Risk is where a prop firm’s money actually lives or dies, and it’s the function most often run on the least infrastructure. A spreadsheet, a couple of analysts, and a lot of after-the-fact review works until it doesn’t, and when it stops working, it stops working expensively. Modern risk infrastructure is what lets a firm enforce its rules consistently at any scale.
The Manual Wall
Manual risk review has a hard ceiling. An analyst can hold a few hundred accounts in their head and spot the obvious problems. At a few thousand, the obvious problems hide in the volume; at tens of thousands, review becomes sampling, and sampling means the breach you didn’t look at is the one that pays out. Hiring more analysts pushes the wall back a little but never removes it: the model is linear, and trader volume isn’t.
The Four Layers of Real Risk Infrastructure
A risk system that scales is built in layers, each handling what the one above can’t:
Why Real Time Matters
The defining property of modern risk infrastructure is that it operates in the same timeframe as the trading it watches. Risk reviewed at month-end tells you what you already lost. Risk evaluated as trades happen lets you act before a fraudulent payout is released or a breach compounds. The shift from retrospective to real time is the single biggest difference between a firm that absorbs losses and one that prevents them.
"The mistake firms make is treating risk as a report they read later. Modern risk infrastructure is a system that runs in the moment, enforcing rules as trades land, surfacing patterns as they form. You can’t prevent what you only see at month-end."
Leonard BreitkopfChief Technology Officer · QTG
The Role of AI: Support, Not Replacement
AI belongs in risk infrastructure, but in a specific role: handling the scale and surfacing the signal, not making the final call. Pattern detection across tens of thousands of accounts is genuinely beyond human capacity. That’s where models earn their place. But the decision to enforce, to withhold a payout, to ban an account carries consequences and accountability, and that judgment should stay with a risk professional. The right design uses AI to prioritize and evidence; it leaves the verdict to a human.
Evidence, Not Suspicion
A risk decision a firm can’t defend is a liability, even when it’s correct. Modern infrastructure treats evidence as a first-class output: every flag arrives with the trade history, the rule it violated, and the context that explains it: ready to support a Terms-of-Service enforcement, settle a dispute, or satisfy a regulator. This is what lets a firm act decisively and fairly at the same time: honest traders get cleared with evidence, bad actors get removed with evidence, and the firm can stand behind every call.
Manual risk review doesn’t scale, and the answer isn’t more analysts, it’s layered infrastructure: automatic rule enforcement, live exposure, behavioral detection, and audit-ready evidence, running in real time. Use AI to handle the scale and surface the signal; keep the judgment with your risk team. That’s what modern risk infrastructure looks like.
Leonard is Chief Technology Officer at Quant Technology Group, where he leads platform architecture and the engineering behind QTG’s risk, CRM, and infrastructure products. He writes about what it takes to build trading technology that scales from hundreds to tens of thousands of accounts without breaking.
More from Leonard
Milos Mosovsky
Product · QuantSentry