Dear CRE Investors: AI is Coming for our Jobs, too
And what systems won't ever be able to replace
Recently, I have seen more and more real estate investors waking up to the transformative power of AI.
Which is about right: real estate is not known for living on the technological cutting edge. So we’d be wise to pay attention before we’re left behind.
Here are the top 10 ways AI is making CRE investors and brokers obsolete, in the rough order you'll be replaced:
(Plus, the one essential element of real estate investing that AI can never replace)
1) Cold calling
Kind of a no-brainer here: The days of grinding through a call list, fingers crossed for a listing, are numbered.
LLMs and voice cloning technology are already here, and word is this is already being tried for real estate cold calls.
There are legal hurdles to surmount (AI-robocalling has been outlawed), but the potential here might just be juicy enough for brokerages to find a way.
2) Comparables /market research
This one's already here, just don't ask RealPage how machine learning for apartment rents is going.
LLMs are really just sophisticated pattern recognition engines (ChatGPT confirmed that this characterization was OK and wouldn’t get me put on any sort of “list”), which is all we humans do when choosing comps and evaluating markets.
And with the tremendous amount of street-level data being generated and collected right now (traffic data, office keycard swipes, point of sale data, etc), systems designed to pick up patterns will soon stay ahead of even the most plugged-in flesh and blood players.
3) Internal ops and efficiency
This one isn't sexy, but as real estate operators struggle to coalesce data from disparate systems, AI is here to help.
Fort Capital in Texas is ahead of the curve on this, and put out a solid podcast episode detailing how their firm has built an AI agent to connect internal teams to their data.
4) OM creation and analysis
Every wonder why every offering memo reads the same?
Just wait until AI-generated OMs become the norm, as systems spin operating data and dynamic market narratives into a tidy package that teams of junior brokers and designers would have spent weeks creating.
Startups are already working on this, and the technology is improving every day.
The flip side of AI-generated OMs is a computer sucking the relevant data out of an OM, synthesizing it, and delivering investors a summary and pre-formatted templates to plug into their pro formas.
Off-the-shelf LLMs are pretty darn good at this already, and startups are developing bespoke tools aimed specifically at real estate investors.
5) Financial modeling
Financial models won't be far behind, so if you think that a decade perfecting mouseless Excel will land you your dream REPE job, think again.
LLMs are still not great at math, but existing plug-ins for Excel are already speeding up modeling tasks, improving iterative analysis and inflating the false precision inherent in fancy pro formas.
Existing systems can already spit out pretty reliable cash flow projections and return metrics, which may soon render the backs of many napkins obsolete.
6) Lease review
The tech is already here to read and summarize PDFs, but be patient on this one.
Just like lawyers have a few years before AI bots take their jobs because matters of law require precision, so do lease reviews.
It’s convenient that AI can pull lease terms out of a lease, but are you ready to trust the results and remove contingencies?
Lease forms are wildly divergent, as are lingo and nuance, and when you throw in scans of printed documents which require OCR (optical character recognition) and a ton of compute, lease review is one of those challenges it will take years for AI systems to master with an acceptably low error rate.
Sure, you could save some legal fees, but your LPs pay those anyway.
7) Investor reporting
The closer we get to impinging on the relationship between GPs and their investors, the further off the impending AI Armageddon gets.
So long as your signature is at the bottom of a report that includes financial statements and delicately delivered bad news, how long will it take GPs to outsource this role to our robot overlords?
Of course, LLMs can help write narratives and even format reports, but anything that includes a direct connection GPs and their capital will be at the tail end of the AI replacement spectrum.
8) Pitch creation
I’ve tried to have AI write my pitch documents for me – it doesn’t work.
Will it be able to eventually? Sure, but try asking ChatGPT or Claude about something you know really, really well, and see what comes back.
Achieving that level of detail and nuance requires a tremendous amount of data and careful prompting, all which must be done by the human who holds that information.
Similar to the investor reporting, while LLMs and other AI can help create the backup material, graphics and data for a pitch, GPs are going to want their personal touch on the final versions of any documents that goes out to a capital partner.
Or they should anyway.
9) Physical due diligence
A lot of people lean on the physical aspect of real estate as a reason that AI isn't an existential risk.
But do you really think your unreliable eyes can take in more detail than a robotaxi or drone programmed to count things like sushi shops, cafes, tagged walls, street entrepreneurs and other "subjective" measures of a neighborhood?
What about a roomba-like robot crawling through a building identifying dry rot, detecting mold, measuring square footage, estimating capex based on previous projects, and filming the whole thing?
That clipboard and site inspection checklist won't seem like much of an edge in the not-too-distant future.
10) Deal evaluation
I could be talking my book on this one, but there's a chasm between pulling rent and operating data out of an OM and knowing in your gut that there's a deal to be had.
I’d like to think there’s an art to this, but alas, this is likely just pattern recognition too, so I should probably ask Chat GPT to write me a new resume.
And lastly, what AI is not coming for:
Relationships
In time, all the thinking tasks we do will be done better by systems.
It’s just a matter of plugging in the right data and writing the right prompts. And eventually even those prompts will be written by the systems themselves.
But AIs don’t drink coffee, don’t sip whiskey, and they definitely don’t laugh and cry and connect with fellow humans (yet).
So if all this talk of AI taking your job is getting you down, turn off your computer (while you still can), go outside, and take a meeting.