AI
Tennis automates marketing reviews for Fintech programs that use sponsor banks as their partner. The service provides real-time feedback on customer-facing materials, an audit trail, and different models depending on requirements (e.g., UDAAP, Fair Lending, Equal Opportunity Credit Act, etc).
🧠 This is needed, but is it reliable or defensible? The FDIC, OCC, and State Regulator enforcement actions against banks who support Fintech and nonbank programs are exploding. If you're new to financial services, you're entering a compliance minefield, and it's incredibly easy to make a mistake. So companies need something that can help them 0 to 1 on compliance.
🧠 The problem is that GenAI can be super inconsistent regarding seemingly simple things like UDAAP or FDIC disclosure reviews. It's less AI and more intelligence augmentation (IA). You need a human in the loop to check the outputs. Tennis not only recommends that but also offers a fractional CCO who can do it for you. This also feels like it should be a feature in every other compliance SaaS (that's a compliment). They also offer a fractional CCO. Which triggered a thought. Does GenAI make every compliance consultant fractional?
2. Mercoa - Embedded AP Automation
Mercola provides an API for Fintech companies, marketplaces, and vertical SaaS companies to offer AP automation to their customers. Payers can pay by card, BNPL, consolidate invoices, or extended terms. Vendors can be offered virtual cards, working capital, or instant transfers.
🧠 The Russian dolls of embedded finance. AP Automation, BNPL for B2B, and virtual cards are becoming default features for the tech and Fintech sectors to consume, so why not re-sell? We saw this happen with payments. Why shouldn't it happen for all B2B commerce? If you run a marketplace or vertical SaaS, this increases the platform's volume and monetizes that volume. Lovely.
3. Architect Capital - Debt Funding for X-border Fintech
Architect Capital provides funding from $10m+ to finance lending activity at Fintech companies. They specialize in emerging markets and early-stage Fintech. Products range across asset finance and include invoice financing, credit facilities, revenue-based finance, and tranche funding for BNPL.
🧠 Not all capital markets are created equal. Domestic lenders in the US or Europe have a solid supply of well-known asset managers in this specialized lending market. International markets don't always have this infrastructure and have a massive opportunity. With great opportunity comes great risk, and the larger firms may not always be able to price that like a specialist. Expect more winners who play in niches as private credit expands and Fintech moves up the complexity curve.
4. Wealt - Open Finance for Wealth (UK)
Wealt aggregates multiple assets from bank accounts, investments, wealth, and NFTs or alternatives like wine and real estate. It then provides a single dashboard of your portfolio to manage the entire family's accounts.
🧠 Unlike in the US, where this service is common, the UK doesn't have good investment app aggregation. It's common for the under 40s to have multiple pensions, tax-free savings accounts (ISAs), investment apps, and Crypto. I have at least 8. For me, the household, my kids. I'd love a service that aggregates that and helps spot tax advantages and access alternative investments. The team has a background in boutique wealth advisory, so they know how to do this, but I can't tell how real the app is. The company was formed in 2021 and has zero job listings on the website. The market needs this service, but this app, in particular, seems to be a zombie.
Things to know 👀
Wise is hitting its stride. Revenue reached £276.6m ($350m) with more than 7.5m consumers and 400k business customers. Customers are up 30% YoY, and interest income (net of consumer benefits) is up 126% YoY. Core fee revenue is also up 40%.
🧠 There are two types of companies in this interest rate environment: those attracting deposits and those losing them. Fintech companies like Wise are attracting them and benefitting them. Big banks are seeing the opposite.
🧠 Wise is benefitting from a rise in interest rates. Years ago, a cynical banker would say to me, "They'll never be profitable, especially when interest rates get back to normal." Wise pre-tax profits were £180m last year, and they've benfitted much more from rate rises
🧠 Wise is a beneficiary of the freelancer and remote working boom. The amount of people who work across borders is increasing every year. SMBs and business customers have much higher balances and move larger volumes. The Wise take rate is ~0.90%, so the higher the volume, the more they win.
🧠 Zing by HSBC has a long way to go to catch up; Wise has a strong lead and is a fantastic product (I'm a customer). Who's the ideal customer for HSBC Zing? Perhaps, like Zelle in the US, there's a gap in the market for the boomer version of the millennial app?
Citigroup revenue is down 3% to $17.44bn in the quarter. Charges included exiting municipal bonds, Argentina, and the distressed debt sector.
Goldman's profit is up 50% to $2bn on revenues of $11.3bn as it shrinks its investment bank, pulls back from some consumers, and focuses on asset and wealth management.
JP Morgan revenue is up 12% YoY, reporting $50bn of profit in 2023 on strong credit quality and net interest income. Shares are up 27% YoY while the bank index is down 5%
🧠 There is no more big 5; there's a big 1. The noteable difference between JP Morgan and everyone else is the narrative. Jamie Dimon is "sh*t scared of Fintech" and concerned about the economy. Other leaders are talking up their innovation efforts. The biggest bank in the USA is still hungry and gets how much the market has shifted.
🧠 Bankers are great at BS'ing themselves. But the hypothesis of death by 1000 cuts from Fintech is playing out. $17.77bn quarterly revenue for Citi is not nothing but not vintage performance. Eroding at 3% YoY is not a legacy any CEO wants. I'll give Jane Fraser this, though. She's got a very hard job and is aggressive in trying to do it. She has the courage to come out and say what needs to be said. I really hope she succeeds and shows it can be done.
🧠 Goldman's new strategy appears to be working. They now have $150bn in deposits, which changes their funding mix; they've exited unprofitable business lines and focussed where the money is on wealth and asset management. They're slowly becoming Morgan Stanley with a shrinking Investment Bank, a Global Transaction Bank, and a BaaS operation on the side.
Just dropped: FinCEN report finds $212bn of identity-related suspicious activity. Scammers use a test-and-learn process to probe for weak KYC/AML processes. The report says attackers i) Impersonate others, ii) exploit insufficient verification processes, and iii) Use compromised credentials to get access.
🧠 KYC is the most critical step in catching the bad guys. I saw a stat from NASDAQ today that estimated $3.1trn of illicit payments in 2022. The scale of this problem is staggering. That's more than the GDP of the United Kingdom.
🧠 Nobody starts a Fintech company to deal with drug cartels, human trafficking, or organized crime. But those groups exploit smaller companies because they know they're the weak spot in the system.
🧠 KYC is not a checkbox process. (LOUDER FOR THE PEOPLE IN THE BACK). Onboarding customers is the most critical step. If you treat it as a conversion optimization problem, you'll end up with customers that look legit but are actually committing fraud. Let's be honest; much of the Fintech industry treated KYC solely as a conversion issue.
🧠 Fraud is by far the biggest cause for concern through KYC. Inadequate checks at onboarding are leading to losses for consumers, businesses, and Fintech companies. That's more than a cost of doing business; it's a BSA/AML control failure if it could have been caught at the KYC step.
🧠 All risk problems are data science problems. We don't have to add friction to get smarter at onboarding. We can reduce KYC costs by passively screening devices, behavior, and other data before customers start the KYC process. You can screen out more bad actors before collecting documents from a user.
Good Reads 📚
The FICO score is the single most important number in the lives of US consumers. Lenders often use old versions that don't include new data sources or rely on the competitor Vantage Score from the 3 credit bureaus. That's changing as the Vantage score will now be required in mortgages, and many consumers live in products that rely on other data sources, such as open banking.
🧠 A single central data source is useful. The FICO made life simple for lenders, even if they used the wrong version. It gave a pricing mechanism for secondary markets and handy "reason codes" for rejecting credit.
🧠 But its inaccuracy forces competition. More data (like rent and utilities) and the use of proprietary credit models by lenders detached the score from reality. Despite upgrades, the FICO score became a proxy that didn't always represent reality.
🧠 New sources of data help us lend to the excluded. Open banking and payroll APIs to the soup of data BNPL providers and new lenders use and help people with no credit history access credit for the first time.
🧠 But new data sources are only helpful when they can be shared and reported to other lenders. BNPL providers or new lenders often submit data to the credit bureau and FICO, but that data isn't getting to other lenders or impacting scores. Either because the bureau isn't using that data well or because the lender is using an older version of the score.
🧠 Isn't it nuts that your chances of getting credit are impacted by what software version a lender uses? More data sources make it harder for lenders in the spider's web of regulation and risk.
🧠 Privacy vs fairness vs accuracy is a constant trade-off. For the best accuracy we need all CRAs and bureaus to be consistently updated with data. Fairness means we can't share too much about a consumer (even if it impacts risk) because we could be discriminatory. Privacy means we should limit what is shared and how. The lending data trilemma.
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 :)
(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.