- Generative Finance
- Posts
- Apple, CFPB + Excel Spreadsheets
Apple, CFPB + Excel Spreadsheets
The latest in generative AI x finance
Pranjal Daga
October 29, 2024
What’s up, everyone – Pranjal here.
There was a LOT that happened this week.
Before we get into it - who else is at Money2020? It seems like everyone in fintech (me included) is in Vegas right now for the Coachella of US fintech. While I’m looking forward to a few days of events, happy hours, and dinners - I’d love to meet new faces.
Message me if you’re around!
My favorite finds of the week.
NEWS
When Big Tech meets Big Banking…
Apple and Goldman Sachs just got an $89 million reminder that innovation without infrastructure is a recipe for disaster. The CFPB's latest enforcement action reveals the messy underbelly of what happens when Silicon Valley's "move fast" mentality collides with banking's regulatory realities.
The backstory: In 2019, Apple and Goldman Sachs launched the Apple Card with much fanfare, promising to reinvent the credit card experience. The card was meant to be the perfect marriage of Apple's user-friendly design and Goldman's financial expertise. But beneath the (very sleek titanium) surface, things weren't quite so polished.
Now... The CFPB has hit both companies with $89 million in fines for mishandling customer disputes and complaints. The specifics are telling:
Apple failed to forward tens of thousands of consumer complaints to Goldman
Goldman bungled investigations when they did receive complaints
The bank mismanaged credit reporting inaccuracies
Both companies fumbled communication about interest-free financing terms
Apple and Goldman built what appeared to be a seamless front-end experience but neglected the unglamorous back-end infrastructure needed to handle disputes and customer service at scale.
THE TAKEAWAY
Many people have written about this news, but here’s what I think a lot of them are missing: This isn't just another story about big companies dropping the ball on compliance.
The real insight lies in the type of complaints that were mishandled. These weren't exotic edge cases—they were basic dispute resolution issues that any traditional card issuer handles thousands of times per day. This reveals a fascinating blind spot: both Apple and Goldman were so focused on reinventing the "exciting" parts of credit cards (sleek design, digital integration, instant approvals) that they underinvested in the "boring" operational infrastructure that makes credit cards actually work.
But here's the non-obvious part: This failure exposes a fundamental misalignment in how tech and banking companies view "minimum viable products." In tech, an MVP can ship with bugs because you can patch software later. In banking, an MVP needs every compliance and operational component fully built because you can't "patch" a customer's credit score or financial well-being.
Even more telling is what this says about the future of tech-finance partnerships: The next wave of successful fintech products won't be built by companies that excel at either technology or finance—they'll be built by companies that excel at bridging the cultural and operational gap between the two.
This isn't just about having both tech and banking expertise under one roof. It's about having leadership that understands how to merge Silicon Valley's innovation velocity with banking's operational rigor. Without this bridge, we'll keep seeing more $89 million lessons about the cost of treating banking like a software problem.
MY TAKE
The quiet war between AI and Excel
As someone building AI tools for finance, I spend my days talking to people who want to "move off Excel." The conversation usually starts with their frustrations about manual processes and ends with plans to modernize their workflows. But between the enthusiastic head nods about AI, there's a fascinating pattern emerging: the persistence of parallel systems.
It's creating a ubiquitous but invisible problem that not many people are talking about.
Banks are investing heavily in AI systems, but the real decisions are being documented and analyzed in Excel files that never feed back into the AI. The models aren't learning from actual decisions because the actual work happens somewhere else.
The implications are staggering. Think about it:
AI models are being trained on "official" decisions that don't reflect the full analysis
The most valuable analyst insights never make it back to the training data
Banks are getting an incomplete view of their AI adoption
But here's where it gets really interesting: These Excel files aren't just standalone spreadsheets – they're rich repositories of institutional knowledge. They capture nuanced adjustments, industry context, and evolving market insights that could make AI models even more powerful. Each file represents years of accumulated expertise that could be transformative if properly integrated into AI systems.
What's even more fascinating is what this tells us about the gap between how we think AI is being used in finance and how it's actually being used. While case studies tout successful AI adoption, the reality on the ground is far messier.
This isn't about resistance to change (many in finance and banking have adopted new tech with open arms). It's about a fundamental misunderstanding of how financial decisions actually get made. The best analysts don't follow a fixed set of rules – they build intricate mental models based on years of pattern recognition. These patterns are being captured in their Excel files, not in the official AI systems.
There’s something to be said about recognizing these "shadow" systems might actually contain the most valuable training data we're not capturing.
The future of AI in finance might not be about replacing Excel – it might be about learning from it.
Until next time,
Pranjal
How I can help
We can help speed up your compliance and onboarding process.