Blockchain, Distributed ledger technology, AI
The meteoric rise of AI in the past year has captured the world’s attention. With hundreds of millions of users flocking to tools like ChatGPT and the deluge of other AI-powered applications, investors and startups have quickly shifted their focus, directing significant investments toward AI projects. This surge of interest hasn’t just been limited to Big Tech, it has also sparked curiosity in the world of decentralized finance.
Notable crypto and blockchain focused investors, such as Framework Ventures and Peter Thiel’s Founders Fund, are now championing a new wave of “crypto + AI” projects like Sentient or Space & Time. While much of these new cross-industry startups have centered around using crypto to challenge AI tech incumbents, we’ve yet to see more traditional financial institutions thinking about how the combination of AI and blockchains might play a role in their technology stack, that is until last month.
Recently, the team behind the major oracle protocol, Chainlink, released a report unveiling that they have been working on an initiative that combines AI, oracles, and blockchain technology to address the lack of real-time and standardized data around corporate actions.
A Who’s Who Of Global FMI providers
The companies involved in the project are a who’s who of global financial market infrastructure (FMI) providers including Swift – the world’s largest inter-bank messaging platform and Euroclear – a global clearing and settlement firm, along with investment management companies such as Franklin Templeton and Wellington Management, and major banks, including UBS, CACEIS, Vontobel, and Sygnum Bank.
Oracles are entities that provide services to blockchains, handling tasks that they cannot typically manage on their own, such as piping in live data from the real world or facilitating transactions across and between different blockchains. In the crypto industry, Chainlink is the most widely adopted oracle network that has enabled over $16 trillion in total transaction value with its data feeds and Cross-Chain Interoperability Protocol (dubbed “CCIP”), a bridge between chains.
Financial institutions around the world face complex data fragmentation problems, particularly with data on mergers, dividends, stock splits, and more. Data relevant to multiple parties in a deal often must go through a complex journey through the hands of custodians, brokers, fund managers, exchanges, and investors.
As the data travels through the pipes, it often enters various inconsistent formats and states (think PDFs, source press releases, spreadsheets, etc.), leading to a confusing array of duplicative or differing sources, formats, terminologies, data cleaning problems, and plainly incorrect information. Processing corporate actions specifically is a decade-long problem that many, including giants like DTCC, have tried to solve.
What may sound like a minor issue to an outsider is actually a massive and complex one in the world of post-trade finance: current inefficiencies in corporate action processes with thousands of regional investors, brokers, and custodian businesses facing costs of $3-5 million each annually and 75 percent of firms having to re-validate custodian and exchange data manually.
Chainlink’s initiative introduces a potential solution using decentralized blockchain oracles. The goal is to mitigate the need for manual reviews by creating a “unified golden record” around corporate actions that were ported across blockchains and accessed in real-time by custodians, asset managers, and other relevant parties.
AI For Data Cleansing
Chainlink oracle networks are used in combination with large language models (LLMs) like OpenAI’s ChatGPT 4o, Google’s Gemini 1.5 pro, and Anthropic’s Claude 3.5 sonnet to validate and deliver key financial data onto the blockchain, or onchain. The on-chain corporate actions data is then moved across both private and public chains using Chainlink’s CCIP.
“The combination of AI and oracles is a powerful tool for taking corporate actions data and turning it into highly reliably structured data,” said Chainlink Co-Founder Sergey Nazarov, “Solving this problem creates a lot of advantages for asset managers, banks, and financial market infrastructures to all be in sync many times faster than today, at a fraction of the cost, and with a massive reduction in costly errors that affect the financial system.”
So why is this initiative important?
According to Laurence Moroney, AI Researcher and Best-Selling AI Author stated in a written statement to Forbes, “While this project focuses on corporate actions data, Chainlink’s approach to combine AI, oracles, and blockchains can be applied to other types of unstructured data in financial services and beyond.”
The possibilities are extendable if this approach is applied to other types of unstructured data. Most of the world’s human-generated information like legal documents, insurance contracts, real estate agreements, surveys, voice recordings, social media posts are not easily readable by machines. The potential impact is significant and can transform how industries handle everything from contracts to customer interactions, especially using AI.
In sectors like financial services, where much of the data remains unstructured, this solution might not yet be a silver bullet as significant hurdles remain. LLMs are still prone to hallucination, meaning they can still generate inaccurate or fabricated information. In fact, a study comparing 11 public LLMs showed hallucination rates ranging from 3 to 27 percent. To minimize the risks of LLMs, a pre-trained model is likely required for true scaling, ensuring the system is well-equipped to handle complex and nuanced datasets.
Even with oracles verifying outputs via a consensus-based model and the use of multiple LLMs, there must be mechanisms in place to flag uncertainty around outputs that trigger a manual review, especially in high-stakes environments like finance, where a lot of investors make their decisions based on this data. Many specialists are also now focusing on Small Language Models (SLMs), data that is curated and better structured, especially for regulated firms.
The New Plumbing
For the powerful combination of AI and blockchain to really gain traction in a way that is both impactful and enduring, institutions must find practical applications for blockchain technology. Often, these applications will come in the form of key and mission critical operational business systems and processes.
As Stéphanie Lheureux, Director of the Digital Assets Competence Center at Euroclear, pointed out, the combination of oracles and AI “can address major pain points and redesign workflows for greater efficiency, transparency and value.”
If nothing else, this initiative demonstrates that some of the most meaningful innovations might be the ones happening quietly behind the scenes, solving unglamorous complex problems in large financial institutions that power the global financial services sector.
Digital financial market infrastructure (dFMI) isn’t the sexiest of topics for consumers, businesses, and policymakers – few get excited about having to change the old plumbing for new – but it a critical digital transformation that will keep our money flowing freely for more people and business, at cheaper costs, with better products, services and accuracy on Web3 for the 21st century, and into the future.
