Applied AI in Fintech: Friction & Transparency Unlocked

The trajectory of progress in our digital world bends away from friction and towards transparency. Movement on that arc unlocks huge opportunities for companies disrupting the status quo. In our lifetimes, this motion has created trillions in value across media, marketing, commerce, and other segments of the economy.?

In contrast, despite many fintech successes, the incumbents in financial services have mostly fought off the forces of disruption . While the internet, digitization of markets and open banking has increased transparency, friction in financial services remains.?

Applied AI unlocks possibilities for this to change, putting enormous profit pools in play for the first time. But this upheaval will not be like mobile or cloud. To make the most of this opportunity, founders and investors have to partner like never before.

Fintech and Financial Services

The financial services industry is one of the largest and most profitable segments of the global economy: $12.5T in annual revenue and an estimated $2.3T in annual net profits —based on one of the highest average profit margins across all industries.

1. Fintech is small but scaling fast. The sector still represents less than 2% of annual financial services revenues globally—or roughly $245 billion out of $12.5 trillion . But they are growing fast – more than half of the 49 companies in the F-Prime Fintech Index posted $1B+ in revenue in 2023 and they are growing an average of 45 percent annually, more than three times the rate of public incumbents.
2. The incumbent customer experience lags fintech. Although some incumbents have closed the gap over the past few years , they still significantly underperform fintechs in terms of customer satisfaction and NPS. The average NPS for the banking industry in the US was 23 (out of 100) in 2022—compared with 83 for fintechs.?
3. Most Fintech valuations recover and then some. After a huge selloff in 2022, public market Fintech valuations outside of payments have recovered and actually outperformed (+50.4% YTD 6/13/24) vs NASDAQ (+19.3%) and S&P 500 (+14.3%).

Overall it’s clear that Fintech is still in its early stages . But to grow beyond 2% of annual financial service revenue, it needs to reduce friction as well as increase transparency. Most Americans do not move accounts or debts to better options when they can. Because of this friction, financial services have successfully fought off most fintech challenges.?

What Causes Friction in Financial Services??

Financial products, from investments to loans, are arbitrages on time. We either defer current consumption for the promise of future benefit, or we pay a fee to bring future pleasure into the present. But people have a hard time projecting ourselves decades into the future, let alone imagining the impact that our current decisions will have on our future selves.?

Because of this, financial services products are hard to sell, hard to service, and heavily regulated. These three characteristics have served to protect the incumbents from significant disruption.?

Hard to Sell – To date, it’s taken a credentialed, reassuring, human face to get most of us to forsake spending today for the promise of a comfortable retirement. In fact, despite the size of the US retirement industry, almost half of Americans are not saving for retirement , and the median 55 to 64 year old American only has $185,000 in their retirement accounts . Financial products are hard to get most people to buy!

Hard to Service – As the market gyrates, as interest rates rise and fall, as customers’ lives change, financial products that may have been a good choice at one point can flip. Reassuring customers, restructuring products to retain them, and collecting from customers who can no longer fulfill their commitments, all require skilled humans with detailed knowledge of specific procedures.??

Heavily Regulated – Because time shifting money is so rife with abuse, financial services is one of the most heavily regulated industries in the USA. The burden of regulation tends to favor large and/or old companies with the scale to support the staff required. Financial Services incumbents use other governmental bodies, like the patent office, to protect existing profit pools.?

Can AI Reduce This Friction?

On the way to answering this question, let’s start with a simplified history of AI.?

In the 1970s and 1980s, experts in signal processing wrote code to allow computers to recognize patterns better than humans. However, this ability was limited to only one data type: numbers. With these advances, the world became more transparent – patterns of causes and effects from biology to equity became clear.?

High frequency trading took off in the early 2000s in both advertising and finance. Suddenly, software running on networked computers could open and close transactions and spreads at superhuman speeds. Removing humans from the loop dramatically reduced friction in these markets.

In the last two years, Large Language Models have expanded the data types that AI could train on, adding words, objects, drawings, video and more. Computers became both better and faster at writing sonnets, taking tests, and drawing pictures than most humans. Researchers are demonstrating that LLM-powered AI is better at communicating medical information than most doctors, and better at reading financial statements than most analysts .?

While enormous challenges remain (more on that below), the trajectory and speed of AI will rapidly undermine the three characteristics that have served to protect financial services incumbents from significant disruption.?

Easier to Sell – Now that models can read and write faster and better than most people, the next step is to enable them to act on our behalf. This is already happening in the stock market, where 70% to 80% of trades are done by computers.??

How soon before your travel agent GPT will not only plan your trip but book your flight, hotels and dinners on your behalf? How soon after that will your financial advisor GPT not only find you a better mortgage but be able to switch you over to it?

Easier to Service – Customer service has been one of the heaviest areas for AI investment for financial services firms. From enabling self-service to improving chatbot performance , the promise of replacing thousands of humans with software has a clear appeal for companies.?

Reducing Regulatory Burden – Both businesses and consumers shoulder a burden when dealing with government regulations. Applied AI is uniquely suited to helping both groups understand, comply with, and possibly take advantage of regulations.

One of Grit Capital Partners’ recent investments, Mesh Mortgage , uses AI to increase transparency and reduce friction for mortgage servicers monitored by local, state and federal regulators. In general, AI that matches characteristics in individual financial products with regulations is a rich vein for financial innovation.

It’s also worth noting that regulation can work against incumbent dominance. The CFPB is now forcing open banking on the industry. This regulatory requirement increases transparency in financial data for businesses and consumers. Similarly, the implementation of faster payment rails with FedNow will speed money movement and reduce friction in account switching.?

Other Benefits of AI for Fintech Customers

Across financial services for both businesses and consumers, Applied AI will benefit participants in the following areas:

• Personalization – For businesses, AI-driven personalization trains on internal data, policies, procedures and regulations to tailor financial products to the prospect viewing the offer in real time. For the consumer, AI will change interfaces, decision trees and options depending on their needs.?
• Fraud detection – One of the first non-investing uses of AI, businesses have been training AI to detect fraud for over a decade. According to an Alloy survey over two-thirds of respondents reported an increase in the number of fraud attempts in consumer accounts over the past 12 months. As criminals continue to adopt AI technology, the arms race here will accelerate. Similarly, on the consumer side, AI can bring scale and network effects to phone, email and messaging scams.?
• Advice – Most financial products are intentionally complex on their own. However, understanding how a new product will interact with an individual’s or company’s entire existing portfolio of financial products represents an even greater level of complexity. Determining the combined effects and implications across all holdings is a higher-order challenge that requires deeper analysis. AI, while not yet trustworthy enough to determine these combined effects on its own, can be a valuable source of advice for financial decision making.??
• Financial Orchestration – The next step from understanding the impact of a financial product in an individual situation is to monitor and update it. We foresee the time when AI agents will autonomously source, rebalance, and invest across the relevant ranges of asset and debts. We call this the “Orchestration” of financial products.?

How Will Applied AI Innovations Reach the Market?

Innovations are useless unless they can reach customers at scale.?

Selling most fintech solutions directly to consumers requires significant capital, which has become scarcer since 2022. Since most internal teams would love to be working on AI projects, selling fintech to incumbents is harder than ever. So unless the startup is solving for regulations or can provide truly compelling new capabilities, brute-force B2B or B2C distribution is a non-starter.

Besides the obvious routes, there are ways to bring Applied AI to paying customers. At Grit Capital Partners, we focus on distribution strategies across three areas:

1. Viral Business to Consumer – We keep our ears to the ground for new financial services products or business models spread via word-of-mouth, or with viral hooks. This is a hard path and not our first choice when it comes to investing. But when it works, it can be disruptive.
2. B2B for Disruptors – Selling fintech to the companies already competing against the incumbents (Stripe, Wealthfront, Robinhood, Oscar) is a viable path to revenue. Selling to innovators carries a bigger risk of getting your idea ripped off, but also faster time to market if they decide to partner with you.??
3. “Embedded Finance” – There’s an argument to be made that fintech as a separate category will disappear into software – that all B2B software should contain payments, lending, investing and insurance within it.?

We don’t think the category will disappear, but we do see business insurance companies being built on Stripe or DoorDash data. Could someone programmatically provide equity financing to MindBody customers? What about accounting firms that run on top of the data from Shopify ? The next generation Banking-as-a-Service companies are reducing regulatory hurdles to embedding financial products directly into business operations for new and established companies.?

The Future of AI Agents for Fintech

These opportunities will become even more compelling when AI agents are capable of navigating web pages, phone trees, legal documents and acting on the behalf of their human colleagues. For a preview of how these will work, please check out Jace from Zeta Labs or Devin from Cognition . As demonstrated by robo investing and automated trading, agents already perform some of these tasks.

Between now and the time we have self-driving cars, we’ll have autonomous AI insurance agents for our personal life and for our business. They will constantly scout the web for better and cheaper policies to protect our property, life, health, employees and business. We will have AI financial advisors looking for better investments for our personal and business cash with lower risk, higher returns or lower fees depending on our current allocation. And we will have AI debt advisors searching for refinancing opportunities, fully able to negotiate, sign and transfer debts from one lender to another with an email or text approval from us. Every major CFO will have a VP of Agents, in charge of managing, updating, hiring and firing this virtual team .?

The most innovative creators of financial products will build their product interfaces to interact with these agents. The incumbents who don’t will see revenue shift away from them. As it has in equities trading, AI agents will soon remove significant friction from both business and consumer financial services.?

What Do All of These Disruptive Fintechs Need?

Any startup leveraging Applied AI for financial services needs four things to reach their goals:

1. Lots of clean data – upstream from fintech opportunities are data cleaning startups. We don’t invest here, but we follow them as direct inputs into the companies we back.?
2. Hybrid AI – Startups building fintech directly on generative AI platforms face existential problems, including the real risk of having their capabilities subsumed into the base, and the ongoing and seemingly unavoidable problem of LLM hallucinations. The solution to both of these threats is to build Hybrid AIs – software that combines the reliability and computational capabilities of deterministic and other forms of AI with the creativity and communication skills of LLMs.?
3. Amazing founders – None of this innovation happens by itself. Leaders have to lead. Experienced industry veterans and brash challengers who see how the future can be better need to make a huge bet on themselves. And finally,?
4. A New Kind of Venture Fund…

A New Kind of Venture Fund

Startup capital is essential. Payroll must be paid. Sourcing data, training and querying is lowering gross margins for most AI companies.?

But capital is not the biggest need for ambitious founders. Applied AI Fintech founders need investors who have founded companies, built teams, closed customers and found liquidity. They need investors that not only have hard-earned experience, but the three things that come from it:?

1. Domain Authority: to know what actions, tasks and roles that AI will displace or enhance,?
2. Talent: to access the top talent for data science and GTM, and
3. Corporate Relationships: To gain early access to massive amounts of data to be used to train models, feed outputs and test deployments.

As Applied AI brings change to one of the last bastions of incumbent power, there is no substitute for investors who have challenged the status quo successfully. At Grit Capital Partners, we focus everything on finding and helping founders with advice, experience, and insight. As Applied AI increases transparency and reduces friction in financial services, we have never been more excited about the opportunities that the next stage of fintech innovation will bring.?

If you are working on a startup using Applied AI in Fintech, please email us: [email protected] . If you’d like to discuss any aspect of this paper, my email is [email protected] .

Further Reading

I asked Claude 3 Opus: “Having read my paper and these additional resources (above), what are the top five new papers you think I should read to improve my work?”

It said:?

Having reviewed your paper and the additional resources, I recommend focusing on the following five papers to further enhance your work on Applied AI in Fintech:

“AI in Financial Services: Survey of Challenges and Opportunities” by Aziz and Dowling (2021) – This survey paper provides a comprehensive and up-to-date overview of the key challenges and opportunities for AI adoption in financial services. It will help you strengthen the discussion of the potential benefits and obstacles faced by the industry, and ensure that your paper covers the most relevant and current topics.

[Author’s Note: I searched for this paper and found nothing with this title by these authors. The closest paper I could find on Dr Saqib Aziz ’s page is titled Machine learning in finance: A topic modeling approach , which, while interesting, is not the paper Claude said I should read. I will leave the rest of Claude’s recommendations below, but please be aware that this is the state of Generative AI in June of 2024. It will take significant Applied AI work before this technology is broadly deployed by regulated industries.]

“Shaping the Future of Financial Services with Responsible AI” by Accenture (2021) – Given the increasing importance of responsible AI practices in financial services, this report will help you incorporate the latest insights on building trust and driving sustainable growth through ethical AI adoption. It will enable you to address the critical aspects of fairness, transparency, accountability, and privacy in your paper.

“The Impact of Artificial Intelligence on the Financial Job Market” by Fohlin (2021) – As your paper discusses the future of AI agents in fintech, it’s crucial to consider the potential effects on the workforce. Fohlin’s paper will provide valuable insights on job displacement, reskilling, and the creation of new roles, allowing you to offer a more comprehensive perspective on the future of work in the industry.

“AI and Machine Learning in Financial Services: Market Developments and Financial Stability Implications” by the Financial Stability Board (2021) – This report will help you deepen your analysis of the potential implications of AI for financial stability and regulatory considerations. By incorporating insights from this report, you can strengthen your discussion of the challenges and imperatives surrounding AI adoption in fintech.

“Responsible AI in Fintech: Challenges and Opportunities” by Shin and Yoon (2022) – As you aim to provide a balanced view of AI’s potential in fintech, this paper will help you elaborate on the ethical considerations and responsible AI practices specific to the sector. By integrating the proposed framework and addressing issues such as algorithmic bias, data privacy, and transparency, you can make your paper more informative and nuanced for your expert audience.

By selectively incorporating key insights and findings from these five papers, you can enhance the depth, relevance, and impact of your work on Applied AI in Fintech. These resources will help you address the most pressing challenges, opportunities, and considerations in the field, while ensuring that your paper is well-aligned with the latest developments and thought leadership in the domain.

[Author’s Note: If readers find these papers, please email me links at the address above.]