AI is great at helping you think through path dependencies. If this happens, what changes next? And then what?

But a big mistake I see when I run corporate trainings at hedge funds and PE funds is that people jump straight to tickers.

They have a view on a particular theme, so they ask AI which stocks to buy. Or they give it a few names and ask how those companies benefit.

That makes the answer overfit the starting point. The names become anchors, and AI builds a story around them instead of working through the economics from first principles.

This week, I'll show you a technique for getting more differentiated answers when you have a big, high-level thesis and want to drill down into investment ideas.

The example is the K-shaped economy. Specifically, the less glamorous side: what happens when households have to make their money go further?

Was this email forwarded?
Subscribe to get more AI investing breakdowns just like this!

Start with the path dependencies

You first need to define what 'thesis' you want to test. Before AI finds tickers, you need to both express what you believe AND test if it's true.

The process is pretty simple:

1. Define the scenario. Describe who is affected and what changes in their finances.
2. Map the transmission paths. Ask how people adapt, where spending moves, and what happens to suppliers, competitors and bottlenecks.
3. Pick the paths you believe. Use your own judgment to decide which mechanisms deserve research.
4. Collect data and test the logic. Check consumer behavior, industry data and company filings. Look for evidence that would break the chain.
5. Map the tickers again. With a better understanding of the economics, decide which companies actually express the idea. Replace weak early matches.

To follow along, use the full research prompt below. It sets up the context, asks AI to propose a plan and resolve questions, then pauses for you to select pathways before the detailed research.

​Download the full Claude Code research prompt​

The useful part is how much wider this makes the search. You can explore several branches before deciding which ones are worth your time.

The Results: The other side of the K-shaped economy

Here's some interesting finds of my K-Shaped Economy research.

“Consumers trade down” sounds straightforward. But households have several ways to spend less.

They can switch brands, cook at home, share housing, keep an older car, replace paid services with their own labor, or delay a purchase entirely.

Those responses have different consequences. A cheaper product might gain customers. A retailer might gain bargaining power. A company selling products for new households might lose demand.

Here are the paths I'd take further.
“Consumers trade down” sounds straightforward. But households have several ways to spend less.

They can switch brands, cook at home, share housing, keep an older car, replace paid services with their own labor, or delay a purchase entirely.

Those responses have different consequences. A cheaper product might gain customers. A retailer might gain bargaining power. A company selling products for new households might lose demand.

Here are the paths I'd take further.

Where the food budget moves

Private labels → shelf control. Shoppers keep buying essentials but stop paying the national-brand premium. That shifts shelf space and negotiating power toward retailers with trusted own brands (e.g., WMT).

Restaurant meals → home cooking → grocery baskets. Replacing restaurant meals with groceries moves spending from service and delivery toward ingredients, prepared supermarket food and the retailers selling them. Household cooking time becomes part of the saving.

Expensive beef → cheaper proteins. Households can change the meal itself, switching from expensive beef to chicken, turkey or other lower-cost proteins. The resulting volume shift affects processors, feed demand and production capacity.

Benefit-payment cycles → shopping peaks → inventory and staffing. Shopping around benefit-payment dates concentrates demand for essentials. That creates recurring replenishment and staffing peaks for retailers, even when total monthly spending does not change.

What households stretch, share or replace

Expensive housing → shared homes → fewer starter purchases. People respond to high housing costs by living with family or taking roommates. Fewer separate households mean fewer kitchens to equip, rooms to furnish and utility accounts to open.

Manufactured homes → demand for permitted sites. A lower-cost structure still needs a permitted, serviced place to sit. That redirects attention from home factories to community owners and the supply of usable sites (e.g., UMH).

Keeping older cars → repairs and parts. Postponing a new car purchase extends the working life of the existing fleet. Necessary repairs then create demand for workshops, parts distribution and replacement components (e.g., ORLY).

Paid services → DIY → supplies. Replacing paid help with household labor shifts some spending toward materials, tools and consumables. The same budget pressure can reduce service visits while supporting the supplies needed to do the work.

Telecom bills → cheaper plans. Households can keep connectivity while cutting the monthly bill through cheaper plans. Demand moves toward value brands, changing the customer and revenue mix within networks that may own both tiers.

Follow the wealthy side too

Complex wealth → custody, reporting and governance. Larger family fortunes can mean more entities, accounts and investments to coordinate. That creates recurring work in custody, consolidated reporting and governance, beyond choosing the next investment (e.g., NTRS).

Private aviation → maintenance and training. An aircraft purchase creates ongoing needs for maintenance, parts and trained crews. Following the asset after the initial sale reveals a service economy around keeping it available, reliable and ready to fly.

​Explore the full interactive map

What do you do with this? Use AI to run further diligence

With a ticker shortlist, this is not enough to just allocate to your portfolio. You need to continue pushing on individual subthemes that stand out to you.

Take the older-car path. I'd ask AI to collect evidence on repair volumes, parts demand and customer behavior. Are people repairing their cars, or putting off repairs because they still cannot afford them?

Then compare the parts distributor with the component supplier. Which has the more relevant exposure? Where does pricing power sit? How much of the business is driven by something else?

That research can change the ticker list. It can also make you abandon the path entirely.

This is where your judgment matters. AI can help you generate more possibilities, collect the evidence and find holes in the reasoning. You decide which explanations hold up.

The next time you have a big macro view, resist jumping straight to a stock list. Ask what changes, follow the consequences, and choose the paths worth testing. You'll give yourself a much better chance of finding an idea beyond the obvious names.

-Felipe