How AI Agents Are Changing the Way We Work

ByBNW TEAM

I have been trading on and off for six years, badly enough to know my problem is not analysis. It is discipline. I see the setup, I hesitate, I enter late, I move my stop. So when a friend told me he had handed his whole swing strategy to an AI agent and stopped watching charts, I was sceptical and curious in equal measure.

This is a walkthrough of the first week, written for someone who has never touched an API key.

Day 1: connecting an account (15 minutes)

I used https://cubrik.ai, mostly because it listed the broker I already use. Signing up took the usual two minutes. The part I was dreading, connecting the exchange, turned out to be a form: pick the venue from a marketplace, paste the API key and secret the exchange generates for you, save. The key I created had trading permission only. No withdrawals. That mattered to me.

Once connected, the app looked less like a trading terminal and more like a chat tool with a sidebar: New chat, Automations, and a trade history that was empty.

Day 1, continued: writing the instruction

This is the whole product, really. I opened a new automation and wrote, more or less:

“Trade XAUUSD only. Look for a daily close above the previous 20-day high and go long on the next session open. Risk 1% of the account per trade, stop under the breakout candle low, trail the stop to breakeven after 1R, take profit at 3R. Never hold more than one position. Do nothing on Fridays.”

Trigger: every day at 00:05 UTC after the daily candle closes. I hit save and went to bed.

Day 2 to 4: nothing

The agent woke up each night, pulled the candles, concluded there was no breakout and wrote a one-paragraph log entry explaining why. Boring, and exactly what I wanted. In six years I have never managed three days of doing nothing when the rules said to do nothing.

Day 5: the first trade

Gold closed above the 20-day high on Thursday. At the open the agent went long with the position size it had calculated from my balance, placed the stop under the candle low, and logged the reasoning: the levels it used, the size, the risk in https://cubrik.ai. I checked the exchange app; the order was there, exactly as described.

Friday, per my instruction, it did not trade. Over the weekend the position sat open with the stop in place.

Day 7: a mistake, and it was mine

Monday the price pushed to about 1.2R, then pulled back. The agent did not move the stop to breakeven because I had written “after 1R” and its interpretation was that the move had to close beyond 1R, not just touch it. The log said so plainly. The trade eventually stopped out for a small loss.

I rewrote the line to “as soon as price touches 1R” and that was the end of it. The lesson is the same one every programmer learns: the machine did what I said, not what I meant. The difference is that here the fix was one sentence in plain English.

What I would tell a friend

  • Start with a strategy you have actually traded by hand. The AI executes; it does not invent an edge for you.
  • Write your risk rules first, your entries second. Position size and stop logic are what keep you in the game.
  • Read the logs for the first week. Not to catch the agent, but to catch your own ambiguities.
  • Use a trading-only key and a separate sub-account if your broker offers one.

I am still running it. The account is up a little, which is not the point. The point is that for the first time, the strategy on paper and the strategy on the account are the same strategy.

Looking Ahead

AI trading agents are not a shortcut to guaranteed profits, but they can make following a strategy more consistent. The key is to start small, test every rule carefully, and monitor performance over time. Technology can help reduce emotional decisions, but it cannot eliminate market risk. In the end, successful trading still depends on a well-tested strategy, sensible risk management, and realistic expectations.