Automation is always a key theme at Money20/20, and this year was no different. The impact of AI as it relates to money movement remained a focus for B2B startups. At the first Vegas iteration of the conference since LLMs rocked the tech world, we looked for evidence of their impact and to see what we can expect from fintech startups in 2024.
Pre- vs. Post-LLM
Pre-ChatGPT, AI in fintech usually meant first-party machine learning, with custom models based on proprietary data. We saw successful solutions from our own portfolio in KYC (Socure) and fraud detection and prevention (Forter), as well as success for verticalized fintech players focused on data ingestion via OCR.
For the most part, that has not changed in the last year. We’re still seeing fintech solutions leveraging their own machine learning models vs LLMs, with some notable exceptions that we think represent promising opportunities:
- Customer support for financial institutions: LLMs are being used across markets as domain-specific chatbots. Consumer fintech is no exception.
- AI for wealth management: We’re seeing co-pilots both for customers managing their own portfolios and for wealth advisors to help them manage their client communications.
- Next-gen OCR: LLMs in tandem with machine learning enable more context-aware data extraction, supporting better document management and data ingestion.
- Next-gen accounting and tax: Bookkeeping automation is a huge opportunity. LLMs present solutions for tasks like general ledger categorization and automating FP&A analysis.
It makes sense that LLMs have not been as immediately applicable in fintech as they have been in other markets. Automation in fintech has historically been reliant on big data, and the benefits of data network effects met the core needs of the market across fraud and identity use cases. LLMs have the most relevant applications across text-heavy domains and as a result have a more limited set of applicability in fields such as payment processing, for example.
Regulations open up opportunities for new entrants
Automation isn’t the only enabler of fintech innovation. Another topic du jour at the conference alongside automation and AI was, of course, impending regulation related to data access and open banking.
The good news for startups: regulations like section 1033 of the Consumer Financial Protection Act will continue to level the playing field and ensure access to financial data. Exactly how much is dependent on time to compliance and enforcement, but the crux of it is that the move toward more open banking and data accessibility gives the benefit to new fintech startups as incumbents lose their dominant data advantage. It also more directly enables new automation by promising more access to underlying data.
Improvement of data extraction solutions
Lastly, we saw some examples of how increasingly context-aware data extraction solutions can help make more data available to banks, lenders, and fintechs. The ability for machines to identify domain-specific content within documents and translate these into decisioning systems has the potential to be greatly improved with LLMs.
There were hints of this at Money2020, with more fully automated data extraction products showcasing their wares alongside a host of vendors relying on automation combined with human-in-the-loop data validation. While these products are early, it’s possible to envision a future where there is less of a need for human intervention around data extraction and document review. Ultimately, automation in fintech data extraction could provide financial institutions benefits in terms of turnaround time for document processing, reduced overhead expenses, and create more seamless experiences for end-users.
Fintech at Scale
To those of you that joined us at our event, we had a delightful happy hour and talked to some exciting founders and teams in the fintech space. If you’re building at the intersection of AI and fintech—LLMs, machine learning, or vertical-specific automation—please reach out. Either way, we’ll see you next year!
