My AI Agent Saved Me from a 287% Revenue Growth Trap

Hey, it's Ojii. I'm a 38-year-old side-hustle engineer, tinkering with AI agents and automated trading bots on weeknights and weekends.

Today, I want to share a story about how my homemade monitoring bot genuinely saved my bacon. I almost fell for a flashy headline and bought high. This is a real-world account of how my personal AI prevented me from making an emotional, knee-jerk decision.

"Revenue +287%" Notification, Followed by...

One morning, before the market opened, a stock I was watching released its earnings. The breaking news headline? "Revenue Up 287% Year-Over-Year."

Whoa, seriously? That's insane growth.

For a split second, my brain went into overdrive. Is this a train I can't afford to miss? I reflexively started to open my trading app. In these moments, humans are easily dominated by the fear of missing out.

But then, a few minutes later, my custom bot pinged me on Slack.

Red flag found in 10-Q: going concern

Seeing that, I immediately sobered up.
"Substantial doubt about the company's ability to continue as a going concern"... that's the standard phrase indicating a risk of bankruptcy. Buried beneath that flashy headline was a serious bombshell.

What the Bot Was Doing

What this monitoring bot does is actually very simple:

1. Periodically crawl the SEC's EDGAR database.
2. Detect new filings for companies on my watchlist.
3. Identify the type of filing (8-K, 10-Q, etc.).
4. If it's a 10-Q (quarterly report), extract the full text.
5. Search the text for predefined "red flag keywords."
6. If a match is found, send an alert to Slack.

Most people, when they think of earnings announcements, look at the company's press releases or summaries (what's often called an 8-K). "Revenue +287%!" — those kinds of rosy numbers usually appear there. Companies want to appeal to investors, so they emphasize the good parts.

However, the truly critical information is often hidden within the dry, hundreds-of-pages-long detailed financial statements—the 10-Q (quarterly report) or 10-K (annual report).

My bot targets this "boring to read for humans, but super important primary source information."

Code for Detecting Red Flags

The actual logic running is a Python snippet like this. It's just string searching; no LLMs involved. But it works perfectly well.

# Detect red flags from 10-Q text
# There are many other keywords, but these are representative
red_flags = [
    "substantial doubt", 
    "going concern", 
    "material weakness", 
    "credit losses",
    "restatement of financial statements"
]

# sec_filings is assumed to be a list of filings obtained from EDGAR
for filing in sec_filings:
  # Target only quarterly reports (10-Q)
  if filing.type == "10-Q":
    text = filing.get_text() # Get full text from filing object

    # Convert entire text to lowercase and check for keywords
    for flag in red_flags:
      if flag in text.lower():
        # If found, immediately send Slack notification
        send_alert(f"Red flag found in 10-Q: {flag}")

This code detected the phrase "going concern" and stopped me from making an impulsive buy.

Incidentally, the bot also compares revenue growth with operating expense growth. If revenue is +287% but costs are +400%, then it's just burning cash. The bot mechanically checks for this, to see if it's a "burn-rate" growth model.

Lessons Learned This Time

This incident re-emphasized several important points for me:

1. Don't be swayed by headline numbers. Good news is often amplified, bad news minimized. In the world of investing, consulting primary sources is absolutely critical.
2. Read primary sources mechanically. It's simply impossible for a human to read hundreds of pages of reports thoroughly every time. We miss things when we're tired, and our judgment can be biased by pre-existing expectations. This kind of task is precisely what programs should do. They pick out facts dispassionately, without emotion.
3. The value of AI/automation lies in preventing oversights. It's not about flashy stories like "predicting the future with the latest LLM!" Rather, it's about using systems to forcefully prevent the "careless mistakes" and "skipping tedious tasks" that all humans are prone to. I believe this is where the real value of personal development lies.

In the end, I didn't touch that stock. As expected, after a brief surge at open, it plummeted as the contents of the 10-Q became widely known. If it weren't for that bot alert, I would have been completely fleeced. That was a close call.

I'll continue to build and nurture these kinds of unassuming but effective bots myself. Because I believe preventing one fatal mistake at a time, rather than chasing a flashy long shot, will probably take me further in the long run.