Generating social media captions, writing emails, summarizing stuff you could just read yourself. Not really moving the needle.
Where I've actually seen AI make a real difference is in the operational side. Things like texting follow up sequences that respond based on what the lead actually said, deal analysis that pulls comps and runs numbers without you touching a spreadsheet, and making sense of large amounts of data fast so you know where to focus.
That's where it actually saves time and closes helps people close more deals
Curious about this, honestly, always trying to find new ways to integrate AI into my own business. Would love to hear what you guys are actually using it for and if it's making a real difference.
I 100% still review and analyze everything. I have tested AI on analyzing data and it works lets say 90% of the time (my number nothing to back that up), but the 10% it is off, it can be very detrimental. I use it to be more efficient, not to replace me.
I 100% still review and analyze everything. I have tested AI on analyzing data and it works lets say 90% of the time (my number nothing to back that up), but the 10% it is off, it can be very detrimental. I use it to be more efficient, not to replace me.
Yes, in deal analyzing I wouldn't rely on AI 100%, but I would use it to speed up my process!
However, incoming AI has been an absolute game changer! saved me a lot of money
I 100% still review and analyze everything. I have tested AI on analyzing data and it works lets say 90% of the time (my number nothing to back that up), but the 10% it is off, it can be very detrimental. I use it to be more efficient, not to replace me.
I think we're still looking at AI primarily as a productivity tool.
Read a PDF.
Extract data.
Summarize a report.
Populate a spreadsheet.
All useful applications.
But the bigger shift may happen when AI stops acting as a single assistant and starts acting as a system.
One process reviews documents.
Another compares them against historical outcomes.
Another identifies inconsistencies and potential risks.
Another prioritizes what actually deserves human attention.
Ironically, that may also be where accuracy improves.
Many of today's errors come from asking a single model to do everything by itself. In practice, experienced operators don't work that way either. Analysis, verification, and decision-making are usually separate steps.
At that point, the value isn't in generating another report. It's in compressing the time required to reach a decision.
Most investors aren't suffering from a lack of information. If anything, they have too much of it.
The real bottleneck is sorting signal from noise.
The operators I've learned the most from aren't valuable because they know more facts. They're valuable because they recognize patterns faster, identify risks earlier, and spend their time on the opportunities that matter.
A lot of what we call experience is really pattern recognition built over hundreds of deals.
That's why I think the long-term opportunity for AI isn't replacing judgment. It's helping people apply judgment faster and more consistently.
The biggest winners probably won't be the tools that generate the most content.
They'll be the tools that help someone avoid spending three weeks analyzing the wrong deal.
I’ve used AI only after I’ve got all the right data and buy box defined. I provide AI with a deterministic financial model and uses that data against the model to scan through properties in my target market. It helps me take a first pass at the properties that pass the initial sniff test. I then use AI to analyze property metadata against that market to understand any key risks that I may otherwise miss. I don’t use AI to underwrite and I validate all the numbers.
I’ve used AI only after I’ve got all the right data and buy box defined. I provide AI with a deterministic financial model and uses that data against the model to scan through properties in my target market. It helps me take a first pass at the properties that pass the initial sniff test. I then use AI to analyze property metadata against that market to understand any key risks that I may otherwise miss. I don’t use AI to underwrite and I validate all the numbers.
I’ve used AI only after I’ve got all the right data and buy box defined. I provide AI with a deterministic financial model and uses that data against the model to scan through properties in my target market. It helps me take a first pass at the properties that pass the initial sniff test. I then use AI to analyze property metadata against that market to understand any key risks that I may otherwise miss. I don’t use AI to underwrite and I validate all the numbers.
I’ve used AI only after I’ve got all the right data and buy box defined. I provide AI with a deterministic financial model and uses that data against the model to scan through properties in my target market. It helps me take a first pass at the properties that pass the initial sniff test. I then use AI to analyze property metadata against that market to understand any key risks that I may otherwise miss. I don’t use AI to underwrite and I validate all the numbers.
I’ve used AI only after I’ve got all the right data and buy box defined. I provide AI with a deterministic financial model and uses that data against the model to scan through properties in my target market. It helps me take a first pass at the properties that pass the initial sniff test. I then use AI to analyze property metadata against that market to understand any key risks that I may otherwise miss. I don’t use AI to underwrite and I validate all the numbers.
Thanks! I build AI products at work and pretty familiar with its uses and limitations at scale.
Good point - I haven’t found a reliable data source to pull title issues with a property. That would be great to have. I currently assess risk based on market conditions and to some extent the property. But this next level deeper would be great.
Im also looking around for any programmatic tools to integrate off market properties, if anyone has any leads.
At this point, AI is a lot of noise on the deal side for me.
I would not want AI just telling me if a deal is good or bad. In my experience it takes ~50 (slight exaggeration) questions to get it to think about what I already know to be true. But if an investor already has the assumptions, formulas, rules, and buy box built, then I can see AI being very helpful in applying a decision faster.
The part that worries me is that experienced investors will usually spot when the answer is wrong because they have seen enough deals. Newer investors may not catch it, and they do not know the questions to ask AI to discover the output it got wrong. Unfortunately, they are probably the ones most likely to trust it too much, which is dangerous.
And then, analyzing a single property is one thing. A portfolio decision is a different animal. Buy a new rental, refinance, 1031, sell another and now your equity, debt, cash flow, taxes, and future buying power can all change. It is not just whether that one property looks good. It is what that decision does to the portfolio.
So I agree with the operational side. That is where AI has been useful for me. Pull public data, flag possible issues, summarize PDFs, show what changed, and help me get to a right administrative decision faster.
But I still think the model and the judgment have to come from the investor. AI should speed up the process, not become the process.
At this point, AI is a lot of noise on the deal side for me.
I would not want AI just telling me if a deal is good or bad. In my experience it takes ~50 (slight exaggeration) questions to get it to think about what I already know to be true. But if an investor already has the assumptions, formulas, rules, and buy box built, then I can see AI being very helpful in applying a decision faster.
The part that worries me is that experienced investors will usually spot when the answer is wrong because they have seen enough deals. Newer investors may not catch it, and they do not know the questions to ask AI to discover the output it got wrong. Unfortunately, they are probably the ones most likely to trust it too much, which is dangerous.
And then, analyzing a single property is one thing. A portfolio decision is a different animal. Buy a new rental, refinance, 1031, sell another and now your equity, debt, cash flow, taxes, and future buying power can all change. It is not just whether that one property looks good. It is what that decision does to the portfolio.
So I agree with the operational side. That is where AI has been useful for me. Pull public data, flag possible issues, summarize PDFs, show what changed, and help me get to a right administrative decision faster.
But I still think the model and the judgment have to come from the investor. AI should speed up the process, not become the process.
Exactly. Chats are useful for a lot of things, but the real benefit is feeding it structured, real data and letting it summarize/reason against that instead of guessing, or worse, spending hours feeding it data you already know but need it to know.
Give it a defined model. Pull the actual comps. Then let it analyze from there. That is very different than asking it cold.
The part I would add is that the model underneath still has to be deterministic. If your foundation is solid and AI is working on top of a proven system, you can get speed without giving up trust.
But if the model itself is fuzzy, AI would not fix that. It just gives you the wrong answer faster.
I agree. The content-generation side of AI gets most of the attention, but the bigger value I’ve seen is in helping operators make decisions faster and automate repetitive work.
In real estate, I’ve used AI for things like:
The biggest benefit hasn’t been replacing people—it’s reducing the time spent on tasks that don’t directly generate revenue.
I think we’re still early. The investors who figure out how to combine AI with their existing acquisition, disposition, and operations processes are going to have a significant advantage over the next few years.
I’m always interested in seeing how others are applying it as well. If anyone is experimenting with AI for acquisitions, underwriting, lead management, or portfolio operations, I’d love to hear what’s working. And if I can be helpful based on what we’ve built and tested, feel free to reach out.
We use AI to deal with the texting and letting us know when someone is ready to sale. That ad rating the properties for us so we dont waste time calling people that wont sale. We tested it 3 times and its a for sure thing that we get contracts with every cold text within 48 hours of the campaign launching. Those are the two main ways we use it for now that I am 100% confident on.
We use AI to deal with the texting and letting us know when someone is ready to sale. That ad rating the properties for us so we dont waste time calling people that wont sale. We tested it 3 times and its a for sure thing that we get contracts with every cold text within 48 hours of the campaign launching. Those are the two main ways we use it for now that I am 100% confident on.
@Rauph Souleimanov Thats a lot of messages bro! I'll hit that number in 2 years haha dang But yea AI helps a lot, im honestly thinking it won't be this good for long, and it will heat weaker or way more expensive to only the money makers can continue. Hopefully im just tripping though
@Rauph Souleimanov Thats a lot of messages bro! I'll hit that number in 2 years haha dang But yea AI helps a lot, im honestly thinking it won't be this good for long, and it will heat weaker or way more expensive to only the money makers can continue. Hopefully im just tripping though
lol it is for sure!, finding so many properties is becoming my problem for sure!
For now its pretty cheap we can run 100k messages total with everything only for about $800-850! including messaging costs etc.
But who knows what it will be like eventually! but right now it is the best time to take advantage of it
We do a lot of the similar things above. It's actually really good at comps once trained. We have skills trained to process 95% of the applications very quick. I also build small apps specific to our company. I have a estimate tool for large rehab projects. I build a project management tool to manage turnovers. Some others. These have saved the team many hours every week, compared to what we used to do. They are also pretty easy to deploy to the whole team and use internally.
If I had to guess I'd say it saves me 8hrs a week and then I'm currently spending 15 hours a week messing around with it. 😂
We do a lot of the similar things above. It's actually really good at comps once trained. We have skills trained to process 95% of the applications very quick. I also build small apps specific to our company. I have a estimate tool for large rehab projects. I build a project management tool to manage turnovers. Some others. These have saved the team many hours every week, compared to what we used to do. They are also pretty easy to deploy to the whole team and use internally.
If I had to guess I'd say it saves me 8hrs a week and then I'm currently spending 15 hours a week messing around with it. 😂
RCIC Operators are generating quality opportunities aka leads in abundance. Full intelligence including loan information, lender information and lender attorney information.
Now these are coming with comps, mra and the info you need to determine the arv range (before rehab and scope of work factors) and skip tracing if their BuyBox matches.Then offer generation, capital passport with lender matching and transaction management.
AI can do what you train it to do. A lot of people will like they say be replaced by it.
WOPR (War Operation Plan Response), also known as Joshua, is the artificial intelligence supercomputer featured as the main antagonist in the 1983 film WarGames.
Rauph, I agree completely that the more specific the application of AI to your operations, the higher the leverage. I think of it in this terms (for customers): where do you spend the most time OR money, and how can AI reduce that? Sometimes in not so obvious, but mostly the answer comes around to operations (time) or headcount (money) for companies.
At my job we built a tool for lease abstraction using AI (not trying to push this, it's very specific for our system). Have you seen any of this in the wild? I'm new to RE so was suprised how much money + time was being saved by this.
AI adds a lot of value when the task is repetitive, clerical, or research-assisted. But it has an important weakness: when it does not know the answer, it may still give one.
I use AI every day, but not to analyze investment properties. That work requires judgment, local market knowledge, and data that must be verified. AI can help organize information, but it should not replace experience or original analysis.
Where AI helps me most is with writing and research support.
Used this way, AI is a helpful assistant. It saves time, improves clarity, and helps locate useful information.
However, Rentometer, Zestimate, and AI fail because they have invalid inputs.
Rentometer fails because of its method. It averages comparable rentals within a radius using a few structural fields, such as bedroom count, bathroom count, square footage, and property type.
That sounds useful, but it misses the details that actually determine rent.
It cannot see the property’s current condition. It cannot hear freeway noise. It cannot smell smoke or pet damage. It cannot tell whether the photos are current. It cannot understand whether one side of a subdivision rents well while the other side does not. It cannot judge renovation quality, street appeal, or tenant demand at the micro-location level.
In other words, Rentometer is structurally blind to the factors that drive rent and price.
AI has the same problem.
AI is not wrong because it is AI. It becomes unreliable when it is fed the same thin, structural data. If a model reads only a Zestimate-style feature set, it may be just as wrong as Rentometer. The limiting factor is not the technology. The limiting factor is the input.
Give a model, or a person, better information and the estimate improves. That means current renovation photos, the exact street location, recent signed leases for true comparable properties, concession data, days on market, and local property manager feedback.
Even then, the information must be verified. Photos may be outdated. Listing descriptions may exaggerate condition. A property may look acceptable online but fail in person because of noise, odor, poor layout, deferred maintenance, or a bad location within the neighborhood.
This is why we do not rely on automated rent estimates or AI-generated rent estimates by themselves. They can be useful starting points, but they are not investment analysis. Real analysis requires verified data, local judgment, and a clear understanding of what the target tenant will actually rent.
Hey Eric,
I think of AI as augmentation, not replacement. On rent estimation, though, it's pushing the boundary further than Zestimate and Rentometer suggest.
I've built a tool that "sees" renovation quality from photos and adjusts prices using a pretrained ML model. The idea is to use every input that exists as data. Odor isn't one of them, and neither is human judgment/localized knowledge.
AI is just moving the starting point you mentioned much closer to reality.
@Josh C. the turnover tool is the one that jumps out — that's usually the messiest thing to systemize since it touches so many people (vendors, tenants, leasing) at once. How are you tracking status across the different stages — one shared view everyone checks, or does it ping you when a stage completes?
Also that last line made me laugh, the 15 hours "messing around with it" is a very real tax nobody warns you about.
Agree completely on the follow up sequencing point, that's the one place I've seen AI actually change outcomes rather than just save typing time, because it's reacting to what the lead said in real time instead of a generic template. The captions or emails stuff is a nice to have but the response time stuff is the one that shows up in closed deals.
I'm on the agent side rather than investing, and I'd agree with part of this. Asking AI to "write me a caption" or "summarize this" as a one-off is low value. You could've done it yourself in the same time.
Where it's actually saved me time is when I stopped treating it like a one-off tool and started reusing the same well-built prompt over and over. Same structure every time: property type, price point, target buyer, tone, etc. For listing descriptions and follow-up emails after showings. The first time you write that prompt takes effort, every use after is basically instant.
The other place it's genuinely helped: drafting 2-3 response options for a tricky negotiation message when I give it the full back-and-forth, not just "write my reply." I still edit before sending, but it kills the "staring at a blank message" time.
Where I'd fully agree with you — anything needing real local market knowledge (comps, HOA rules, that kind of thing) it just makes up if you're not careful. Never let it touch anything client-facing without checking facts myself.