Hey, fintech fam 💜

I genuinely cannot believe we're already almost in fall. This season snuck up on me this year, but I'm not mad about it.

Maybe it’s the shift in pace, maybe it’s the energy of a busy fall calendar, but I’m coming into these next few months feeling especially energized about what’s happening in fintech and what we're building with FTW: San Francisco.

And today’s stories are a good example of why.

Last week in Chicago, I hosted a conversation at Brex Mode about what it actually means to run an AI-native finance organization. And one question kept coming up: What should AI be trusted to do, and where do humans still need to make the call?

That same tension shows up across all three stories today.

Brex is using AI agents to investigate company spending without a finance team touching every transaction. Flanks is tackling the messy financial data AI needs before it can be trusted in wealth management. And Kita is using agents to turn fragmented borrower information into something lenders can actually use.

Different parts of financial services. Same shift.

AI is increasingly doing the processing. Humans are keeping the judgment.

And I think we’re starting to get a much clearer picture of what AI-native financial services will actually look like because of it.

Let’s get into it. ✨

#SPONSORED

Researching accounts. Building lists. Writing sequences.

There's a better use of your team's time.

Apollo is the AI revenue engine that handles the busywork, so you can stay focused on selling.

Plus, everything you need is in one place:

Stop doing busywork and start building pipeline, faster.

With Apollo — the AI revenue engine powering 4M+ users.

Sign up for free

#TRENDING

Every Thursday, I break down the fintech stories that matter most — grounded in my reporting, interviews with industry leaders, and what I’m seeing unfold across the industry.

AI removes enough labor from the workflow to free humans to focus on judgment.

That was a clear takeaway from a fireside chat I hosted last week at Brex Mode in Chicago, focused on the evolving role of the Chief Financial Officer in the AI era. 

Brex’s CFO, Erica Dorfman, pulled up a $28.75 DoorDash charge from an employee named Elena. On its face, nothing about it looked wrong. 

But Brex's AI audit agent had already flagged it, because the charge landed on a day the company had an in-office lunch scheduled, and the policy says expensed meals don't apply on those days. 

No human reviewer would have caught that without cross-referencing a calendar by hand. 

The agent did it automatically, messaged Elena through Brex's assistant, got her explanation, recategorized the expense, and closed the case. Nobody on the finance team touched it.

That's the version of "AI-native" Dorfman and Legora’s CFO, David Eckstein, were actually describing on stage.

The interesting question isn't whether AI replaces the finance team. It's which decisions still deserve the finance team's time.

Straightforward tasks, such as cross-referencing a company calendar with clearly defined rules against a DoorDash charge, can be handled by agents, but add one complex step and you need human judgment. 

Brex's audit agent now reviews 100% of company spend, up from the 15% most finance teams manage to sample manually. 

It can also read receipt line items, not just totals, which is how it caught an employee expensing liquor as "office supplies" and asked for context before resolving it against policy. 

But it still escalates the harder calls; a $1,300 dinner at Carbone required the employee to submit attendee lists, and a memo before the case even reached a human reviewer, at which point most of the ambiguity was already gone.

Legora is drawing the same line in legal work. 

Eckstein said the company's AI now handles roughly 98% of the work on an NDA or a share purchase agreement, and closed one acquisition in 13 days from signature to close by routing nearly all of the diligence through its own platform. 

But he was direct about where he still pays for outside counsel: 

"They provide a judgment layer that AI will never produce," he said, framing it as reserving legal spend for that moment “where I need that judgment, where it's the difference between a $5 billion acquisition and a $2 billion acquisition."

🤑 BTW – Legora’s Chief Marketing Officer Zeynep Inanoglu Ozdemir will be unpacking more of this at FTW: San Francisco!

Using AI to compress an acquisition process into 13 days is one of the clearest operational ROI examples I’ve seen yet.

Get your tickets now and learn how Legora is using AI every single day. VIP tickets grant you backstage access to meet the speakers.

The harder problem neither company has fully solved is measurement. 

Eckstein described moving from tracking headcount savings to tracking "cost to serve," blending human and token costs into one number, because gross margins that used to sit at 80–90% are compressing as consumption-based AI pricing scales with usage. 

The ROI math for agentic tools doesn't behave like the ROI math for software licenses, and most finance teams don't have a model for it yet. 

Eckstein said this in closing: "AI is going to do a lot of heavy lifting, but it still comes down to all of you in this room to apply judgment." 

The companies figuring out AI-native operations right now aren't the ones automating the most, but instead being most precise about what not to automate.

It's exactly the kind of specificity FTW: SF is built around. Expect to see operators from AI-native to evolving fintech companies showing you behind-the-scenes mechanics, not panels about AI in the abstract. 

Wealthtech company Flanks announced this week that its wealth data platform is now live inside Perplexity, letting advisors query portfolio holdings, transaction history, and investment positions in natural language and act on them without leaving the platform. 

The partnership covers Flanks' aggregation across more than 700 European financial institutions, a segment of the market that has historically been nearly impossible to standardize.

Advisors typically hold client accounts across three to five banks that don't share formats, systems, or even basic data definitions.

That’s the real story in Flanks’ announcement: the EU standardization problem, not the AI layer sitting on top of this tech. 

"No AI model, however capable, can reason well over data that isn't complete, accurate, and compliant to begin with," said Flanks CEO Joaquim de la Cruz.  

Isn’t that the truth?

Whether you’re in the EU market, the U.S., or elsewhere: bad & fragmented data leads to bad & fragmented AI outputs.

Flanks is regulated as an Account Information Service Provider by the Bank of Spain under European Central Bank supervision, and processes more than 8.2 million portfolios monthly across 33 countries, representing over €43 billion in assets. 

That regulatory footprint is part of what Flanks brings to Perplexity's connector ecosystem: standardized financial data with an existing compliance layer underneath it.

According to Perplexity's Jeff Grimes, who said the bar for inclusion is "the level of data quality our users need to trust AI with real portfolio decisions."

It's the same distinction Laurel Taylor draws at Candidly.

Enterprise clients like Bank of America and Charles Schwab require what she calls the "sacred truths" of deterministic math, audit trails, and a hard wall between the AI conversation layer and the underlying calculations. 

Many fintech founders claim to be AI-native, as Taylor has told me; but few have actually built the compliance harness required to deploy that AI inside a bank. 

The compliance requirement Taylor describes for the AI layer is exactly what Flanks is building for the data layer underneath it.

Drew Glover predicted this exact shift on a recent podcast episode: 

As financial institutions become more protective of proprietary data and reluctant to hand control of it to outside model providers, Glover expects a wave of B2B infrastructure companies to emerge between enterprise data and the models acting on it.

Then banks and wealth managers won’t have to staff up on AI engineers themselves. 

The diligence behind Flanks' regulatory status and Candidly's enterprise harness is an  operating standard, and most companies claiming to be AI-native haven't matched it. 

The model may be the intelligence layer. But in wealth management, the company that makes fragmented financial data usable, permissioned, and auditable may control an equally valuable piece of the AI stack.

Get in the room with people who are building the future of fintech and position yourself as one of the most competitive companies in the market today. 

Carmel Limcaoco and her co-founder Rhea Malhotra became best friends after randomly ending up as roommates in a one-bedroom Boston apartment, and started building Kita together while doing their master's in computer science at Stanford. 

That dorm-room project just raised $4.5 million in seed funding led by BoxGroup, with participation from Y Combinator, Golden Gate Ventures, and Apex Star Capital, the family office of Xiaomi co-founder Lin Bin.

Kita's pitch centers on a specific problem in lending: the data that best shows whether someone can repay a loan usually isn't sitting in a credit bureau file. 

It's scattered across bank statements, payslips, invoices, and other documents that look different for every borrower and every market, which is why credit teams spend hours manually reviewing files and chasing applicants for paperwork, stretching a lending decision from days into months. 

Kita built three AI agents to close that gap: 

The company says work that used to take days to months now runs in under 60 seconds, and it has already processed more than $130 million in loan volume for lenders in the U.S., Southeast Asia, and Latin America.

The name itself carries the thesis: Kita comes from the Tagalog word (shout out to my fellow Filipinas!) meaning both "to see" and "earnings," a detail that traces back to Limcaoco's time working with lenders in Manila, where she kept encountering the same issue – regardless of market.

Thin-file and new-to-credit borrowers aren't necessarily uncreditworthy. They're often illegible to systems built around a narrow, standardized data format.

Fixing that legibility problem, rather than the underlying risk model, is what actually expands who gets access to capital. 

Kita's bet is that AI doesn't need to loosen underwriting standards to expand access to credit. It can make more of the information lenders already need usable in the first place.

Importantly, Kita isn't positioning the agents as the final credit decision-maker. The company says human judgment and oversight remain fundamental to the process.

It's also a live example of AI infrastructure getting built specifically for underserved and emerging markets first, rather than as an afterthought once U.S. and European use cases are solved.

📰 Share this newsletter with a friend and start growing your network.

🔗 Connect with me on LinkedIn for daily insights on female leadership.

🤝 Grow your business through content & community by partnering with me.

📣 Promote yourself to 50,000 subscribers by sponsoring this newsletter.

🎤 Host an epic event by booking me as a speaker, moderator, or emcee.

📚 Increase your expertise by ordering your copy of my book, Fintech Feminists: Increasing Inclusion, Redefining Innovation, and Changing the Future for Women Around the World.

That wraps up today’s edition—thanks for reading! Until next week, keep innovating and challenging the status quo.

See you Tuesday!

Love,

Nicole 💜