EP 387 AI Is Making Bad Investors Worse

APPLE PODCASTS | SPOTIFY

If someone asked you to choose between three real estate markets, which one would you pick?

The fastest-growing one?

The one everyone is moving to?

The city with the biggest employers, strongest job growth, and cranes seemingly everywhere?

That’s probably where we would’ve started too.

But after more than seven years of investing in multifamily real estate, we’ve learned something that sounds obvious only after you’ve learned it the expensive way:

A great city doesn’t automatically make a great real estate investment.

My sister Nancy and I have built a real estate portfolio approaching $500 million across multiple markets. We’ve returned more than $45 million to investors and helped create more than $110 million in tax savings.

We’ve underwritten through changing interest rates, shifting rents, booming markets, and soft ones.

We’ve gotten things right.

And we’ve gotten enough things wrong to know that the market everyone is talking about isn’t necessarily the market where your money should go.

That’s why we started building something different.

An AI Market Research System that could take the way we evaluate a real estate market and turn it into a repeatable process.

Then we tested it.

We gave it three North Carolina markets:

Raleigh. Charlotte. Greensboro.

And the result surprised even us.

But before we tell you which one won, you need to understand why researching real estate markets is harder than it looks.

“Is Raleigh a Good Real Estate Market?”

This is how a lot of AI market research starts.

Open ChatGPT.

Type:

“Is Raleigh a good market for real estate investing?”

And you’ll probably get something that sounds pretty convincing.

Growing population.

Strong employment.

Business-friendly environment.

Great universities.

Growing technology sector.

Wonderful.

Except…

What do you actually do with that?

Where’s the construction pipeline?

What’s happening with absorption?

Are rents growing?

Are concessions increasing?

What’s happening with occupancy?

How quickly can a landlord remove a non-paying tenant?

And how current is the information you’re looking at?

This is one of our biggest frustrations with generic AI research:

It gives you adjectives when you need arithmetic.

So You Open 17 Browser Tabs Instead

We’ve all done this.

One website for population.

Another for rent growth.

Another for employment.

Then a broker report.

Then a Census page.

Then someone sends you a “Top 10 Real Estate Markets for 2026” article.

Before you know it, your browser looks like it needs an intervention.

And there’s a bigger problem.

Some of the people publishing those lists have an economic interest in the markets they’re recommending.

Maybe they have properties to sell there.

Maybe they’re raising capital for a deal there.

Maybe they’re a brokerage with listings there.

That doesn’t mean they’re lying.

But it does mean you should understand who is giving you the information and why.

Then there’s the data itself.

One source’s “rent growth” might mean asking rents.

Another might mean effective rents after concessions.

Those aren’t necessarily the same number.

Now you’ve taken seven numbers from seven sources, measured at seven different times using seven different definitions…

And put them into one spreadsheet as if they’re perfectly comparable.

At some point, that’s not really research anymore.

It’s a scavenger hunt.

Or We Do What Humans Naturally Do: Invest Somewhere Familiar

This one we understand.

You invest in your backyard.

You know the neighborhoods.

You’ve watched that coffee shop open.

You know which freeway everyone hates.

You remember when that entire development was still dirt.

It feels safer because it’s familiar.

But this is one of the most important distinctions we’ve learned:

Familiar is not the same as good.

And the opposite is also true.

A market you’ve never visited isn’t automatically a bad place to invest.

Your backyard.

Someone else’s “best markets” list.

A podcast recommendation.

A generic ChatGPT answer.

They’re all different versions of the same problem:

We’re substituting familiarity for fundamentals.

So We Built the Market Research System We Wanted

We wanted something repeatable.

Give the system a city.

Pull current market information.

Then evaluate that market using the same framework every single time.

Our system grades markets across seven factors and ultimately puts them into one of three buckets:

Invest-grade. Watch. Pass.

But there’s another piece we care deeply about.

Sources.

We want to know where the number came from so we can verify it ourselves.

Because AI shouldn’t simply hand you a confident answer.

It should help you get closer to the evidence.

Some of the multifamily supply and absorption information institutional investors use can also sit behind expensive data subscriptions.

That’s part of what makes AI so interesting to us.

The data wasn’t necessarily hidden from individual investors. A lot of it was simply priced away from them.

So we decided to run a real test.

First up:

Raleigh.

Raleigh Looked Almost Perfect

If you only looked at the big-picture story, Raleigh would be pretty easy to love.

Job growth?

A-minus.

Migration?

A.

Industry diversity?

A-minus.

The market has major technology, life sciences, pharmaceutical, higher education, and government employment.

Population was growing around 2.4% annually in the data we ran, with roughly 39,000 people moving into the market net each year.

That’s the kind of story investors love.

More jobs.

More people.

More demand.

What could possibly go wrong?

Then we got to rents.

C-plus.

Rent growth was essentially flat at about 0.1%.

Occupancy was around 93.1% and slipping.

Wait.

How does one of the fastest-growing markets in the region have almost no rent growth?

This is exactly where market research gets interesting.

Demand Wasn’t Raleigh’s Problem

Here’s the number that changed the story.

Raleigh had absorbed approximately 19,400 units since 2024 in the data set we analyzed.

People were renting the apartments.

Demand existed.

So why weren’t rents moving?

Supply.

Inventory had grown roughly 15% across 2023 and 2024.

Think about that.

A market can have incredible population growth and incredible job growth…

And still experience flat rents if developers build apartments faster than demand can absorb them.

That’s the part the headline misses.

But then we saw another number.

The pipeline had fallen roughly 30% over the prior year.

And this is where the distinction between data and judgment starts to matter.

Most people look at how many apartments are under construction today.

We also want to know which direction that number is moving.

Because today’s construction tells you about today’s competition.

The direction of the pipeline starts telling you about tomorrow’s.

Raleigh had already experienced the supply wave.

Now the wave appeared to be shrinking.

So while the system gave rent trends a C-plus, we saw something else:

That C-plus might have an expiration date.

One Number We Think More Investors Should Watch

There’s another metric we pay close attention to:

Rent-to-income.

Because rents can’t rise forever simply because an investor’s spreadsheet says they should.

At some point, a human being has to write the check.

Raleigh’s median income was around $85,000 in our data, putting rent-to-income at roughly 22%.

Why does that matter?

Because once renters become too stretched, your ability to continue pushing rents becomes constrained.

We like to say:

Rent doesn’t rise to what the market will bear. It rises to what a paycheck will bear.

Raleigh still appeared to have room.

So the system’s final verdict?

WATCH.

Great demand.

Soft current pricing.

Shrinking supply pipeline.

Not necessarily the market we’d choose for day-one cash flow, but one we’d keep watching.

And this is where we disagreed slightly with our own AI.

The system was grading what had already happened.

We were thinking about what might happen next.

The system graded the weather. We were looking at the season.

Then We Ran Charlotte

Charlotte’s migration score came back even stronger.

A-plus.

In the data we analyzed, Charlotte was adding roughly 157 residents every day.

That’s an extraordinary demand story.

Then we looked at rents.

Down roughly 3.2%.

Not for one quarter.

For 11 straight quarters.

If you stopped there, you might conclude:

Charlotte is broken.

But then came the number that changed our interpretation.

Vacancy was only around 6.2%.

That’s important.

If nobody wanted the apartments, you’d expect empty units everywhere.

But that wasn’t what the data showed.

People were renting them.

They were simply renting them at lower prices while a huge wave of new supply worked its way through the market.

Roughly 12,000 units had been absorbed against approximately 13,000 delivered.

That isn’t disappearing demand.

That’s competition.

Falling rents with relatively low vacancy can be a supply story, not a demand story.

And supply stories can eventually end.

Then Greensboro Ruined Our Beautiful Theory

At this point, Raleigh and Charlotte were telling similar stories.

Amazing migration.

Strong economies.

Soft rents.

Large amounts of new supply.

Then we ran Greensboro.

Migration?

C-plus.

The weakest demand grade of the three.

Population growth was around 1%.

If you were building a “Top North Carolina Growth Markets” list, Greensboro probably wouldn’t be your winner.

Then the system gave us its verdict.

INVEST-GRADE.

Wait…what?

The market with the weakest growth story was the only invest-grade market?

That’s when we started digging.

Supply had grown around 7%, compared with roughly 15% in Raleigh.

There were approximately 1,400 units under construction, with that figure declining year over year.

Rents were up around 2.5%.

Occupancy was above 95% and improving.

Median home prices were also considerably lower than Raleigh and Charlotte.

Then there’s the Toyota battery plant.

Roughly $14 billion.

Around 3,000 jobs.

And importantly, production had already begun.

Not an announcement.

Not a ribbon-cutting planned for five years from now.

Actual economic activity.

But Even Good News Can Become a Risk

Here’s the part we don’t want investors to skip.

A $14 billion plant bringing thousands of jobs sounds fantastic.

And it is a meaningful catalyst.

But Greensboro still only received a B for industry diversity.

Why?

Because a single employer can be both an opportunity and a concentration risk.

Every job that employer creates makes the market stronger.

And simultaneously makes part of the market more dependent on that employer.

Both can be true.

That’s why we don’t want AI simply telling us:

“Toyota investment = good.”

We want it asking:

“What happens if we’re wrong?”

Good news should still be underwritten.

Not simply celebrated.

The Fastest-Growing Market Isn’t Necessarily the Best Investment

This was the big lesson from running all three.

Raleigh and Charlotte had the strongest growth stories.

Greensboro had the weakest.

And yet Greensboro was the only market our system rated invest-grade at that moment.

Why?

Because here’s the strange thing about growth:

Growth attracts capital.

Capital attracts developers.

Developers create supply.

Supply creates competition.

And competition can suppress rents.

In other words, some of the exact things that made Raleigh and Charlotte so attractive also helped fill their skylines with cranes.

That’s why we don’t think Greensboro was necessarily “better” than Charlotte.

It may simply have been earlier.

And that distinction matters enormously.

Because the best market isn’t always the market with the most growth.

Sometimes it’s the market where demand and supply are currently most mispriced in your favor.

And This Is Where AI Stops

Here’s the part of this experiment we care about most.

The AI system could collect the information.

It could organize it.

Grade seven different factors.

Compare three markets.

Source the numbers.

And give us a verdict.

But it couldn’t hand us the judgment required to act on that verdict.

Take Charlotte.

Eleven consecutive quarters of falling rents.

That’s a fact.

One investor reads that and says:

“Stay away.”

Another looks at the same number, sees relatively low vacancy and a declining construction pipeline, and thinks:

“Supply is clearing.”

Same data.

Completely different decision.

So which one is right?

That’s where experience enters the room.

We’ve underwritten through a full rate cycle.

We’ve watched assumptions fail.

We’ve made mistakes.

We’ve seen what happens after the spreadsheet ends.

And that experience changes how we interpret the numbers.

The system gave us the read. The scar tissue helped give us the answer.

That’s the Part of AI Nobody Should Skip

AI can dramatically change real estate market research.

We’re watching it happen in our own business.

It can help investors gather information faster.

Compare markets faster.

Find inconsistencies faster.

See relationships that would’ve required hours of manual research.

But speed isn’t the ultimate advantage.

Judgment is.

Because eventually someone is going to send you a deal.

Maybe it’s Raleigh.

Maybe Charlotte.

Maybe Phoenix, Dallas, Atlanta, Nashville, Tampa—or whatever market everyone starts talking about next.

You’ll see the population growth.

You’ll see the jobs.

You’ll see the rent number.

You’ll see the construction pipeline.

And maybe AI will even give you the correct grade.

Then comes the harder question:

What does the grade actually mean?

Is that C-plus a market that’s deteriorating?

Or a market that’s early?

Is falling rent evidence of disappearing demand?

Or temporary supply pressure?

Is that massive employer announcement a catalyst?

Or a concentration risk?

The data can help you see the pattern.

But eventually, you still have to make the decision.

Want to See How We Actually Use AI to Invest?

If this made you realize there’s a layer beneath simply asking ChatGPT “Is this a good real estate market?”—that’s exactly what we want you to see next.

We created a free masterclass called:

Become the Commercial Real Estate Investor Who Uses AI to Outperform Everyone Else

Inside, we go deeper into how we use AI against real deals, how these systems are built, and—most importantly—how we apply investment judgment on top of what AI finds.

Because building a sophisticated AI system isn’t the goal.

Building a sophisticated AI system pointed at the right questions is.

And if you’re wondering where your own wealth structure may be carrying more risk than you realize, we also built Where Wealth Breaks.

It’s completely free, takes less than three minutes, and uses seven questions to help identify where your current wealth architecture may be most exposed.

The biggest takeaway from this experiment wasn’t that Greensboro beat Raleigh or Charlotte.

Those rankings will change.

Markets always do.

It was this:

You don’t need someone else’s list of the best markets.

You need a process for understanding the market in front of you.

Because better access to data is useful.

Better access to deals is useful.

But the thing that ultimately changes the way you invest?

Access to your own judgment.

And that’s what we’re really trying to build.