Risk scenarios at the speed of news.
I spent years as a market risk manager at a tier-1 bank. I know the pain. Scenarios are the most important tool in risk management — and at most banks, they're stale, outdated, and disconnected from reality.
The Problem
A Head of Risk spent a week stress testing the 2024 election. By the time the board saw the results, the world had already moved on.
But that wasn't even the scary part.
His team ran ten stress scenarios every week. Most were broken. Some referenced risk factors that no longer existed. One was still modeling LIBOR exposure — LIBOR was discontinued in 2023.
The rest were historical replays: “What if 2008 happens again?” “What if we see another March 2020?”
The problem with historical scenarios isn't that they're wrong. It's that the world has changed. Regulations changed. Market structure changed. Correlations changed. The leverage that blew up banks in 2008 doesn't sit on balance sheets anymore. A 2008 replay today wouldn't hit the same places.
And the hypothetical scenarios weren't better. Shocks applied to arbitrary risk factors, with arbitrary magnitudes. No one could explain why an oil shock would propagate the way the model said it would — because the model had no mechanism. It assumed a static correlation matrix.
A correlation matrix tells you oil and airlines usually move in opposite directions. It doesn't tell you why. And when you don't know why, you can't distinguish between a supply shock crushing airline margins and a demand boom lifting both oil and travel.
Correlation tells you what moved together. Causality tells you why — and when that relationship breaks.
This is the problem I built StressGen to solve. Most scenario engines are built on correlation. StressGen is built on causality.
StressGen starts from what's happening now — news, policy signals, macro data, market regime — and builds scenarios grounded in real-world transmission. Not: “Based on history, how do airlines react when oil prices double?” But: “If the Strait of Hormuz is blocked, how does that shock propagate to my airline exposures?”
That shock propagates through multiple channels. Each channel operates on its own timeline — hours to weeks. Each has a mechanism. Each can be challenged.
Given what's happening today, what breaks tomorrow?
What This Means in Practice
Causal reasoning is not a slogan here — it changes what a stress scenario is made of. Four things follow from it.
Scenarios start from today, not from a template
Every stress scenario is built on live context — macro releases, market data, news, prediction markets and SEC filings — so it reflects the risks in front of you rather than last year’s assumptions.
Shocks propagate through mechanisms
A causal transmission map decides how a shock travels, and C-Vine copulas put statistically coherent numbers on the secondary moves. Nothing falls out of a static correlation matrix.
Every number has provenance
Each figure in the narrative is traceable to the evidence that produced it — the source, the reading and the date. Numbers that cannot be traced are marked as unverified rather than quietly presented as fact.
Built to be challenged, not trusted blindly
Three AI economists argue the primary shocks before anything ships, and the output is written for review. Your risk team adjusts and signs off — the pipeline is augmentation, not automation.
The result is stress testing your risk team can argue with. Every shock in every scenario carries a rationale, a source and a magnitude you can challenge — which is the only kind of scenario analysis worth taking to a board. See how AI stress testing works.
See What Your Scenarios Are Missing
Schedule a technical demo with our team. We'll walk through scenario generation, the causal engine, and how 10 AI agents work together to stress test your portfolio.
Questions? Get in touch