1. Find the bets you didn’t know you were making
You might own a chipmaker, a utility and a data-center company. Different sectors. But how much of each investment depends on the same AI spending cycle?
Give AI your holdings and weights. Have it map the underlying business exposures, then build a factor analysis using market data.
You can look for shared exposure to rates, momentum, credit or particular customers, rather than stopping at sector labels.
The useful question: how many independent bets am I actually making?
I previously explored this through Robinhood’s factor exposures.
I wrote this $HOOD factor analysis a while ago using old LLMs - it can be even better with Claude Code now!
2. Check whether your diversification holds up on bad days
Two stocks might behave differently most of the time, then fall together during a sell-off.
AI can help you compare your portfolio’s behavior across ordinary markets, sharp drawdowns and different macro environments.
Which positions helped cushion losses? Which ones became another version of the same exposure?
This is more useful than looking at one correlation matrix and assuming the relationships will hold forever. The result gives you a better set of assumptions to stress-test.
3. Look for crowded trades that could unwind together
A stock can fall because other investors need to sell, even if its business hasn’t changed much.
Use AI to compare ownership filings, identify overlapping positions and combine that with liquidity and available short-interest data.
You can also investigate thematic trades: which groups of stocks might be vulnerable to long selling or short covering at the same time?
The goal is to identify potential pressure points. Similar price moves alone don’t prove that investors hold the same trade, and public filings only show part of the picture.
4. Ask what a macro shock would actually do to your holdings
“Oil goes higher” is a market view. What does it mean for the companies you own?
Have AI trace the impact through fuel costs, suppliers, consumer spending, margins and financing. Then rank which exposures matter most to your portfolio.
I did a version of this with Pershing Square’s disclosed holdings and an oil-shock scenario.
(of course, you can model other scenarios too)
The interesting part is often the second-order effect. For instance using oil as the continued example - A company doesn’t need to buy much oil directly to be hurt by customers having less money to spend.
5. Figure out how hard it would be to exit
Being able to buy a position doesn’t mean you can get out of it easily.
AI can build a dashboard comparing your position sizes with trading volumes and realistic selling assumptions. Stress what happens if liquidity gets worse. If you use leverage, check whether you have enough cash and time to manage the exit.
I built a liquidity and crowding dashboard around Situational Awareness’s disclosed portfolio.
This is especially useful for small caps and less-liquid positions. The question is simple: could I still control the decision to sell?
6. Find companies that might need cash at the wrong time
A sell-off becomes much more uncomfortable when a company also needs to refinance debt or raise equity.
Have AI pull together debt maturities, interest costs, cash burn and capital commitments from the filings. Then test weaker sales, lower margins or more expensive financing.
This can help you better understand individual position risks.
Which companies have room to wait? Which might need fresh capital under worse conditions?
You can also check whether reported earnings are translating into cash. This is an AI technique I use a lot especially for my shorting framework for companies with iffy accounting standards.
7. Compare different ways to hedge
Buying puts is one option. So is trimming the position.
AI can help compare those choices alongside put spreads, collars or an index hedge. Look at the protection, cost, expiry and upside you give away under different scenarios.
I shared an example of using AI to explore options hedges.
If you can describe different scenarios in plain English, AI can help you structure the right instruments to structure hedging.
8. Work backward from a loss you can’t tolerate
Instead of choosing a scenario first, start with your limit.
What would have to happen for this portfolio to suffer a loss I’m unwilling or unable to take?
Have AI build a model that tests combinations of weaker earnings, lower valuations and changing correlations. Then compare how different position sizes or hedges affect the result.
This helps connect the risk analysis to an actual decision. You can see which changes meaningfully reduce the vulnerability, without pretending the model knows exactly what will happen next.
9. Look through the smooth valuations in private equity
This one is for the PE guys.
A private investment’s reported value might barely move while comparable public companies are selling off.
That doesn’t necessarily mean the underlying business is less exposed.
I used AI to take a quant approach to a private-equity portfolio: map holdings to public comparables, examine shared exposures and test downside scenarios.
You can extend that to earnings pressure, leverage and lower exit multiples, then look for overlap with your public investments.
Treat it as an estimated view of economic risk. It isn’t a precise new valuation of the private assets.
Where I’d start
Pick one question about your portfolio that you’ve been putting off because answering it seemed like too much work.
Maybe several holdings depend on the same customer. Maybe you’ve never checked how your hedge behaved in previous sell-offs. Maybe your largest position would take longer to exit than you realize.
Start there.
You still need to check the data, assumptions and calculations. But AI can make it much easier to get from “I wonder if this is a risk” to an analysis you can inspect.
Then you can decide whether to keep the exposure, reduce it or hedge it. At least it becomes a deliberate choice.
-Dave
Personal
I'm in Australia this week, speaking at several events including the Citi Australia Conference and GLG events.
If you're attending any of these, hit reply and give me a shout!
