July 9, 2026 —
Systemic risk originates in the financial system—though it could be broader—with the possibility of adversely impacting the real economy. Agentic AI is the perfect tool for monitoring systemic risk.
Introduction
My sense is that many investors are bullish on the fundamentals of the AI build-out, but remain concerned about high valuations and the fact that so much of the build-out is debt financed. There are also concerns about the impact of the Iran War on inflation and growth expectations. And multi-year conflicts in Ukraine and Israel, as well as China’s interest in Taiwan, all continue to linger.
I recently participated in a Columbia Business School / EY executive education class where we had a rich discussion of systemic risk with a room of risk professionals hailing from a broad cross-section of industries. We covered all of the above issues, as well as several others, all equally concerning.
Systemic risk is
“a risk of disruption to financial services that is (i) caused by an impairment of all or parts of the financial system and (ii) has the potential to have serious negative consequences for the real economy.”
Source: Joint report of IMF, BIS, and FSB (https://www.imf.org/external/np/g20/pdf/100109.pdf)
My own view is systemic risk should also cover events originating outside of the financial system that can have serious negative consequences for the real economy.
Using this expanded definition, the class came up with a long list of possible risks, some imminent, others further off. I won’t recount every topic we discussed, but they included climate risk, credit risk, AI overbuild, geopolitical risk (especially regarding Taiwan), and even the risk of a Kessler cascade in orbit (a satellite collision, intentional or unintentional, resulting in a debris field which triggers other collisions).
These risks can be ranked across three dimensions: severity (they’re all pretty severe conditional on occurrence, though some are more severe than others), likelihood, and timing. Arguably, none of the systemic risks that we discussed in class have yet materialized, but they are all plausible. So the pressing questions are:
- What are these risks?
- How likely are they to happen in the near- to intermediate-term?
- How can we prepare today for their occurrence?
I don’t have great answers to these questions, but I do have the sense that they are worth thinking about. Markets feel benign now, but we know that won’t always be the case.
Monitoring
Agentic AI provides an ideal toolkit for monitoring systemic risks. AI agents can help construct a list of topics worth monitoring, then search the web for news about each topic, and finally synthesize the news into a human-digestible summary form. I’ve been an avid user of Claude Code since the start of this year, and this project seems perfectly suited to take advantage of Claude Code’s ever-increasing capabilities.
Together with Claude Code (Opus 4.8), I put together a list of systemic risk topics to monitor:
Macro-financial risks
- Monetary policy and interest rates
- Credit and leverage
- Commercial real estate
- Banking and funding markets
- Market liquidity and structure
- Equity valuation and concentration
- Geopolitical
- Trade, tariffs, and supply chains
- Sovereign and fiscal
- Currency and international monetary system
- Digital assets and stablecoins
- Real economy
- Tail and contagion risk
Thematic and idiosyncratic
- Physical / climate risk
- Technology / cyber risk
- Competition / obsolescence
- Space / orbital risk (Kessler cascade)
- AI capex / data-center risk
These categories are not exhaustive and are subject to refinement. For example, residential real estate does not have its own category. It is covered by credit and leverage, the real economy, and physical / climate risk (e.g., impact of natural disasters on housing). It’s not obvious that digital assets and stablecoins deserve their own line. Ditto for space and orbital disaster risk.
The great thing about this setup, though, is that Claude Code—with guidance—designed the entire workflow: from the output template, to the risk taxonomy, to the data-fetch strategy. Modifying aspects of the workflow is easy because Claude Code handles all the bookkeeping, e.g., changing the number of risk categories across many different specification files (e.g., CLAUDE.md, a skills directory, document templates, and so on). Finally, Claude Code handles the entire update process, from doing the new data fetch to comparing the new output against the old one to looking for changes in risk severity or likelihood to documenting sources searched for each risk category.
This is an example of Claude literally doing the work of at least one and possibly more human analysts.
Systemic Risk Scorecard
Rather than me summarizing the output of this monitoring process, I think it’s better to let the output speak for itself. The scorecard rates each risk category across three dimensions—severity, likelihood, and timing—and flags what has changed since the prior run. I plan to refresh the scorecard periodically—monthly or quarterly—so it becomes an ongoing lens into how these risks are evolving rather than a one-time snapshot. You can download the PDF file of the most recent systemic risk run at this link.
Keep in mind that agentic AI, while it is a powerful monitoring tool, can and will make mistakes. It can overlook important information, misinterpret source material, reach incorrect conclusions, and understate the severity of future adverse outcomes. We view the systemic risk scorecard as only one input into a complex and subtle decision process.
I would love to hear your thoughts on any and all aspects of the scorecard. Did we forget important risk categories? Are there risks within each category which the process did not identify? Do you disagree with the model’s severity, likelihood, or timing estimates?
A healthy dialogue around the capabilities and limitations of this process should produce a better systemic risk monitoring framework. When markets become less benign, this should prove to be a useful tool.
Working with QuantStreet
QuantStreet Capital is a registered investment advisor. It offers financial planning, wealth management, model portfolios, and portfolio analytics. The firm’s approach is systematic, data-driven, and shaped by years of investing experience. If you are an advisor interested in joining our platform, or you want to learn more about our model portfolios or how to invest with us, please reach out at hello@quantstreetcapital.com or sign up for our email list. Note that registration does not imply any particular level of skill or training.