Anyone who’s worked in or around finance would probably attest that a good chunk of the work is tedious, especially when evaluating the performance of companies over time and their future performance outlook.
The work requires pulling forms and documents such as 10-Ks, 10-Qs, and 8-Ks, presentation decks and spreadsheets of results for publicly traded companies, compiling them, reviewing them for pertinent information, or trying to cross-reference them to see if certain words or product names are mentioned over time. This requires a great deal of attention, patience and effort. Fund managers, financial analysts and private equity firms in particular must do or have someone do all this when deciding whether or not to invest in a company, to divest, to bet against it, or to acquire it.
But it also seems like the perfect work for a fine-tuned or financially savvy AI program. That’s exactly the premise of a new startup, Metal, which emerged from venture capitalist Paul Graham’s esteemed Y Combinator startup accelerator program last year and is now announcing a new product that does all this.
“We’re relaunching the product as an AI application that helps fund analysts, for example, in venture capital and private equity, conduct research, perform diligence on investment opportunities, and help fund managers actually monitor their portfolios,” said Taylor Lowe, Metal’s co-founder and CEO, in an exclusive interview with VentureBeat. Lowe previously worked as a product manager (PM) at Meta and co-founded Ripple, among other gigs.
Analysts can feed whatever data they’d like into Metal’s platform about a company, and it will securely store and analyze it on demand in the form of a knowledgable and savvy chatbot, one whose hallucinations are controlled through retrieval augmented generation (RAG) techniques and citations to the underlying data.
The program is available starting now as a software-as-a-service subscription, billed per seat. Lowe declined to specify how much per seat but invited interested parties to reach out to the company.
Founded by Lowe, Sergio Prada, and James O’Dwyer, Metal was launched with $2.5 million seed funding led by Swift Ventures along with Y Combinator and Chapter One.
This investment is earmarked for expanding Metal’s AI platform, specially tailored for large enterprise customers.
According to Lowe, Metal’s success lies not in experimentation but in deploying real-world AI solutions that enhance enterprise operations.
Regarding the underlying LLMs used in Metal, Lowe stated:
“We don’t build or deploy our own models; we use models that customers have a preference for. If we swap one model out for the other because it does a better job at the same tasks, that’s something we want to do because it’s going to deliver more value to our customers”.
“We’ve seen a number of providers offer that switch between different models on the back end, but by and large, [OpenAI’s] GPT is still probably the most common.”
That’s notable at a time when enterprises have begun experimenting with open source LLMs, which Metal also says it can support at customer request.
Streamlining AI for financial services
Metal’s unique offering handles complex infrastructures like data transformation and storage faster than current tools.
“Our product basically follows analysts along their path or with diligence and research, and it helps them consume and parse a ton of information,” he explained to VentureBeat. “They can just move faster during research or diligence as a result.”
In addition, Metal allows users to store and segment information about the companies they are researching in different verticals — e.g. tech, retail, aerospace and defense, CPG, etc — and answer questions about how the entire sector is doing, or one individual company or product line relative to the others in the space.
“So if I’m a fund manager, and I want to know how our tech portfolio is doing… if all of that data is ingested in a single place, Metal, we can start to enable these really interesting queries that LLMs can also support,” Lowe noted.
He provided the following example of how a financial analyst or fund manager might use Metal:
“Whether it’s throwing a 10-K form into Metal and asking it a bunch of questions — ‘Hey, tell me about how XYZ sales went compared to the previous 10k,’ Or whether it’s extracting specific quotes from a call transcript to provide customer anecdotes, ‘Hey, don’t take it for me. This is what management said.’ A holistic picture is really the output that these teams are after.”
Already, the young startup boasts that it has aided funds and analysts in saving time and understanding a company’s performance, outlook and growth prospects for the future — or lack thereof.
“Over the past year, we’ve witnessed this impact firsthand. Metal has accelerated diligence workflows by an order of magnitude,” Lowe wrote on Metal’s website earlier this month. “Every expert call transcript, SEC filing, or financial statement processed by Metal has saved countless hours of effort for funds. And in today’s hypercompetitive deal environment, this can be an unfair advantage.”
Metal processes and parses extensive data sets — from financial statements to board meeting notes — accelerating research and due diligence processes. This capability is not just about speed; it’s about building trust and compliance in a sector where these qualities are paramount.
Unlocking new workflows
The future of Metal looks promising. It’s not only enhancing existing processes but is set to unlock entirely new workflows in financial services.
Today, Metal is rolling out its services on a fund-by-fund basis, offering a tailored approach to each client. This strategy ensures dedicated resources and success for each partnership.
With a focus on real-world solutions and a commitment to trust and compliance, Metal wants to transform how financial services operate, one fund at a time.
VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.