AI-Powered Personalized Trading System: The Best Era for Individual Investors (With Investment Process Checklist)

After the previous article Let AI Handle the Grunt Work of US Stock Investing was published, page views quickly surpassed 100,000, and many people messaged me asking for the GitHub link.

People always want a ready-made answer they can directly apply — that’s perfectly understandable.

But what I want to say is: you can copy the dashboard, but using it might feel a bit off. Some data you never look at, while things you check daily are missing. That dashboard was built around my investment approach, and it may not suit you.

That’s why I wanted to write this second article. The first one gave you homework you could copy directly; this one talks about the thinking behind the homework.

1. Your Broker Data, Programmable for the First Time

Let’s start with data — because it’s the foundation of everything.

In the past, your holdings, trades, and P&L data were locked inside the broker’s app. You could view it, but you couldn’t recalculate it your own way, draw a chart, or combine it with other data — you were limited to the official interface.

The privilege of reprocessing that data belonged only to institutions and a few developers.

IBKR’s MCP server tore down that wall. It opened up real-time and historical quotes, account holdings, balances, trade records, open orders, option chains, contract search — all these permissions — to AI.

With these permissions, AI gets a key to freely access your account data. Once data can be freely called, it transforms from “a report you can only look at” to “programmable raw material.” That’s a step from 0 to 1.

Many worry: Will AI place random orders and blow up my account?

IBKR has already thought of that: AI can only generate draft orders — you still need to manually confirm before any order goes through. If you’re more conservative, you can even disable the draft generation permission entirely.

So under this architecture, fund security is essentially a non-issue.

2. Investing = Judgment + Process, AI Should Only Take Over the Latter

Once the data is clear, the next question is: What do we use this data for?

I break investing into two parts: judgment and process.

Judgment is what to buy, at what price, and how to position. This is the core of investing — your knowledge, experience, and risk tolerance combined — and shouldn’t be handed over to AI. A model can tell you that NVDA’s valuation percentile is 30%, but whether to actually buy and how much is your decision, your responsibility.

Process is all the repetitive work before and after judgment: checking recent company news, financials and valuation percentiles; calculating how much your account has earned this year; seeing if the market is overvalued or undervalued; scanning for options with decent annualized returns. This repetitive work is exactly what AI is best suited for.

AI should take over the entire process, to assist your judgment.

How do you decide if something is a process that should be automated? There’s a popular rule of thumb: Any action that you repeat more than twice a week, with a fairly fixed set of steps, should be handled by AI.

Monitoring account asset fluctuations, reviewing returns over a period, checking market valuation percentiles, scanning for options opportunities — all four fit: I do each repeatedly, and each has fixed steps. They shouldn’t keep consuming my time and emotions; they should become something in a sidebar that I can access anytime. Automate these, and save time and mental energy for what truly matters.

3. AI Can Make Mistakes, So “Verification” Is Also Part of the Process

One of the most common questions after the first article: How do you handle AI hallucinations? How do you ensure data accuracy?

This concern is valid. Even today, large models sometimes confidently fabricate numbers, grab wrong contracts, or use stale quotes.

But “AI can make mistakes” and “AI is unusable” are two different things. Just like humans, no one can claim they never make mistakes at work. So the real discussion isn’t whether AI will err, but how to minimize errors, keep them small, and catch them in time.

I use two layers of protection to mitigate this risk.

The first layer is written into the research skill. I hardcoded a few rules: separate facts from judgments, and every judgment must be labeled “My Judgement”; each key data point must come with a source and timestamp; conflicting sources must be listed for cross-checking; data that cannot be obtained must honestly say “not retrieved,” and using memory to fabricate is forbidden. These rules don’t guarantee AI won’t make mistakes, but they force AI to lay out the data’s provenance for you to see, rather than blindly assert a fabricated number.

The second layer is in your own hands. For any critical number that affects an order — price, financial data, positions — cross-check it in your broker app before acting. In short, you let AI help you work, but you’re the ultimate boss.

By compressing the time you spend gathering information and calculating returns from hours to minutes, you should naturally reserve some of that saved time for verification. A reliable process never assumes AI is always correct; it assumes AI will make mistakes and plans how to catch them in advance.

4. Tools Are the Map of Your Process

Back to the original question: Why might this dashboard feel awkward to use?

Because I didn’t start with a dashboard and then find uses for it. Rather, my investment process has these fixed steps, and the repetitive work in each step annoyed me enough to build tools for them. Tools are the map of your process — what your dashboard looks like depends on your process.

I invest by starting with fundamental research on a company, so I needed something that could automatically run through the entire research workflow. Hence the /equity-research skill.

After research, I decide whether to trade. I want each decision to have a basis, and be reviewable later, so I built a trade decision system that forces me to score before placing an order and log it afterward.

I don’t want those logs to live only in Claude’s local storage and vanish with an update. I want them permanently retained, so I sync data to Google Drive.

Besides stocks, I also trade options and indices, so I added an options radar and a market valuation panel.

Every tool corresponds to a specific need.

5. There Is No One Perfect Dashboard

Because tools map your process, and everyone’s process is different, there’s no standard dashboard that fits all investors.

My dashboard is essentially built for investors who focus on fundamentals, value investing, and longer holding periods. It emphasizes research, discipline, and review — because that’s how I make money. But if you’re a different type of investor, your workflow should look different.

If you’re a quant or high-frequency trader, your focus isn’t on writing a data-dense research report — it’s on integrating as many data sources as possible and letting AI quickly calculate signals, approaching the boundary of automated decision-making. You need speed and data breadth, not research depth.

If you’re an event-driven or swing trader, your primary need is a monitoring + alert panel: earnings dates, lockup expirations, key price level breakouts — alerts when they happen.

If you’re a long-term, minimalist DCA investor, you might only need one panel: remind you to execute each periodic investment on time and log it.

To determine what kind of dashboard you should build, use a simple criterion: Look at where you spend your time and where your mistakes come from. Automate the steps where time is repeatedly consumed; constrain the steps where mistakes repeatedly occur with rules. Following these two threads, you’ll find a workflow unique to you.

6. Final Thoughts

A few years ago, for an individual investor to turn their investment process into a system, they either had to write code themselves or pay someone to develop it. Now, you just need to clearly explain your needs. The only barrier left is: Have you thought through your own process? From this perspective, this truly is the best era for individual investors.

At the end of this article, I’ve attached an investment process checklist to help you identify steps you should automate — I recommend saving it. If you’d like to refer to the complete skill and prompts of the trading system I built, follow my account and send me a private message to get them.