AI Is Standardizing Note Pricing – Good News for Buyers, Tougher for Low-Rate
Member since 2026 · 35 posts · 12 votes
AI Is Standardizing Note Pricing – Good News for Buyers, Tougher for Low-Rate
Most Popular Reply
Investor · VA · Member since 2015 · 21k+ posts · 19k+ votes
7moI will throw my 2 cents in without using AI...
1. AI speeding up note valuation - Borrower credit and comps have been able to be reviewed quickly forever. a AVM is AI. A credit score is AI. They are algorithms to determine a datapoint which now everything gets coined AI.
2. Pricing has always been fairly consistent, the key is finding the needles in haystacks.
3. Low rate notes have never been very attractive. If we see a low rate we have a simple factor of being able to look at the rate and how far its delinquent and say "no" without having to use AI. Its pretty easy to see when you have run 1000x.
4. Bruce is correct, buyers target yield. Thats how I have been doing it the past decade.
5. Buyers say they are using AI, but are they? What competitive advantage is it doing? It may replace a VA from sorting through collateral, but its replacing a $5 VA.
6. AI is not speeding up due diligence. It still requires chasing the seller for the collateral files, chasing down title company for title reports, chasing attorneys to review collateral. That has not changed.
7. AI is not going to pick up the borrowers BK and mindset, what the inside of the property looked like, many NPL borrowers do not act rationally or what you would think is the norm, which is what AI is based on.
8. Automation is not reducing servicing costs.
The ones winning in notes in 2026 are those who know how to manage the notes since property values are declining and costs are increasing.
7e investments53 Reviews
14 Replies
- Real Estate Broker · Coppell, TX · Member since 2011 · 5k+ posts · 4k+ votes7mo
I disagree with all your assumptions. I doubt hardly anyone is using AI to price their notes. Are you? Anyone you know?
I'm not a note buyer, but I would guess most note buyers have a specific yield number they want to hit and don't care much about the rate? They just adjust pricing to hit their target yield. Are you ignoring all 5% and lower notes?
- Member since 2021 · 82 posts · 79 votes7mo
Both of you are having the wrong argument. AI isn't "standardizing" note pricing — it's exposing the spread that's always existed between what sellers think their note is worth and what the yield math actually demands. And Bruce is right that yield-to-target pricing hasn't changed. But he's wrong that nobody's using AI to get there faster.
Here's what's actually happening in the note market in 2026, and why it matters whether you're buying or selling.
The yield math hasn't changed. The speed of the math has.
A performing 1st lien note at 5% with 25 years remaining and a $150K UPB has always been worth whatever discount produces the buyer's target yield. That's not new. What's new is that a buyer can now pull borrower credit, property valuation, payment history, comparable note sales, and default probability in minutes — not weeks. The "standardization" Frances is describing isn't AI changing the formula. It's AI compressing the timeline between "I see a note" and "I know exactly what I'll pay."
That benefits disciplined buyers, not sellers.
The tax angle nobody in this thread mentioned:
Under IRC §1272, if you purchase a debt instrument at a discount to face value, the discount is treated as Original Issue Discount (OID) and accrues as ordinary income over the life of the note — taxed annually whether or not you receive it. This is the hidden tax cost of buying discounted notes that most buyers don't model until their CPA sends them a surprise bill.
However, if the note qualifies as a "market discount bond" under IRC §1276, you can elect to defer recognition until disposition — meaning you don't pay tax on the phantom income annually. The election must be made on your return for the year of acquisition. Miss it, and you're stuck with annual OID inclusion.
On a $150K UPB note purchased at $112K (targeting 9% yield), that's $38K in market discount. At a 37% marginal rate, the difference between OID
treatment and market discount deferral election is $14,060 in tax timing — money that stays in your pocket compounding instead of going to the IRS annually.
The financial reality on low-rate notes:
Frances is correct that discounts are widening. Here's the math:
┌───────────┬──────────┬────────────────────┬────────────────┬──────────┐
│ Note Rate │ UPB │ Buyer Target Yield │ Purchase Price │ Discount │
├───────────┼──────────┼────────────────────┼────────────────┼──────────┤
│ 5.0% │ $150,000 │ 9% │ $103,400 │ 31.1% │
├───────────┼──────────┼────────────────────┼────────────────┼──────────┤
│ 5.0% │ $150,000 │ 10% │ $95,200 │ 36.5% │
├───────────┼──────────┼────────────────────┼────────────────┼──────────┤
│ 5.0% │ $150,000 │ 12% │ $82,100 │ 45.3% │
├───────────┼──────────┼────────────────────┼────────────────┼──────────┤
│ 7.0% │ $150,000 │ 9% │ $126,800 │ 15.5% │
├───────────┼──────────┼────────────────────┼────────────────┼──────────┤
│ 7.0% │ $150,000 │ 10% │ $118,400 │ 21.1% │
└───────────┴──────────┴────────────────────┴────────────────┴──────────┘
A 5% note has to be discounted 31-45% to hit current buyer yield targets. A 7% note only needs 15-21%. That's not AI — that's arithmetic. But
AI lets buyers run these scenarios across 500 notes in the time it used to take to underwrite one.
Bruce's point stands: buyers have a target yield and back into the price. AI just makes that calculation instantaneous across larger pools. The
sellers who get hurt are the ones who think their 5% performing note is worth par because "payments are current." Payments being current doesn't change the fact that a buyer can get 5.5% in a Treasury with zero risk. Your 5% note has to compensate for credit risk, liquidity risk, servicing cost, and default probability — ON TOP of competing with risk-free rates. The operational truth from 26 years of managing the collateral behind these notes: The piece neither of you mentioned: collateral condition risk. AI can pull property valuations, tax assessments, and even satellite imagery to assess collateral. But it cannot assess deferred maintenance, which is the #1 driver of collateral value erosion in residential notes.
A note secured by a property with $40K in deferred maintenance is worth less than a note secured by a well-maintained property — even if the
borrower pays perfectly. Because if the borrower STOPS paying, your recovery through foreclosure is based on collateral value, not UPB.
Properties with deferred maintenance lose 15-30% of their value at foreclosure auction versus well-maintained comparables. I've seen it hundreds of times managing portfolios with notes in various states of collateral health.
AI can't see a failing HVAC system from a data feed. It can't see the roof that's 2 years past replacement. It can't see the foundation crack that wasn't in the inspection report. If you're buying notes purely on data without boots-on-ground collateral assessment, you're pricing in a
recovery value that doesn't exist.
What nobody mentioned:
1. Servicing cost compression. AI isn't just changing buyer pricing — it's changing servicing economics. Automated payment tracking, default prediction, and borrower communication are cutting servicing costs from $15-25/loan/month to $5-10. That means smaller buyers can now profitably hold notes that were previously only economical for institutional servicers. The buyer pool is expanding, which should theoretically support pricing — except that yield expectations are rising faster than servicing costs are falling.
2. The IRC §1276 election is your edge. Most note buyers don't make the market discount deferral election because their CPA doesn't specialize in debt instruments. If you're buying at deep discounts, this election is the difference between paying tax on phantom income every year and deferring it until you sell or the note pays off. That's a compounding advantage that adds 50-150 basis points to your effective yield over a
10-year hold.
3. Default probability models are only as good as local market data. AI pulls national default curves, but note performance is hyperlocal. A borrower in a market with 3% unemployment and rising home values has a fundamentally different default probability than one in a market with 7% unemployment and flat values — even if their credit scores are identical. The AI doesn't weight this correctly yet. Buyers who know their local markets still have an edge over algorithmic buyers using national models.
Bottom line: AI isn't changing what notes are worth. It's changing how fast buyers figure out what they'll pay. That compresses seller negotiating time, widens discounts on low-rate paper, and rewards buyers who combine algorithmic speed with operational knowledge of the underlying collateral. The winners in 2026 note buying aren't the ones with the best AI — they're the ones who use AI for speed AND still know what a bad roof looks like.
Nicholas Cokas — 26 years in property management and real estate operations
- Thanks for the AI slop response to the AI slop question
Quote from @Nicholas Cokas:
Both of you are having the wrong argument. AI isn't "standardizing" note pricing — it's exposing the spread that's always existed between what sellers think their note is worth and what the yield math actually demands. And Bruce is right that yield-to-target pricing hasn't changed. But he's wrong that nobody's using AI to get there faster.
Here's what's actually happening in the note market in 2026, and why it matters whether you're buying or selling.
The yield math hasn't changed. The speed of the math has.
A performing 1st lien note at 5% with 25 years remaining and a $150K UPB has always been worth whatever discount produces the buyer's target yield. That's not new. What's new is that a buyer can now pull borrower credit, property valuation, payment history, comparable note sales, and default probability in minutes — not weeks. The "standardization" Frances is describing isn't AI changing the formula. It's AI compressing the timeline between "I see a note" and "I know exactly what I'll pay."
That benefits disciplined buyers, not sellers.
The tax angle nobody in this thread mentioned:
Under IRC §1272, if you purchase a debt instrument at a discount to face value, the discount is treated as Original Issue Discount (OID) and accrues as ordinary income over the life of the note — taxed annually whether or not you receive it. This is the hidden tax cost of buying discounted notes that most buyers don't model until their CPA sends them a surprise bill.
However, if the note qualifies as a "market discount bond" under IRC §1276, you can elect to defer recognition until disposition — meaning you don't pay tax on the phantom income annually. The election must be made on your return for the year of acquisition. Miss it, and you're stuck with annual OID inclusion.
On a $150K UPB note purchased at $112K (targeting 9% yield), that's $38K in market discount. At a 37% marginal rate, the difference between OID
treatment and market discount deferral election is $14,060 in tax timing — money that stays in your pocket compounding instead of going to the IRS annually.
The financial reality on low-rate notes:
Frances is correct that discounts are widening. Here's the math:
┌───────────┬──────────┬────────────────────┬────────────────┬──────────┐
│ Note Rate │ UPB │ Buyer Target Yield │ Purchase Price │ Discount │
├───────────┼──────────┼────────────────────┼────────────────┼──────────┤
│ 5.0% │ $150,000 │ 9% │ $103,400 │ 31.1% │
├───────────┼──────────┼────────────────────┼────────────────┼──────────┤
│ 5.0% │ $150,000 │ 10% │ $95,200 │ 36.5% │
├───────────┼──────────┼────────────────────┼────────────────┼──────────┤
│ 5.0% │ $150,000 │ 12% │ $82,100 │ 45.3% │
├───────────┼──────────┼────────────────────┼────────────────┼──────────┤
│ 7.0% │ $150,000 │ 9% │ $126,800 │ 15.5% │
├───────────┼──────────┼────────────────────┼────────────────┼──────────┤
│ 7.0% │ $150,000 │ 10% │ $118,400 │ 21.1% │
└───────────┴──────────┴────────────────────┴────────────────┴──────────┘
A 5% note has to be discounted 31-45% to hit current buyer yield targets. A 7% note only needs 15-21%. That's not AI — that's arithmetic. But
AI lets buyers run these scenarios across 500 notes in the time it used to take to underwrite one.
Bruce's point stands: buyers have a target yield and back into the price. AI just makes that calculation instantaneous across larger pools. The
sellers who get hurt are the ones who think their 5% performing note is worth par because "payments are current." Payments being current doesn't change the fact that a buyer can get 5.5% in a Treasury with zero risk. Your 5% note has to compensate for credit risk, liquidity risk, servicing cost, and default probability — ON TOP of competing with risk-free rates. The operational truth from 26 years of managing the collateral behind these notes: The piece neither of you mentioned: collateral condition risk. AI can pull property valuations, tax assessments, and even satellite imagery to assess collateral. But it cannot assess deferred maintenance, which is the #1 driver of collateral value erosion in residential notes.
A note secured by a property with $40K in deferred maintenance is worth less than a note secured by a well-maintained property — even if the
borrower pays perfectly. Because if the borrower STOPS paying, your recovery through foreclosure is based on collateral value, not UPB.
Properties with deferred maintenance lose 15-30% of their value at foreclosure auction versus well-maintained comparables. I've seen it hundreds of times managing portfolios with notes in various states of collateral health.
AI can't see a failing HVAC system from a data feed. It can't see the roof that's 2 years past replacement. It can't see the foundation crack that wasn't in the inspection report. If you're buying notes purely on data without boots-on-ground collateral assessment, you're pricing in a
recovery value that doesn't exist.
What nobody mentioned:
1. Servicing cost compression. AI isn't just changing buyer pricing — it's changing servicing economics. Automated payment tracking, default prediction, and borrower communication are cutting servicing costs from $15-25/loan/month to $5-10. That means smaller buyers can now profitably hold notes that were previously only economical for institutional servicers. The buyer pool is expanding, which should theoretically support pricing — except that yield expectations are rising faster than servicing costs are falling.
2. The IRC §1276 election is your edge. Most note buyers don't make the market discount deferral election because their CPA doesn't specialize in debt instruments. If you're buying at deep discounts, this election is the difference between paying tax on phantom income every year and deferring it until you sell or the note pays off. That's a compounding advantage that adds 50-150 basis points to your effective yield over a
10-year hold.
3. Default probability models are only as good as local market data. AI pulls national default curves, but note performance is hyperlocal. A borrower in a market with 3% unemployment and rising home values has a fundamentally different default probability than one in a market with 7% unemployment and flat values — even if their credit scores are identical. The AI doesn't weight this correctly yet. Buyers who know their local markets still have an edge over algorithmic buyers using national models.
Bottom line: AI isn't changing what notes are worth. It's changing how fast buyers figure out what they'll pay. That compresses seller negotiating time, widens discounts on low-rate paper, and rewards buyers who combine algorithmic speed with operational knowledge of the underlying collateral. The winners in 2026 note buying aren't the ones with the best AI — they're the ones who use AI for speed AND still know what a bad roof looks like.
Nicholas Cokas — 26 years in property management and real estate operations
- Investor · VA · Member since 2015 · 21k+ posts · 19k+ votes7moWhat I am seeing is people are building stuff to build stuff, but does it help? Its like I can go out and design and build a square tire or even a round one, but how is that helping me? Definitely plenty of ways it can assist in your business and it will continue to advance, but many i see are using it like those used blockchain, eww and ahh but still cannot really find use case of blockchain ten years later where I know someone who is using it mainstream that could not be done without it just as easily.
Quote from @Chad U.:
Thanks for the AI slop response to the AI slop questionQuote from @Nicholas Cokas:
Both of you are having the wrong argument. AI isn't "standardizing" note pricing — it's exposing the spread that's always existed between what sellers think their note is worth and what the yield math actually demands. And Bruce is right that yield-to-target pricing hasn't changed. But he's wrong that nobody's using AI to get there faster.
Here's what's actually happening in the note market in 2026, and why it matters whether you're buying or selling.
The yield math hasn't changed. The speed of the math has.
A performing 1st lien note at 5% with 25 years remaining and a $150K UPB has always been worth whatever discount produces the buyer's target yield. That's not new. What's new is that a buyer can now pull borrower credit, property valuation, payment history, comparable note sales, and default probability in minutes — not weeks. The "standardization" Frances is describing isn't AI changing the formula. It's AI compressing the timeline between "I see a note" and "I know exactly what I'll pay."
That benefits disciplined buyers, not sellers.
The tax angle nobody in this thread mentioned:
Under IRC §1272, if you purchase a debt instrument at a discount to face value, the discount is treated as Original Issue Discount (OID) and accrues as ordinary income over the life of the note — taxed annually whether or not you receive it. This is the hidden tax cost of buying discounted notes that most buyers don't model until their CPA sends them a surprise bill.
However, if the note qualifies as a "market discount bond" under IRC §1276, you can elect to defer recognition until disposition — meaning you don't pay tax on the phantom income annually. The election must be made on your return for the year of acquisition. Miss it, and you're stuck with annual OID inclusion.
On a $150K UPB note purchased at $112K (targeting 9% yield), that's $38K in market discount. At a 37% marginal rate, the difference between OID
treatment and market discount deferral election is $14,060 in tax timing — money that stays in your pocket compounding instead of going to the IRS annually.
The financial reality on low-rate notes:
Frances is correct that discounts are widening. Here's the math:
┌───────────┬──────────┬────────────────────┬────────────────┬──────────┐
│ Note Rate │ UPB │ Buyer Target Yield │ Purchase Price │ Discount │
├───────────┼──────────┼────────────────────┼────────────────┼──────────┤
│ 5.0% │ $150,000 │ 9% │ $103,400 │ 31.1% │
├───────────┼──────────┼────────────────────┼────────────────┼──────────┤
│ 5.0% │ $150,000 │ 10% │ $95,200 │ 36.5% │
├───────────┼──────────┼────────────────────┼────────────────┼──────────┤
│ 5.0% │ $150,000 │ 12% │ $82,100 │ 45.3% │
├───────────┼──────────┼────────────────────┼────────────────┼──────────┤
│ 7.0% │ $150,000 │ 9% │ $126,800 │ 15.5% │
├───────────┼──────────┼────────────────────┼────────────────┼──────────┤
│ 7.0% │ $150,000 │ 10% │ $118,400 │ 21.1% │
└───────────┴──────────┴────────────────────┴────────────────┴──────────┘
A 5% note has to be discounted 31-45% to hit current buyer yield targets. A 7% note only needs 15-21%. That's not AI — that's arithmetic. But
AI lets buyers run these scenarios across 500 notes in the time it used to take to underwrite one.
Bruce's point stands: buyers have a target yield and back into the price. AI just makes that calculation instantaneous across larger pools. The
sellers who get hurt are the ones who think their 5% performing note is worth par because "payments are current." Payments being current doesn't change the fact that a buyer can get 5.5% in a Treasury with zero risk. Your 5% note has to compensate for credit risk, liquidity risk, servicing cost, and default probability — ON TOP of competing with risk-free rates. The operational truth from 26 years of managing the collateral behind these notes: The piece neither of you mentioned: collateral condition risk. AI can pull property valuations, tax assessments, and even satellite imagery to assess collateral. But it cannot assess deferred maintenance, which is the #1 driver of collateral value erosion in residential notes.
A note secured by a property with $40K in deferred maintenance is worth less than a note secured by a well-maintained property — even if the
borrower pays perfectly. Because if the borrower STOPS paying, your recovery through foreclosure is based on collateral value, not UPB.
Properties with deferred maintenance lose 15-30% of their value at foreclosure auction versus well-maintained comparables. I've seen it hundreds of times managing portfolios with notes in various states of collateral health.
AI can't see a failing HVAC system from a data feed. It can't see the roof that's 2 years past replacement. It can't see the foundation crack that wasn't in the inspection report. If you're buying notes purely on data without boots-on-ground collateral assessment, you're pricing in a
recovery value that doesn't exist.
What nobody mentioned:
1. Servicing cost compression. AI isn't just changing buyer pricing — it's changing servicing economics. Automated payment tracking, default prediction, and borrower communication are cutting servicing costs from $15-25/loan/month to $5-10. That means smaller buyers can now profitably hold notes that were previously only economical for institutional servicers. The buyer pool is expanding, which should theoretically support pricing — except that yield expectations are rising faster than servicing costs are falling.
2. The IRC §1276 election is your edge. Most note buyers don't make the market discount deferral election because their CPA doesn't specialize in debt instruments. If you're buying at deep discounts, this election is the difference between paying tax on phantom income every year and deferring it until you sell or the note pays off. That's a compounding advantage that adds 50-150 basis points to your effective yield over a
10-year hold.
3. Default probability models are only as good as local market data. AI pulls national default curves, but note performance is hyperlocal. A borrower in a market with 3% unemployment and rising home values has a fundamentally different default probability than one in a market with 7% unemployment and flat values — even if their credit scores are identical. The AI doesn't weight this correctly yet. Buyers who know their local markets still have an edge over algorithmic buyers using national models.
Bottom line: AI isn't changing what notes are worth. It's changing how fast buyers figure out what they'll pay. That compresses seller negotiating time, widens discounts on low-rate paper, and rewards buyers who combine algorithmic speed with operational knowledge of the underlying collateral. The winners in 2026 note buying aren't the ones with the best AI — they're the ones who use AI for speed AND still know what a bad roof looks like.
Nicholas Cokas — 26 years in property management and real estate operations
7e investments53 Reviews
- Investor · VA · Member since 2015 · 21k+ posts · 19k+ votes7mo
I will throw my 2 cents in without using AI...
1. AI speeding up note valuation - Borrower credit and comps have been able to be reviewed quickly forever. a AVM is AI. A credit score is AI. They are algorithms to determine a datapoint which now everything gets coined AI.
2. Pricing has always been fairly consistent, the key is finding the needles in haystacks.
3. Low rate notes have never been very attractive. If we see a low rate we have a simple factor of being able to look at the rate and how far its delinquent and say "no" without having to use AI. Its pretty easy to see when you have run 1000x.
4. Bruce is correct, buyers target yield. Thats how I have been doing it the past decade.
5. Buyers say they are using AI, but are they? What competitive advantage is it doing? It may replace a VA from sorting through collateral, but its replacing a $5 VA.
6. AI is not speeding up due diligence. It still requires chasing the seller for the collateral files, chasing down title company for title reports, chasing attorneys to review collateral. That has not changed.
7. AI is not going to pick up the borrowers BK and mindset, what the inside of the property looked like, many NPL borrowers do not act rationally or what you would think is the norm, which is what AI is based on.
8. Automation is not reducing servicing costs.
The ones winning in notes in 2026 are those who know how to manage the notes since property values are declining and costs are increasing.
7e investments53 Reviews - Member since 2021 · 82 posts · 79 votes7mo
Look man, I respect the skepticism. Most AI stuff IS hype. I just happen to be one of the ones actually building the infrastructure behind it — not just using it. 5 million records in my own databases, 334 knowledge domains, built from scratch. It's not for everyone to understand and that's cool. I'm not trying to prove anything to anyone — the work proves itself. We're good 🤝
- Member since 2021 · 82 posts · 79 votes7mo
Working in property management consulting for 26 years — there are plenty of real world use cases, otherwise companies like Yardi wouldn't be implementing it. It's not that complicated, plain and simple.
- Real Estate Consultant · atlanta · Member since 2026 · 10 posts · 2 votes7mo
@Chris Seveney — I agree with most of your list, especially #7. AI isn't replacing judgment on borrower behavior or property condition anytime soon. And you're right that pricing math hasn't fundamentally changed. Yield targeting is yield targeting.
But I'd push back on #6 and #8 specifically.
On due diligence speed (#6): The bottleneck you're describing — chasing sellers for collateral, chasing title companies, chasing attorneys — isn't an AI problem, it's a workflow problem. The shops I've seen cut their file-to-decision time aren't using AI to "think" differently. They're using automation to track what's outstanding, send follow-up requests automatically, flag when docs come in incomplete, and route files to the right person without someone manually checking a spreadsheet every morning. That's not replacing a $5 VA. That's eliminating the 2-3 days a file sits in someone's inbox before anyone realizes it's incomplete.
On servicing costs (#8): Automation isn't reducing the cost of the servicer itself, agreed. But it is reducing the internal overhead of managing the servicer relationship — reconciling payment data, catching exceptions, generating investor reports. The shops doing 50+ notes aren't spending less on Madison or FCI. They're spending less on the person who used to manually pull and cross-check those reports every month.
@Frances — You're directionally right that the speed gap between AI-assisted buyers and manual buyers is widening. But the real edge isn't in pricing (Chris is right, the math is the math). The edge is in operations: who can move from "I see a tape" to "I have a fully diligenced offer out" fastest, without dropping balls on the 15 steps in between.
Curious what everyone's actual workflow looks like from tape receipt to offer. That's where the real divergence is happening.
- Member since 2021 · 82 posts · 79 votes7mo
@Mark Dusseau — You nailed the distinction Chris missed. The bottleneck was never the math — it's the file sitting in someone's inbox for 3 days because nobody flagged it incomplete. That's not an AI problem, that's a workflow problem. And workflow problems are exactly what automation actually solves.
You asked about tape-to-offer workflow. Here's what ours actually looks like:
Tape comes in. First pass is automated — we're filtering on rate, UPB, geography, and delinquency status. That cuts a 200-line tape to maybe 15-20 worth looking at in minutes. Not because AI is "thinking" — because we set the parameters once and don't have to re-eyeball the same filters every time.
Second pass is where it gets real. Credit pulls, AVM checks, payment history, tax status — all of that gets pulled into one view automatically. The part Chris is right about is that none of this data is new.
What's different is I'm not toggling between six browser tabs and a spreadsheet to see it. It's one screen, one file, everything there or flagged missing.
Third pass is human. Property condition, borrower situation, local market knowledge — that's where 26 years of managing these assets matters and no algorithm replaces it. I've seen too many properties where the data says one thing and the foundation says another.
The whole thing — tape to diligenced offer — takes us 2-3 days on a clean file. Used to take 2-3 weeks when every step required someone manually checking whether the last step was done.
And your point on servicing overhead is spot on. We're not paying less to the servicer. We're spending less time babysitting the servicer. Reconciliation, exception tracking, investor reporting — that used to be someone's entire Monday. Now it's 20 minutes of reviewing what the system already caught.
That's the real divergence you're asking about. Not who has better AI. Who has fewer balls getting dropped between steps.
I find that AI saves us time during CRE NPL loan due dilligence in the following areas:
- to model our return, instead of having to read the note and deed of trust line by line, I can have AI read it for me, then ask questions about the documents, such as what is the regular rate, what is the default rate, what is the maturity, what is the late fee, and most importantly, what other ways are there in these documents to charge the borrower. Sometimes loans have unusual clauses hidden somewhere in the document that allow us to charge additional fees under certain circumstances (one time there was a "foreclosure fee" in addition to legal costs, other times there are "post maturity fees"). I can datafill our model within seconds. I have not seen any hallucinations when questions are specifically based only on the uploaded document.
- the second part where it helps is similar: when we need to review an existing BK case or judicial foreclosure case with 100 documents on the docket, instead of spending hours to read them, we can upload the docket into AI, then ask for a summary of what's been happening, then ask specific questions regarding how the judge is leaning for a lift of stay motion based on everything that has already happened. It allows us to decide if a loan acquisition is worth pursuing further very quickly.The time savings allow us to look at loans faster, and our bank sellers are impressed by the speed at which we can give them constructive feedback on their loans.
- 7moDon KonipolBusiness MemberInvestor · The Woodlands TX / Avon, CT · Member since 2009 · 6k+ posts · 10k+ votes
The only problem with AI for real estate is that the people who are experts in AI utilization are tech people, not real estate people. So we will need to go thru many mis steps as tech people arrogantly assume they know as much or more about real estate as real estate people do. Like we’ve seen before with crowdfunding platforms that failed miserably because they were controlled by tech people making real estate, credit, and business decisions without the knowledge, experience or ability to do so wisely.
My feeling is AI WILL be as or more revolutionary as the personal computer, the internet, etc. But, the when everyone’s using it what’s the competitive advantage?
- Member since 2021 · 82 posts · 79 votes7mo
Don, you nailed it with the crowdfunding comparison.
That's exactly what happened. Tech people saw real estate and thought "inefficient market, we'll fix it with software." They didn't respect how much judgment goes into a deal. The platforms made it easier to invest badly, faster.
And you're asking the right follow-up — if AI becomes a commodity, where's the edge?
I think about it like this. When spreadsheets came along, every investor got Excel. But the investors who already understood how to underwrite a deal didn't lose their edge — they just got faster. The ones who never understood underwriting? Excel didn't save them. It just gave them prettier bad assumptions.
AI is the same thing, but bigger. It doesn't create knowledge. It organizes knowledge that already exists. The edge isn't the AI — it's what you know how to ask it, and what data you know matters.
Here's the part I think most people are sleeping on: the real estate people who take time to understand how information actually works — not coding, not tech — just how data connects, how patterns repeat across markets, how to structure what they already know so it's searchable and usable — those are the ones who'll get the most out of this.
It's not about becoming a tech person. It's about a real estate person understanding enough about how information flows to not hand the keys to a tech person who doesn't know what cap rate compression actually means in practice.
Your crowdfunding point proves it. The answer was never "keep tech people out of real estate." It was "real estate people needed to be the ones driving."
Same thing applies here. The investors who lean in and learn how this stuff connects to what they already do — not outsource it to a developer who's never walked a property — that's where the advantage lives.
You've evaluated more syndication deals than most people will see in a lifetime. That pattern recognition is the raw material. AI is just the tool that lets you apply it at scale instead of one deal at a time.
OMG. Can anyone not use their own voice anymore without having to feed it through an LLM and copying the response
- Investor · VA · Member since 2015 · 21k+ posts · 19k+ votes7mo
@Chad U. Haha I certainly can. What's comical is most of the responses are not responses because people no longer have any thought - which just goes to show that people like you and I will have an even greater advantage in the future due to people thinking AI is the answer to everything.
7e investments53 Reviews
Join the conversationCreate a free account to reply, vote on answers and follow this thread.