Discover more from Fintech Inside
Banking on GenAI - Part 2 | Fintech Inside - Edition #77 - 21st Aug, 2023
Hope to see you at the Falcon presents Fintech Happy Hour Mumbai on 5th Sept! This edition also has suggestions for potential applications of GenAI - Part 2 in the series.
Welcome to the 77th edition of Fintech Inside. Fintech Inside is the front page of Fintech in emerging markets.
We’re hosting our next Fintech Happy Hour in Mumbai on 5th Sept, 2023. If you’re attending Global Fintech Fest in Mumbai, we’d love to see you at the happy hour designed to help you network with the best minds in the financial industry. Details in the newsletter.
This edition is the second part in the series on Generative AI. In this part, I’ve summarised a few potential and practical applications of this revolutionary technology in the financial services industry. I’d love to hear your views on these applications and if you believe there are other applications of GenAI in financial services.
Thank you for sticking around here. Enjoy another great week in fintech!
Raising funding for your early stage fintech startup? reach out to me at firstname.lastname@example.org
✨ Falcon presents Mumbai Fintech Happy Hour
Attending Global Fintech Fest in Mumbai? We’ve got the best networking evening for the best minds in the financial ecosystem. Come over to the Falcon presents Mumbai Fintech Happy Hour!
This time at the Mumbai Fintech Happy Hour, we will host a panel comprising of special guests and luminaries from the financial industry. The panel will be real talk, no fluff, addressing opportunities and challenges for the fintech ecosystem in 2023. Stay tuned for more details on the panelists!
- Date: Tuesday, 05th Sept, 2023
- Time: 6pm to 9pm
- Venue: BKC, Mumbai (venue details will be shared with approved guests).
Global Fintech Fest is happening from 5th to 7th Sept and will witness 250+ Exhibitors, 850+ Global Speakers, and 20,000+ Delegates.
🤔 One Big Thought
Banking on GenAI - Part 2: Potential Applications in Financial Services
This is the second part in the series on Generative AI and its applications in financial services:
Part 2: What are the tangible applications of Generative AI in financial services?
In the first part we covered that Generative AI is a field of algorithms that help "generate" new content (audio, image, text etc.). It takes in certain input (prompt) to find the probability of the most likely output based on the training data (labeled, unlabelled or semi-labeled) and generates the response. If you're looking for a first principle's understanding of genAI, I'd urge you to read the previous edition.
Before we get into the potential applications of Generative AI in financial services, it is important to establish three things:
AI, like finance, is more a horizontal than a vertical. More a feature than a standalone product. AI is a technology that can be applied on products and features to achieve significantly better results. The user doesn't care if the product has AI or GenAI or blockchain or coded in plain ol' cobol (some insurance companies are still coded with it). The user solely cares about their end goal and the experience in achieving that goal.
It's easy to pronounce many companies dead because of this new shiny technology in generative AI. But that's far from reality. If a company gets disrupted, it'll be because of their own mistakes, not catching up to new innovations or other market conditions not because of generative AI itself.
This is not the first time AI is being applied to the financial industry. AI has previous been used for algorithmic or quantitative or high-frequency trading, payment frauds identification, credit underwriting, money laundering identification, customer service, identity theft avoidance and so on.
In an Interview in 2016, HDFC Bank's Country Head for Branch Banking & Retail Trade FX Business said this
What are some applications of GenAI already in the wild? (picked from the Emphasis Ventures blog post on GenAI).
BloombergGPT: An LLM by Bloomberg, specifically trained on a wide range of financial data to support a diverse set of natural language processing (NLP) tasks within the financial industry. This model is expected to assist Bloomberg in "improving existing financial NLP tasks, such as sentiment analysis, named entity recognition, news classification, and question answering, among others"
Brex, an expense management platform is partnering with OpenAI to launch a chat-based platform to generate insights and benchmarks for customers.
Stripe is also partnering with OpenAI to leverage its GPT4 model to improve customer support, fraud detection, NLP.
Toggle is an AI-powered investment advisory platform that trains LLMs like ChatGPT on chat-based investment advisory.
Glia offers a seamless customer interaction solution across platforms for financial service companies. It has solutions for customer onboarding, sales and personalized customer support.
Mostly AI develops synthetic data for financial service companies to train their fraud detection models and other internal AI/ML models.
SymphonyAI recently launched a platform ‘Sensa Copilot’ to accelerate financial crime detection and management. This solution operates on the OpenAI model.
What are some potential use cases of GenAI in financial services? We're still in the early days of identifying tangible use cases of GenAI. This is the bleeding edge phase of GenAI, so there's a lot of throwing darts in the dark. When the lights come on, we'll know which use cases hit the target and have actual product market fit. Here's a list of my darts in the dark:
Automated personal finance for the masses: Train models to identify recurring transactions, track and predict cashflows, budget spending habits, analyse market conditions alongside asset performance and come up with an optimized savings and investment plan. All day, everyday, wealth manager, affordable for everyone.
Customer services: Called your bank, only to be kept on hold for hours without your problem being solved? Or visit your bank and still don't know which documents you need to apply for that home loan? Deliver the right service and speed up customer engagement including product enquiry, account opening, ongoing support and grievance management with responses fine tuned to be empathetic and comprehensive.
Comprehensive Underwriting: Underwriting is typically rules based and still largely dependent on credit score. As we continue to generate more personal digital footprints in a single day than in our histories, it helps to use some of this data to draw a clearer picture of a borrower's finances. With GenAI, heaps of data can be analysed for increased approval rates with better success.
Targeted Collections: Debt recovery even today is most effective offline. Though lately, you probably noticed that you're receiving more texts and calls to pay your instalment on time. Even if you've already paid it, you still receive these reminders. That's because financial firms don't have the systems to target their messages. GenAI can be used to customise messaging and calls targeted to borrowers for their specific case (due date upcoming, few days late, delinquent, default etc).
Learning and Development: The larger the firm, the more complex their human capital structure and the tougher it gets to train and grow the firm. From internal communication and training (culture, goals, best case practices, tribal knowledge etc.) to external communications (new product launch and features, product changes, updated regulations and much more), all can be achieved with better results via GenAI.
Fraud detection in payments: As the world moves towards real time payments and settlements, scams and frauds are doing its best to keep up. Moreover, the frauds are becoming increasingly sophisticated. Fraud detection has typically been a post-facto activity, meaning, after payments are processed the portfolio is analysed for fraud. It is significantly tough to do before payment is done or during the payment being made. GenAI can help identify fraud patterns early on and trigger alerts for fraudulent transactions.
Claims processing: Insurance frauds are even more common and this has made insurers ultra-conservative: meaning even a genuine insurance claim can be declined. With GenAI, the claims process can be streamlined to understand the case, request adequate supporting documents, analyse the dump of supporting documents (in whatever format) and make claims decisions - all in the fraction of the time it takes today.
Personalised recommendation engines: We've all had that experience of visiting a sales rep and getting sold an entirely unrelated and potentially terrible financial product. Sales reps are not equipped and apps are not designed to analyse a user and recommend personalised products. GenAI can personalise product recommendations at scale while maintaining the firm's distribution targets.
Never forget about compliance: None of us really enjoys filing documents for regulatory and compliance perspectives. It's a necessary evil (not saying don't comply) that takes the fun out of building. A Compliance GenAI bot could pull data from various sources, audit compliance against ever evolving regulations and laws, and wait for an approval or automatically file required forms to stay compliant. Taken a step further, GenAI could even be built for internal policy and rule compliance - cost compliance, HR compliance and so on.
Better market trader: It's not easy to keep a track of thousands of stock market data, macro data, market data and more. Furthermore, it's not easy to keep coming up with stock trade ideas. GenAI can be put to the creative test of ingesting data from these various sources and in various formats, to churn out multiple ideas. Moreover, this use case can be extended to decide money movement and liquidity management between asset classes to improve IRR's.
Issuing legal notices: Legal notices continue to be the most efficient way to improve bad debt collection rates. In cases of delayed loan repayments and in extreme cases wilful defaults, GenAI can understand the borrower's case file, customise and automate legal notices and file for debt collections at scale.
Customisable frontend: Imagine a website or app with copy and images customised "for me", with my profile in mind and my tastes in mind. GenAI can generate and display unique website copies, each potentially generated for the specific user, to improve click through rates and conversion rates.
This sounds like a lot of people will lose their jobs, right? Personally, I don't think jobs will be lost en masse or in a short span of time. It'll be long before some jobs will be made redundant altogether. Some jobs will be made more efficient and new types of jobs will be created. Moreover, no government will let enterprises fire millions of people in one fell swoop. It's just too important a topic.
Why won't GenAI takeover jobs anytime soon?
The financial services industry (FS) is subject to stringent industry regulations and therefore will take a conservative view to implementing new technologies. There's a lot at stake for financial firms to implement new technologies at scale, most important of those - trust. The room for error is minimal. This is why, financial firms look for precedent or the regulator evaluating technologies for implementation and therefore precedent.
Foundational LLM models are built on top of publicly available internet data, so are very good at tasks ‘out-of-the-box’ that require this generalized information. More specialized outputs require a combination of specialized training data and specialized workflows or product wrappers around the AI model. By specialised training data, I mean proprietary data. Proprietary data will become new "oil". Enterprises will seek to collect, protect and monetise proprietary data like never before.
We forget that AI was already applied in the past, and the adoption of text input interfaces is limiting to the majority of users. Text input is a huge drag on our interaction with interfaces. We then brought in speech based AI to improve our interactions with our products - this was supposed to be the holy grail to exponentially grow user adoption, but adoption didn't scale in reality. Unless our input interaction with AI products and features improves, it's tough for large scale adoption of generative AI.
With GenAI, the one thing that's become the most important, the one thing that will create clear winners, the only thing that will matter is going to be distribution. With GenAI, give a prompt and anyone could copy interfaces and whip up near identical entire products - full with the code, integrated with databases and other services. And so, if product is not a moat, capital is a limited moat and loyalty is limited, distribution will be a strong predictor of success. Distribution is king in a Generative AI future.
🏷️ Other Notable Nuggets
🎵 Song on loop
Fintech updates can get boring, so here's an earworm: Wet Sand by Red Hot Chili Peppers (Youtube / Spotify). All time fav song, love the progression and eventual crescendo and final guitar solo (John Frusciante, you legend!). Such feels.
👋🏾 That's all Folks
If you’ve made it this far - thanks! As always, you can always reach me at email@example.com. I’d genuinely appreciate any and all feedback. If you liked what you read, please consider sharing or subscribing.
See you in the next edition.