Transcription
Imagine waking up tomorrow morning and discovering that an AI agent has been managing your investments overnight. While you were sleeping, an AI agent monitored earnings reports, scanned breaking news, harvested tax losses, and adjusted your exposure to market volatility. And you didn't manually place a single trade. That future may be closer than many investors think very soon.
Here we're going to be moving to what I would call 2.0, where firms are actually creating agentic solutions within their own organization or within their own app, where a customer can come into that app, give their life circumstances: I'm Devin. I'm 45. I have three children. Here's my risk tolerance. Here are my goals. Here are the type of strategies that I want to employ. And I want real time tax loss harvesting. I want real time rebalancing. I'd like you to use options overlays, take advantage of volatility, generate income, and then go execute.
When I started reporting the story, I thought agentic trading was simply a more advanced way to ask AI which stocks to buy. Instead, I found an industry moving from AI that gives investors ideas to AI that can take action on their behalf. But how close are we to that future? I was just losing money consistently, and I realized this is a sinking ship. Here's what the future could look like with agentic trading.
Some retail investors are using general purpose AI tools like ChatGPT and Claude to research stocks, analyze markets, and even build custom investing agents. One retail investor we talked to did exactly that. So I use Claude and then more recently, Claude Code and the prompt that I used was one. I generated it, actually, in ChatGPT and then plugged it into Claude, and it was essentially just asking Claude to act as a hedge fund analyst to find undervalued stocks. I gave it kind of a dollar amount and I said, 'hey, I'm looking for undervalued opportunities. Find me some and organize it within this budget.' And so it gave me some. And then I also asked for some option plays and it gave me strike prices and all of that. And so I executed the option contracts that had actually directed me towards. Obi told us the trick he followed made money. That was a small, self-selected experiment. He still reviewed every recommendation and made every trading decision himself. Even after building his own investing agent, he isn't ready to let AI take full control. There always need to be a human level of discernment when picking investments, and I do think that you can train it well. But I will always kind of stand by. AI being a tool as opposed to a replacement.
Ivy is an AI agent built by a startup called Podium Markets. Ivy can access a user's holdings across different brokerage accounts, analyze their portfolios, and tailor its response to the level of risk the user selected. But right now, Ivy can see what's in the user's accounts, but it can't act on that information. The user still has to approve the strategy and place the trade themselves. Ivy is not algorithmic, so it's basically AI informs, but the human decides. When you come in, the first questions we asked you after you entered the site is about risk calibration, your risk profile. That's a really big differentiator. So that's one end of the spectrum. The investor gives context that the AI research is, and then the investor makes the trade. The next frontier is bringing those capabilities directly to the brokerage accounts itself.
Public is an online brokerage where retail investors can trade stocks, bonds, options and other assets. Building AI agents in-house, which it hopes will give average investors access to more sophisticated strategies. It could also reshape the economics of the brokerage industry, because AI agents can trade far more frequently than the average customer. Retail investors have used AI now for a few years for research. But really what this era of agentic is doing, where it's now becoming automated and where AI agents can actually execute investment strategies on your behalf. If today's AI helps investors research, tomorrow's AI may manage much more than a portfolio. It could become a financial operating system that constantly optimizes an investor's entire financial life. Effectively everybody has their own family office that is working 24/7 for them while they're awake or sleeping. It's not just looking at your investing sleeve, it's looking at are your assets in the right place? You have to pay college tuitions and pay for your mortgage. We think that agents are going to be employed to do many things, both financial services related, but also across your broader life. Retail will have their own agents if they choose so, but also there'll be corporate agents and institutional agents, and they'll all be interacting in the same agentic economy. This isn't ten years away. This is coming in the next few years in our opinion.
There are optimists in the space that believe AI investing will continue to evolve from just being a research tool to eventually something much closer to a financial advisor. Phase one was science project phase: connect your brokerage account to a Claude, Open AI. Tinker with it. Have some ideas. See what happens. Low risk. Then the second is a little bit that phase of AI agents living in your portfolio and getting to this point where you can actually use it with your real account and real money. And then the third from there, I think, goes into the sense of like, okay, now how far do we get from just execution into also advice? And that's when it starts to obviously go very directly into the world of financial advisors.
This all begs the question of what could go wrong? Can you trust an AI agent to make decisions with your money? I used to work for a family office based in Switzerland and in Singapore, and I got to see a lot of how trading happens from different teams. So I understood the kinds of research that they're doing. I built a agent development and orchestration platform, essentially letting people build and deploy agents. He later used that experience to build a small, automated trading experiment with about $200. I was just losing money consistently. There was a period where I made money for almost a week, but that then went away. And I started considering what else can I do? And I realized this is a sinking ship. I think it's very good for research, and you can use it to help you do a wider amount of research, deeper amount of research faster. I think if you're using it for an automated system or relying on an agent to do it for you, you probably want a professional, you probably want an environment optimized for it.
Another challenge is how do you turn a human goal into a set of precise investing instructions? For example, an investor might tell an AI agent to grow my portfolio aggressively. But what does aggressively actually mean? Should it take on more volatility? Concentrated portfolio? Use options? That's why many firms are moving cautiously. Public, for example, says users review and approve an agent's workflow before it can carry out any tasks. So rather than giving AI unlimited authority, companies are building guardrails designed to keep humans involved. On Public, the AI agents are essentially designed to help you execute your ideas. You still have the last word, so to say. You approve that workflow. You see what the workflow would be, and you approve it and basically set it up. The AI agent will not have an own mind and suddenly do something you never told it to do. I think the average investor still should be be very much in charge of the final decision. And so at this stage, it's really about trying to get to the best holistic and informed answer to your portfolio in a meaningful way that you can really take your own decision and ultimately do the trade. Essentially, you have to make sure that the customer's best interests are at the forefront. And so if the agent is not behaving as modeled or as you expect, that becomes a risk for the firm because a firm is responsible in delivering that regulatory wrapper for the customer.