Analysis

May 17, 2024

Can AI turn us all into Warren Buffet?

If you were hoping a benevolent AI might soon be able to multiply your savings, you might be disappointed — here's why


Tim Smith

5 min read

Photo: Lexie Yu

When ChatGPT hit the market in 2022, people’s imaginations ran wild with the idea that large AI models — trained on more data and information than any human could ever consume — could be turned into money-printing machines.

Initially, we saw viral social media experiments like HustleGPT, where a language model was tasked with running an e-commerce business. Today entrepreneurs are promising that, before long, we’ll see profitable, zero-employee companies where AIs do the work.

Now, founders are also developing AI tools to help us invest our money across the spectrum of financial markets, building everything from crypto trading bots, to tools to automate the work of VC firms and asset managers. 

Advertisement

But if you were hoping that a benevolent AI might soon be able to reliably make your savings multiply while you put your feet up, you might be disappointed. Founders are focusing on building tools for big institutional firms — which have been using algorithmic trading to make money for decades. They warn that AI is currently still far off being able to advise retail investors on where they should place their money.

Watch out, VCs

One company that’s using AI to try and help VC investors do their work is London-based V7 Labs, which recently released a tool that helps VC firms analyse pitch decks, extracting key metrics to surface startups that are worthy of further analysis.

“VCs are inundated with crap decks,” says cofounder and CEO Alberto Rizzoli. “The hard stuff for them is going through all the garbage that they received, and then finding those diamonds in the rough.”

He believes that this technology could allow VCs to stay leaner, and reduce the need to hire the associates who currently filter through pitch decks. One side effect of this, he hopes, could be that it benefits investors with operational experience, as it will take care of number crunching the more financial data, allowing them to pay attention to the more subjective information on how a startup is faring.

“The data that VCs should be looking at is less financial, it’s more ‘unstructured data.’ So it's quotes from customers, it's like the reactions that the customers have when they see the product the first time. The founders communicate to the team during a period of adversity,” says Rizzoli. “These are really important data points that I think an operator picks up and a finance person does not.”

Another startup working in analysing investment prospects is Stockholm-based Grasp, which uses AI to identify M&A opportunities by scouring the internet for promising companies, serving private equity firms who’d otherwise pay consultancy firms for research.

“Clients have compared our output to what they’ve managed to produce with external support before from consultants and typically it's better, in terms of the number of relevant companies found, versus the manual methods,” he told Sifted last year.

AI as the junior analyst

When targeting institutional investors like these, improving the accuracy and reliability of AI models is the number one concern, according to Siddhant Jayakumar, formerly a research engineer at Google DeepMind and now CEO and founder of London-based Finster AI.

“The margin for error is very low. It's real-time decision making, lots of information, lots of money on the line, real people's money, pensions,” he says. “These things are not reliable — they hallucinate. So where is your guarantee that the number you’re giving is correct to three decimal places?”

Jayakumar’s company is using AI to develop tools to automate work done by public market investors, and is working with asset management firms as design partners to develop its product. The system ingests the same kinds of information that professionals at these firms already use, as well as some proprietary data, to automate the type of research that’s normally done by entry level hires in these firms.

Advertisement

“It can help you look at numbers, it can help you cross reference things like: ‘This is what Apple said five years ago, and this is what they said now, this is how that's evolved,’” says Jayakumar. “I think the thing to be clear on is that AI is playing the role of junior analyst, it's not playing the role of portfolio manager making a decision on a trade.”

He adds that he’s spoken to a number of retail investment platforms while building Finster AI, and believes that consumers should be distrustful of anyone promising that AI can direct individuals on how to trade.

“Will there be similar kinds of tools for very savvy retail investors who want to research more? Yes, it will be possible. Does that mean you should listen when an LLM says: ‘Buy this stock’? Absolutely not. I think that is very dangerous territory to go into,” says Jayakumar. “If you, as an AI company, claim that you are a reliable stock picker, then why are you selling your tech and why are you not a hedge fund?”

Don’t quit your day job

The Finster founder adds that retail investors are still likely to find the best returns on investment by “investing in a passive index fund and never touching it,” and that AI hype doesn’t mean we should all be trying to game the markets. 

“Is it net positive for the world for you and me to become day traders? It's probably not,” says Jayakumar.

There’s also the argument that it makes more business sense to build an AI tool for a big, deep-pocketed institutional investment firm, rather than the more fickle retail investor market.

One founder who learnt this is Fayçal Arbai, who launched Bristol-based startup Absurdia in 2022, originally with the goal of using AI to help crypto investors analyse the performance of different crypto trading algorithm tools. 

The company has now pivoted to building a B2B software product for user management, after learning that crypto traders don’t make the most reliable people to sell to.

“We were just struggling to grow and find better paying customers,” he tells Sifted. “We discovered that this is a very price sensitive market.”

For now, it seems like most of the work going into AI for investment applications will target the institutional market — who make better customers than an individual who’s jumped on the crypto bandwagon. And given the low reliability of generative AI systems today, that might actually be better for society than seeing a bunch of chatbots telling us where to invest our savings.

Tim Smith

Tim Smith is news editor at Sifted. He covers deeptech and AI, and produces Startup Europe — The Sifted Podcast . Follow him on X and LinkedIn