Hudson Labs Co-Founder CEO Kris Bennatti and Chief Technology Officer Suhas Pai
Hudson Laboratory
This year, discussions were dominated by the promises and dangers of artificial intelligence.
As the original natural wonder of “chatbots” that instantly generated lists, communication and analytics wore off, users quickly realized that generative AI output must be timely, reliable and relevant. In addition to data quality assurance, there is also a hidden issue of digital transformation – will new tools help or replace work?
Equity research provides an ideal context for exploring these challenges. Last year, the SEC received nearly 800,000 regulatory filings. That means millions of pages, billions of words, and countless amounts of data for analysts to sift through and dissect. To meet growing market demand, Hudson Labs (formerly Bedrock AI), founded in 2019, develops innovative software powered by finance-specific Big Language Models (LLMs) to automate equity research workflows and generate actionable insights. The firm currently serves a client list with over $600 billion in assets under management, including major financial institutions and funds. The Hudson Labs platform enables capital markets investment professionals to harness the power of industry-tailored artificial intelligence. Their success also highlights three key criteria for AI adoption – specialization, reliability, and compelling hiring acceleration. Trust and verify
ChatGPT and other on-the-fly generative AI tools have brought language modeling into everyday life. Their rapid popularity is due to their remarkable ease of use for simple business tasks such as quick report writing, background research, meeting summaries and key wording for transcription. Across all industries, employers are finding themselves in the midst of a new generation of artificial intelligence. The 2024 McLean HR Trends report found that while 79% of executives surveyed are looking to increase productivity and efficiency by implementing next-generation AI, only 27% of employees see a clear plan for AI adoption, use, and limitations. Furthermore, implementation is difficult from a technical point of view. Because gen AI is neither a database nor a search engine. Popular "general" AI models that are "trained" on web data will have great difficulty classifying high-tech data for specific industries. Common limitations are "hallucinations" (false information that sounds plausible), inference errors, and poor output repeatability. Suhas Pai, chief technology officer and co-founder of Hudson Labs, highlighted the importance of using artificial intelligence in financial tasks. "A doctor of legal sciences is not a panacea. Financial text is very different from what you typically find online, characterized by financial and legal jargon mixed with numbers and a unique linguistic style. Our models are trained on billions of words of financial text, giving them an understanding of financial concepts, text style and structure, and helping them distinguish between standard and material information. "
Pai explains what makes Hudson Labs' approach unique. "Trust and credibility are essential to the success of AI products in finance. There are too many problems with current LL.M.s, including poor reasoning, tendency to deviate from facts, lack of verifiability, etc. Instead of using one complete LLM, we divide a task (eg generating company background notes) into dozens of subtasks. Each subtask is solved independently, including using a special LL.M. In this way, we can develop and deliver highly reliable products that overcome the common limitations that exist in LL.M. "
This solution is essential to achieve difference efficiency, compliance and reliability. Hudson Labs AI accelerates and improves inventory research
Hudson Labs echo test
No research analyst can afford to write a report that is factually incorrect.
So Hudson Labs paired its technology with popular next-generation artificial intelligence tools and finance-specific bots to run a series of queries on randomly selected public companies like Domino's Pizza. The experiment relies on well-known but less concentrated market registrants, as giants like Apple appear more widely in online data.
First, the test asked each platform whether seasonality (normal disclosure) affects the business revenue of the sampled companies. The "Open Forum" bot generated "facts" about seasonality, such as for businesses. For dominoes, for example, the algorithm reports: “School year affects domino sales. Families with children who have less time to cook during the school year can order meals more often. In fact, Domino's describes its business as "non-seasonal" in SEC filings.
Even more specialized, finance-oriented generative bots have run into trouble. When asked to list Domino's listable business segments, the answer was "delivery, takeout and sit down." The correct answer, as revealed by Domino's, is "US stores, international stores and supply chain." Hudson Lab's AI tool delivered perfect results on all test queries, in stark contrast to the mixed or failed responses of the alternatives.
This advantage is essential to enhance the common equity research mission. Chris Bennatti, CEO and co-founder of Hudson Labs, emphasized: "When analysts worry about their work, I remind them that they need to consume a lot of information to form a view that differs from the rest of the market."
"If artificial intelligence makes the process of using information 50% or even 15% easier, their work will stay the same and reduce friction and frustration. For example, one of Hudson Lab's investments in financial artificial intelligence research is a proprietary noise reduction technology that can be applied to company information, call records etc. In a future based on artificial intelligence, you don't have to read ten pieces of nonsense to find the important thing,” she added.
This perspective shows how well-managed artificial intelligence can accelerate and improve work, reduce fears of job replacement, and support meaningful and lasting digital transformation. Outlook
Benatti sees great promise in automating financial services workflows. "Going forward, our AI research and technology in finance will enable Hudson Labs to deliver three key products in a timely manner: earnings record summaries, auditable automated investment notes and AI-generated news feeds for underserved markets." These resources can differentiate research and promote top talent. At the end of the day, it's the insights, not the data. Who is ready - or not yet?
