By Greg Woolf, AI RegRisk Think Tank
Last year I described a renegade agent that spun up a Delaware LLC and began moving value on-chain without any human sign-off. That once-fanciful scenario has now leapt into old-school equities. Two live retail experiments show autonomous bots not only placing real orders but also outrunning their human-managed benchmarks.
Click Here to Learn More About the AI Readiness
Program for Financial & Wealthtech Firms
The first involves e-commerce CTO Morgan Linton, who wired $1,000 into Robinhood and told Perplexity’s new Comet browser to “make as much money as possible.” After a few teething errors—duplicate tickets, mis-clicked tickers, lost context—the agent deployed the full stake across a FAANG-plus basket and even sprinkled in Berkshire Hathaway and a sliver of crypto. Three weeks later the portfolio is modestly positive while the S&P 500 drifts sideways, proving that an LLM with browser-level authority can already manage real money end-to-end.
The second experiment, run by seventeen-year-old Nathan Smith on Reddit, handed GPT-4o just $100 with one constraint: only U.S. micro-caps under $300 million. Four weeks on, the basket is up about 24 percent versus the Russell 2000’s four. The bot even rode out a seven-percent draw-down, reassessed two-dozen alternatives, and held firm—exactly the discipline most novice traders lose after a bad Friday.
An Unfair Advantage
In a recent episode of AI Tools That Will Replace You (or Make You Rich), investor-operator Greg Isenberg framed his daily AI toolkit as a kind of retail quant stack—cheap, flexible, and able to synthesize what once required a team of analysts and engineers. His refrain—“there’s an unfair advantage today if you know how to use these tools”—applies directly to stock picking: whoever systematizes this stack first, he says, will “make a billion dollars” because AI’s risk-to-reward ratio is so skewed.
Isenberg uses large language models to grab lightning-fast sentiment reads, correlate them with historical context, and balance those signals against multiple perspectives. The fluidity mirrors how institutional desks constantly swap in new factors or alt-data feeds.
Beyond idea generation, a non-coder can now scrape earnings transcripts, run an LLM to score tone or guidance strength, dump the results into a sheet, and push trade alerts—an end-to-end research-to-execution loop that would have cost six figures in 2019.
The Effect on Advisors
For investment advisers, the shift is existential. The edge is no longer memorizing P/E ratios; it is designing and governing these AI pipelines—deciding which models to trust, how to audit their outputs, and how to translate machine-ranked opportunities into client-specific portfolios. Voice tools further shrink friction by letting a trader dictate insights on the fly, auto-transcribe them, and drop them straight into a CRM or client memo.
Advisers who master the stack become curators and risk managers of AI-generated insight; those who ignore it will watch fee compression accelerate as robo-platforms bolt the same research agents onto low-cost execution.
Boulders Toppled by Ants
Isenberg warns against what he calls “vibe revenue”—short-lived spikes from curious trial users. Traders face a parallel trap: flashy back-tests or demo accounts can hide brittle strategies. Sustainable alpha will come from pairing readily available AI engines with proprietary data, robust automation plumbing, and a clear governance framework—roles tailor-made for forward-looking advisory firms.
Former SEC chair Gary Gensler warned Congress that “herding algorithms” could amplify volatility if millions of retail bots chase the same prompt-engineered thesis. The bigger near-term risk, however, is compliance latency. Brokerage agreements were written for flesh-and-blood day-traders, not silicon ones that never sleep and never forget a limit order. Suitability checks, trade surveillance, and best-execution rules have no easy place to land when the “customer” is a checkpointed model.
Will humans still press the buy button five years from now? Probably—though only as a sign-off step in workflows largely driven by agents. The upside is radical efficiency; the downside is that ten million identical prompts could trigger the next flash crash in microseconds. Either way, the bots have already found the trading floor—and, for the moment, they’re up.
Greg Woolf is an accomplished innovator and AI strategist with over 20 years of experience in founding and leading AI and data analytics companies. Recognized for his visionary leadership, he has been honored as AI Global IT-CEO of the Year, received the FIMA FinTech Innovation Award, and was a winner of an FDIC Tech Sprint. Currently, he leads the AI Reg-Risk™ Think Tank, advising financial institutions, FinTech companies, and government regulators on leveraging AI within the financial services industry. https://airegrisk.com
