Using Claude Cowork to Underwrite a Deal in 10 mins.

Real Estate Agent · Bowling Green KY ~ Lexington, KY · Member since 2015 · 1k+ posts · 603 votes

How I Underwrote a 47-Unit Value-Add Deal in 10 minutes using Claude Cowork Agent — Full Breakdown

I recently ran a full deal analysis on a 47-unit multifamily property up in Kentucky, and I wanted to share the complete breakdown — the good, the bad, and the red flags — because I think it provides a decent example of how to evaluate a value-add deal. And using AI, it can really save you a lot of time on your deal analysis. I think back to just a year ago, if not less, and all this data entry would easily take a few hours, or days if your waiting on a Broker to call you back or answer an email regarding the deal.

If you don't have any Broker relationships to where they are sending you deals, you can go on Crexi, or LoopNet or any other site and search for what you are looking for, download the Offering Memorandum, T12 and Rent Roll (if available), and upload it to your AI Agent and start vetting the deal.

This is a Class C workforce housing property built in 1972. All utilities are landlord-paid. Current occupancy sits at 74.5%. The seller hasn't publicly disclosed an asking price. Here's how I approached it.

The Property at a Glance

Units: 47 (16 Studio/1BA · 20 One-Bed/1BA · 11 Two-Bed/1BA)
Year Built: 1972 · 2-story
Class: C — workforce housing, aging but functional
Utilities: ALL landlord-paid (gas, electric, water, sewer, trash)
Physical Occupancy: 35 of 47 units = 74.5%
Current Monthly Rent (occupied): $28,000/month
State: Kentucky

The Market

The property sits in a small but stable Kentucky market with a population of roughly 14,300 and steady 1.4% annual growth. The AI Agent broke down the crime rate (very low), and the lack of affordable housing in the area, and comparable properties for rent. It also researched the current vacancy rates.

This is an undersupplied workforce housing market, which creates a favorable backdrop for lease-up.

The below comes from the AI Agent reviewing the Offering Memorandum, T12 (Trailing 12 Months Income), and the Rent Roll.

The T12: What the Numbers Actually Say

Total T12 Income: $314,622

The income side is straightforward. Rental income came in at $296,722, with the remainder from late fees, EFT convenience fees, utility reimbursements, and misc income. Nothing alarming.

The expense side is where it gets interesting.

The as-stated NOI from the seller shows $161,691. When I normalized the expenses, here's what I found:

Biggest adjustment: Management fees. The T12 shows management fees in only 5 of 12 months — dropping to $0 from July through December 2025. That's a red flag. Either the property switched to self-management mid-year, or the owner stopped recording the fee. Either way, you have to normalize it. At 8.5% of EGI, that's $26,743/year — versus only $12,107 actually recorded. That single adjustment accounts for the majority of the normalization difference.

	As-Stated	Normalized
Total Income	$314,622	$314,622
Total Operating Expenses	($129,500)	($145,679)
Property Taxes	($23,431)	($23,431)
True NOI	$161,691	$147,055
Expense Ratio	48.7%	53.7%

Utilities are the other major exposure. The landlord pays ALL utilities — $57,597/year across 47 units. That's $1,225/unit/year. If you buy this and bring in higher-consumption tenants, your costs go up with no offsetting revenue. RUBS (Ratio Utility Billing System) implementation is the play here, and I'll get to that in the value-add section.

The Rent Roll: Where the Real Upside Lives

Current occupancy of 74.5% breaks down like this:

Unit Type	Total	Occupied	Vacant	Avg Rent
Studio/1BA	16	9	7	$691
1 Bed/1BA	20	18	2	$858
2 Bed/1BA	11	8	3	$791
Total	47	35	12	$800

The studio section is the story. 7 of 16 studios are vacant — 43.8% vacancy in that unit type alone. That's your primary management challenge and your primary upside lever.

Loss to Lease (Below-Market Occupied Units)

Eight occupied units are paying significantly below market:

Unit 8: $465/month vs. $750 market (Studio) — 38% below market
Unit 34: $650/month vs. $905 market (2BR)
Unit 35: $600/month vs. $905 market (2BR)
Unit 43: $635/month vs. $905 market (2BR)
Plus four more units ranging 10–18% below market

Total annualized loss to lease from occupied below-market units: $23,040/year — recoverable as leases expire or units turn.

The AI Agent caught the below also. I new this from the build year from having owned older properties, but it was nice to see the AI Agent caught it without me having to suggest it.

CapEx: Don't Skip This on a 1972 Build

A 53-year-old building requires full capital planning. Pre-1975 reserve tables call for $1,800–$2,500/unit/year. Before earnest money goes hard, I'd want inspections on all of the following:

Critical (must know before going non-refundable):

Plumbing system — galvanized pipe is common in 1960s–70s Kentucky construction. Full replacement can run $50,000–$100,000+. This is the single largest unknown.
Roof — unknown condition, $15,000–$30,000 to replace (most likely more)
Electrical panels — likely updated but verify

High priority:

HVAC per unit: $4,000–$6,000/unit if replacement needed (would need to confirm if Central Air, or window units)
7 vacant studios need cosmetic turns: ~$2,000–$5,000 each

Year 1 CapEx budget I used: $100,000 funded at closing.

The studio turn math is compelling on its own: turn all 7 at $3,500 average = $24,500 investment to unlock $63,000/year in new gross revenue. That's a 257% first-year ROI on that capital alone.

The Value-Add Stack

Here's the full upside picture if you execute all levers:

Lever	Annual NOI Impact
Fill 7 vacant studios @ $750/mo	+$63,000
Fill 2 vacant 1BRs @ $880/mo	+$21,120
Fill 3 vacant 2BRs @ $905/mo	+$32,580
Bring 8 below-market units to market	+$21,912
RUBS utility billing @ $75/unit/mo	+$42,300
Total Potential Upside	+$180,912/yr

Not every lever executes simultaneously. My conservative stabilized NOI (90% occupancy, market rents, no RUBS) is $221,857/year. With RUBS, it reaches $264,157/year.

At a 9% cap rate, stabilized value = ~$2,465,000. Against a target purchase price of $1,475,000, that's nearly $1,000,000 in value creation.

The AI Agent also caught the below which some investors may not have thought about.

The RUBS Problem Nobody Talks About

RUBS looks great on paper, I see multiple deals per month where the Broker is using that as a major value add. I’m not saying it doesn’t work, but take into consideration the following which the Agent brought up: Recover $42,000–$47,000/year in utility costs from tenants, slash your largest expense category, add $470,000+ in stabilized value at a 9% cap. It's one of the most cited value-add levers in multifamily underwriting — and it deserves real scrutiny before you run it as a given.

Here's the issue with this specific deal: the current market rents for this property are $750/month for studios, $880 for one-beds, and $905 for two-beds. Those are the rents I've been using as the stabilized target. But those market rents are almost certainly already priced to include utilities — because that's how this market rents. Every comparable in a Class C workforce housing market like this factors in what tenants actually pay all-in.

When you implement RUBS and start billing tenants $60–$90/month in utility charges on top of their rent, you haven't just added income. You've effectively raised their all-in monthly cost to $810–$840 for a studio, $940–$970 for a one-bed, $965–$995 for a two-bed. At that point, you're no longer competitive with the market. You're asking tenants to pay above what similar units in the area cost — and in a Class C workforce housing property, you don't have the amenities or finishes to justify a premium.

The real risk: RUBS-driven vacancy. If tenants balk at the new all-in cost and start leaving — or if you can't fill vacancies because your effective rent is above market — you've traded a utility expense reduction for an occupancy problem. And an occupancy problem at a 47-unit property with 1972-era bones and already 74.5% physical occupancy is a much bigger issue than $57,000 in annual utility costs.

How to think about it correctly:

There are really two ways RUBS works in your favor here, and they aren't the same thing:

If you implement RUBS AND lower base rents to compensate — tenants pay the same all-in, you just shift the cost structure. Your utility expense drops but your rental income also drops. The NOI improvement is real but smaller than the raw RUBS number suggests.
If comparable properties in the market have already shifted to tenant-paid utilities — and yours is one of the last to do it — then adding RUBS actually doesn't push you above market. It brings you in line with it. In that case, the full $42,000–$47,000 recovery is legitimate upside.

The due diligence question that changes everything: Are competing properties in this market charging utilities separately, or are they all-inclusive? If the comps are all-inclusive, RUBS is a retention risk. If the comps have already shifted to tenant-paid, RUBS is a free $470,000 in value creation.

I underwrote this deal conservatively — stabilized NOI of $221,857 with no RUBS. The RUBS upside is real, but I'm treating it as a bonus if the market supports it, not a lever I'm counting on to make the deal work.

Offer Strategy

Because the property is at 74.5% occupancy, conventional/agency financing isn't available yet. The right vehicle is a bridge loan (I modeled 65% LTV at 7.0% IO), with a plan to refinance to conventional debt after stabilization.

	Opening	Target	Walk-Away
Price	$1,350,000	$1,475,000	$1,600,000
Price/Unit	$28,723	$31,383	$34,043
Cap Rate (Norm NOI)	10.89%	9.97%	9.19%
Cash-on-Cash (Yr 1)	12.9%	11.3%	9.8%
DSCR	2.39×	2.19×	2.02×
Total Equity Required	$664,500	$710,750	$757,000

5-Year Projection (at Target Offer):

Year 3 refinance returns ~$597,000 of original $710,750 equity invested
5-year total distributions + exit: ~$1,886,000
Equity multiple: 2.65×
Estimated IRR: ~29–31%

The Verdict

This is a compelling value-add deal — but not without real risks.

Green lights: Strong market fundamentals, excellent safety profile, below-market entry price, multiple independent upside levers, strong DSCR even at current occupancy.

Red flags to resolve in due diligence:

Why did management fees go to $0 in H2 2025? Is this self-managed and operationally stressed?
Plumbing — is it galvanized? That answer alone could change the deal.
Studio vacancy — is it condition-related, pricing-related, or demand-related?

The deal works at $1,475,000 if the inspection comes back clean and the studio vacancy is a management problem (not a demand problem). If the plumbing is galvanized, you're looking at a $50,000–$100,000 budget revision and a renegotiated price.

Some of you may already be using AI to assist you in your deal analysis, if not though, I would encourage you to give it a try. Also, I mainly use Claude Code, its seems to keep getting upgraded each month to where it is smarter and faster than the month before. As a side note, AI isn’t perfect (at least not right now), but it’s helped me save a lot of time vetting deals that on their face value look promising. One thing to note about this deal, is the suggested purchase price seems very low in my view, but the AI is looking at it from a black and white view, they are not thinking "thats really low per door", the AI is just going off the facts of the numbers and giving a suggestion based upon those facts, no emotions involved whatsoever.

What would you want to know before pulling the trigger on a deal like this? Drop your questions below.

@Carl Mcknight One thing that stood out to me was the relatively low occupancy rate despite the landlord covering utilities. You mentioned this as well, but properties with utilities included usually see stronger occupancy, even with smaller unit sizes.

It might be worth looking into whether the studios could easily be converted into one bedroom units. I’d also check whether the rents are priced too high for the market or if the units may need some updating. Those factors could be contributing to the lower occupancy but with workforce housing if clean, functional, and priced right there shouldn't be any problems with keeping units filled.

REI James w/ eXp Realty54 Reviews

Website View Page

1 Reply

Carl Mcknight

OP

Rental Property Investor · TN · Member since 2013 · 54 posts · 50 votes

7mo

Quote from @James Wilcox:

Quote from @Carl Mcknight:

How I Underwrote a 47-Unit Value-Add Deal in 10 minutes using Claude Cowork Agent — Full Breakdown

I recently ran a full deal analysis on a 47-unit multifamily property up in Kentucky, and I wanted to share the complete breakdown — the good, the bad, and the red flags — because I think it provides a decent example of how to evaluate a value-add deal. And using AI, it can really save you a lot of time on your deal analysis. I think back to just a year ago, if not less, and all this data entry would easily take a few hours, or days if your waiting on a Broker to call you back or answer an email regarding the deal.

If you don't have any Broker relationships to where they are sending you deals, you can go on Crexi, or LoopNet or any other site and search for what you are looking for, download the Offering Memorandum, T12 and Rent Roll (if available), and upload it to your AI Agent and start vetting the deal.

This is a Class C workforce housing property built in 1972. All utilities are landlord-paid. Current occupancy sits at 74.5%. The seller hasn't publicly disclosed an asking price. Here's how I approached it.

The Property at a Glance

Units: 47 (16 Studio/1BA · 20 One-Bed/1BA · 11 Two-Bed/1BA)
Year Built: 1972 · 2-story
Class: C — workforce housing, aging but functional
Utilities: ALL landlord-paid (gas, electric, water, sewer, trash)
Physical Occupancy: 35 of 47 units = 74.5%
Current Monthly Rent (occupied): $28,000/month
State: Kentucky

The Market

The property sits in a small but stable Kentucky market with a population of roughly 14,300 and steady 1.4% annual growth. The AI Agent broke down the crime rate (very low), and the lack of affordable housing in the area, and comparable properties for rent. It also researched the current vacancy rates.

This is an undersupplied workforce housing market, which creates a favorable backdrop for lease-up.

The below comes from the AI Agent reviewing the Offering Memorandum, T12 (Trailing 12 Months Income), and the Rent Roll.

The T12: What the Numbers Actually Say

Total T12 Income: $314,622

The income side is straightforward. Rental income came in at $296,722, with the remainder from late fees, EFT convenience fees, utility reimbursements, and misc income. Nothing alarming.

The expense side is where it gets interesting.

The as-stated NOI from the seller shows $161,691. When I normalized the expenses, here's what I found:

Biggest adjustment: Management fees. The T12 shows management fees in only 5 of 12 months — dropping to $0 from July through December 2025. That's a red flag. Either the property switched to self-management mid-year, or the owner stopped recording the fee. Either way, you have to normalize it. At 8.5% of EGI, that's $26,743/year — versus only $12,107 actually recorded. That single adjustment accounts for the majority of the normalization difference.

	As-Stated	Normalized
Total Income	$314,622	$314,622
Total Operating Expenses	($129,500)	($145,679)
Property Taxes	($23,431)	($23,431)
True NOI	$161,691	$147,055
Expense Ratio	48.7%	53.7%

Utilities are the other major exposure. The landlord pays ALL utilities — $57,597/year across 47 units. That's $1,225/unit/year. If you buy this and bring in higher-consumption tenants, your costs go up with no offsetting revenue. RUBS (Ratio Utility Billing System) implementation is the play here, and I'll get to that in the value-add section.

The Rent Roll: Where the Real Upside Lives

Current occupancy of 74.5% breaks down like this:

Unit Type	Total	Occupied	Vacant	Avg Rent
Studio/1BA	16	9	7	$691
1 Bed/1BA	20	18	2	$858
2 Bed/1BA	11	8	3	$791
Total	47	35	12	$800

The studio section is the story. 7 of 16 studios are vacant — 43.8% vacancy in that unit type alone. That's your primary management challenge and your primary upside lever.

Loss to Lease (Below-Market Occupied Units)

Eight occupied units are paying significantly below market:

Unit 8: $465/month vs. $750 market (Studio) — 38% below market
Unit 34: $650/month vs. $905 market (2BR)
Unit 35: $600/month vs. $905 market (2BR)
Unit 43: $635/month vs. $905 market (2BR)
Plus four more units ranging 10–18% below market

Total annualized loss to lease from occupied below-market units: $23,040/year — recoverable as leases expire or units turn.

The AI Agent caught the below also. I new this from the build year from having owned older properties, but it was nice to see the AI Agent caught it without me having to suggest it.

CapEx: Don't Skip This on a 1972 Build

A 53-year-old building requires full capital planning. Pre-1975 reserve tables call for $1,800–$2,500/unit/year. Before earnest money goes hard, I'd want inspections on all of the following:

Critical (must know before going non-refundable):

Plumbing system — galvanized pipe is common in 1960s–70s Kentucky construction. Full replacement can run $50,000–$100,000+. This is the single largest unknown.
Roof — unknown condition, $15,000–$30,000 to replace (most likely more)
Electrical panels — likely updated but verify

High priority:

HVAC per unit: $4,000–$6,000/unit if replacement needed (would need to confirm if Central Air, or window units)
7 vacant studios need cosmetic turns: ~$2,000–$5,000 each

Year 1 CapEx budget I used: $100,000 funded at closing.

The studio turn math is compelling on its own: turn all 7 at $3,500 average = $24,500 investment to unlock $63,000/year in new gross revenue. That's a 257% first-year ROI on that capital alone.

The Value-Add Stack

Here's the full upside picture if you execute all levers:

Lever	Annual NOI Impact
Fill 7 vacant studios @ $750/mo	+$63,000
Fill 2 vacant 1BRs @ $880/mo	+$21,120
Fill 3 vacant 2BRs @ $905/mo	+$32,580
Bring 8 below-market units to market	+$21,912
RUBS utility billing @ $75/unit/mo	+$42,300
Total Potential Upside	+$180,912/yr

Not every lever executes simultaneously. My conservative stabilized NOI (90% occupancy, market rents, no RUBS) is $221,857/year. With RUBS, it reaches $264,157/year.

At a 9% cap rate, stabilized value = ~$2,465,000. Against a target purchase price of $1,475,000, that's nearly $1,000,000 in value creation.

The AI Agent also caught the below which some investors may not have thought about.

The RUBS Problem Nobody Talks About

RUBS looks great on paper, I see multiple deals per month where the Broker is using that as a major value add. I’m not saying it doesn’t work, but take into consideration the following which the Agent brought up: Recover $42,000–$47,000/year in utility costs from tenants, slash your largest expense category, add $470,000+ in stabilized value at a 9% cap. It's one of the most cited value-add levers in multifamily underwriting — and it deserves real scrutiny before you run it as a given.

Here's the issue with this specific deal: the current market rents for this property are $750/month for studios, $880 for one-beds, and $905 for two-beds. Those are the rents I've been using as the stabilized target. But those market rents are almost certainly already priced to include utilities — because that's how this market rents. Every comparable in a Class C workforce housing market like this factors in what tenants actually pay all-in.

When you implement RUBS and start billing tenants $60–$90/month in utility charges on top of their rent, you haven't just added income. You've effectively raised their all-in monthly cost to $810–$840 for a studio, $940–$970 for a one-bed, $965–$995 for a two-bed. At that point, you're no longer competitive with the market. You're asking tenants to pay above what similar units in the area cost — and in a Class C workforce housing property, you don't have the amenities or finishes to justify a premium.

The real risk: RUBS-driven vacancy. If tenants balk at the new all-in cost and start leaving — or if you can't fill vacancies because your effective rent is above market — you've traded a utility expense reduction for an occupancy problem. And an occupancy problem at a 47-unit property with 1972-era bones and already 74.5% physical occupancy is a much bigger issue than $57,000 in annual utility costs.

How to think about it correctly:

There are really two ways RUBS works in your favor here, and they aren't the same thing:

If you implement RUBS AND lower base rents to compensate — tenants pay the same all-in, you just shift the cost structure. Your utility expense drops but your rental income also drops. The NOI improvement is real but smaller than the raw RUBS number suggests.
If comparable properties in the market have already shifted to tenant-paid utilities — and yours is one of the last to do it — then adding RUBS actually doesn't push you above market. It brings you in line with it. In that case, the full $42,000–$47,000 recovery is legitimate upside.

The due diligence question that changes everything: Are competing properties in this market charging utilities separately, or are they all-inclusive? If the comps are all-inclusive, RUBS is a retention risk. If the comps have already shifted to tenant-paid, RUBS is a free $470,000 in value creation.

I underwrote this deal conservatively — stabilized NOI of $221,857 with no RUBS. The RUBS upside is real, but I'm treating it as a bonus if the market supports it, not a lever I'm counting on to make the deal work.

Offer Strategy

Because the property is at 74.5% occupancy, conventional/agency financing isn't available yet. The right vehicle is a bridge loan (I modeled 65% LTV at 7.0% IO), with a plan to refinance to conventional debt after stabilization.

	Opening	Target	Walk-Away
Price	$1,350,000	$1,475,000	$1,600,000
Price/Unit	$28,723	$31,383	$34,043
Cap Rate (Norm NOI)	10.89%	9.97%	9.19%
Cash-on-Cash (Yr 1)	12.9%	11.3%	9.8%
DSCR	2.39×	2.19×	2.02×
Total Equity Required	$664,500	$710,750	$757,000

5-Year Projection (at Target Offer):

Year 3 refinance returns ~$597,000 of original $710,750 equity invested
5-year total distributions + exit: ~$1,886,000
Equity multiple: 2.65×
Estimated IRR: ~29–31%

The Verdict

This is a compelling value-add deal — but not without real risks.

Green lights: Strong market fundamentals, excellent safety profile, below-market entry price, multiple independent upside levers, strong DSCR even at current occupancy.

Red flags to resolve in due diligence:

Why did management fees go to $0 in H2 2025? Is this self-managed and operationally stressed?
Plumbing — is it galvanized? That answer alone could change the deal.
Studio vacancy — is it condition-related, pricing-related, or demand-related?

The deal works at $1,475,000 if the inspection comes back clean and the studio vacancy is a management problem (not a demand problem). If the plumbing is galvanized, you're looking at a $50,000–$100,000 budget revision and a renegotiated price.

Some of you may already be using AI to assist you in your deal analysis, if not though, I would encourage you to give it a try. Also, I mainly use Claude Code, its seems to keep getting upgraded each month to where it is smarter and faster than the month before. As a side note, AI isn’t perfect (at least not right now), but it’s helped me save a lot of time vetting deals that on their face value look promising. One thing to note about this deal, is the suggested purchase price seems very low in my view, but the AI is looking at it from a black and white view, they are not thinking "thats really low per door", the AI is just going off the facts of the numbers and giving a suggestion based upon those facts, no emotions involved whatsoever.

What would you want to know before pulling the trigger on a deal like this? Drop your questions below.

@Carl Mcknight One thing that stood out to me was the relatively low occupancy rate despite the landlord covering utilities. You mentioned this as well, but properties with utilities included usually see stronger occupancy, even with smaller unit sizes.

It might be worth looking into whether the studios could easily be converted into one bedroom units. I’d also check whether the rents are priced too high for the market or if the units may need some updating. Those factors could be contributing to the lower occupancy but with workforce housing if clean, functional, and priced right there shouldn't be any problems with keeping units filled.

Turning the units into one beds may be something I look into. One thing I noticed on the T12 was the lack of a property management expense from July-Dec. I checked in on it, and it looks like it went back to self management, that may have something to do with it, maybe the owner is having his kid manage it who knows nothing about leasing up units, but unkown right now. Thanks for your inquiry James.

0 Reply

James WilcoxBusiness Member

Real Estate Agent · Bowling Green KY ~ Lexington, KY · Member since 2015 · 1k+ posts · 603 votes

7mo

Quote from @Carl Mcknight:

Quote from @James Wilcox:

Quote from @Carl Mcknight:

How I Underwrote a 47-Unit Value-Add Deal in 10 minutes using Claude Cowork Agent — Full Breakdown

I recently ran a full deal analysis on a 47-unit multifamily property up in Kentucky, and I wanted to share the complete breakdown — the good, the bad, and the red flags — because I think it provides a decent example of how to evaluate a value-add deal. And using AI, it can really save you a lot of time on your deal analysis. I think back to just a year ago, if not less, and all this data entry would easily take a few hours, or days if your waiting on a Broker to call you back or answer an email regarding the deal.

If you don't have any Broker relationships to where they are sending you deals, you can go on Crexi, or LoopNet or any other site and search for what you are looking for, download the Offering Memorandum, T12 and Rent Roll (if available), and upload it to your AI Agent and start vetting the deal.

This is a Class C workforce housing property built in 1972. All utilities are landlord-paid. Current occupancy sits at 74.5%. The seller hasn't publicly disclosed an asking price. Here's how I approached it.

The Property at a Glance

Units: 47 (16 Studio/1BA · 20 One-Bed/1BA · 11 Two-Bed/1BA)
Year Built: 1972 · 2-story
Class: C — workforce housing, aging but functional
Utilities: ALL landlord-paid (gas, electric, water, sewer, trash)
Physical Occupancy: 35 of 47 units = 74.5%
Current Monthly Rent (occupied): $28,000/month
State: Kentucky

The Market

The property sits in a small but stable Kentucky market with a population of roughly 14,300 and steady 1.4% annual growth. The AI Agent broke down the crime rate (very low), and the lack of affordable housing in the area, and comparable properties for rent. It also researched the current vacancy rates.

This is an undersupplied workforce housing market, which creates a favorable backdrop for lease-up.

The below comes from the AI Agent reviewing the Offering Memorandum, T12 (Trailing 12 Months Income), and the Rent Roll.

The T12: What the Numbers Actually Say

Total T12 Income: $314,622

The income side is straightforward. Rental income came in at $296,722, with the remainder from late fees, EFT convenience fees, utility reimbursements, and misc income. Nothing alarming.

The expense side is where it gets interesting.

The as-stated NOI from the seller shows $161,691. When I normalized the expenses, here's what I found:

Biggest adjustment: Management fees. The T12 shows management fees in only 5 of 12 months — dropping to $0 from July through December 2025. That's a red flag. Either the property switched to self-management mid-year, or the owner stopped recording the fee. Either way, you have to normalize it. At 8.5% of EGI, that's $26,743/year — versus only $12,107 actually recorded. That single adjustment accounts for the majority of the normalization difference.

	As-Stated	Normalized
Total Income	$314,622	$314,622
Total Operating Expenses	($129,500)	($145,679)
Property Taxes	($23,431)	($23,431)
True NOI	$161,691	$147,055
Expense Ratio	48.7%	53.7%

Utilities are the other major exposure. The landlord pays ALL utilities — $57,597/year across 47 units. That's $1,225/unit/year. If you buy this and bring in higher-consumption tenants, your costs go up with no offsetting revenue. RUBS (Ratio Utility Billing System) implementation is the play here, and I'll get to that in the value-add section.

The Rent Roll: Where the Real Upside Lives

Current occupancy of 74.5% breaks down like this:

Unit Type	Total	Occupied	Vacant	Avg Rent
Studio/1BA	16	9	7	$691
1 Bed/1BA	20	18	2	$858
2 Bed/1BA	11	8	3	$791
Total	47	35	12	$800

The studio section is the story. 7 of 16 studios are vacant — 43.8% vacancy in that unit type alone. That's your primary management challenge and your primary upside lever.

Loss to Lease (Below-Market Occupied Units)

Eight occupied units are paying significantly below market:

Unit 8: $465/month vs. $750 market (Studio) — 38% below market
Unit 34: $650/month vs. $905 market (2BR)
Unit 35: $600/month vs. $905 market (2BR)
Unit 43: $635/month vs. $905 market (2BR)
Plus four more units ranging 10–18% below market

Total annualized loss to lease from occupied below-market units: $23,040/year — recoverable as leases expire or units turn.

The AI Agent caught the below also. I new this from the build year from having owned older properties, but it was nice to see the AI Agent caught it without me having to suggest it.

CapEx: Don't Skip This on a 1972 Build

A 53-year-old building requires full capital planning. Pre-1975 reserve tables call for $1,800–$2,500/unit/year. Before earnest money goes hard, I'd want inspections on all of the following:

Critical (must know before going non-refundable):

Plumbing system — galvanized pipe is common in 1960s–70s Kentucky construction. Full replacement can run $50,000–$100,000+. This is the single largest unknown.
Roof — unknown condition, $15,000–$30,000 to replace (most likely more)
Electrical panels — likely updated but verify

High priority:

HVAC per unit: $4,000–$6,000/unit if replacement needed (would need to confirm if Central Air, or window units)
7 vacant studios need cosmetic turns: ~$2,000–$5,000 each

Year 1 CapEx budget I used: $100,000 funded at closing.

The studio turn math is compelling on its own: turn all 7 at $3,500 average = $24,500 investment to unlock $63,000/year in new gross revenue. That's a 257% first-year ROI on that capital alone.

The Value-Add Stack

Here's the full upside picture if you execute all levers:

Lever	Annual NOI Impact
Fill 7 vacant studios @ $750/mo	+$63,000
Fill 2 vacant 1BRs @ $880/mo	+$21,120
Fill 3 vacant 2BRs @ $905/mo	+$32,580
Bring 8 below-market units to market	+$21,912
RUBS utility billing @ $75/unit/mo	+$42,300
Total Potential Upside	+$180,912/yr

Not every lever executes simultaneously. My conservative stabilized NOI (90% occupancy, market rents, no RUBS) is $221,857/year. With RUBS, it reaches $264,157/year.

At a 9% cap rate, stabilized value = ~$2,465,000. Against a target purchase price of $1,475,000, that's nearly $1,000,000 in value creation.

The AI Agent also caught the below which some investors may not have thought about.

The RUBS Problem Nobody Talks About

RUBS looks great on paper, I see multiple deals per month where the Broker is using that as a major value add. I’m not saying it doesn’t work, but take into consideration the following which the Agent brought up: Recover $42,000–$47,000/year in utility costs from tenants, slash your largest expense category, add $470,000+ in stabilized value at a 9% cap. It's one of the most cited value-add levers in multifamily underwriting — and it deserves real scrutiny before you run it as a given.

Here's the issue with this specific deal: the current market rents for this property are $750/month for studios, $880 for one-beds, and $905 for two-beds. Those are the rents I've been using as the stabilized target. But those market rents are almost certainly already priced to include utilities — because that's how this market rents. Every comparable in a Class C workforce housing market like this factors in what tenants actually pay all-in.

When you implement RUBS and start billing tenants $60–$90/month in utility charges on top of their rent, you haven't just added income. You've effectively raised their all-in monthly cost to $810–$840 for a studio, $940–$970 for a one-bed, $965–$995 for a two-bed. At that point, you're no longer competitive with the market. You're asking tenants to pay above what similar units in the area cost — and in a Class C workforce housing property, you don't have the amenities or finishes to justify a premium.

The real risk: RUBS-driven vacancy. If tenants balk at the new all-in cost and start leaving — or if you can't fill vacancies because your effective rent is above market — you've traded a utility expense reduction for an occupancy problem. And an occupancy problem at a 47-unit property with 1972-era bones and already 74.5% physical occupancy is a much bigger issue than $57,000 in annual utility costs.

How to think about it correctly:

There are really two ways RUBS works in your favor here, and they aren't the same thing:

If you implement RUBS AND lower base rents to compensate — tenants pay the same all-in, you just shift the cost structure. Your utility expense drops but your rental income also drops. The NOI improvement is real but smaller than the raw RUBS number suggests.
If comparable properties in the market have already shifted to tenant-paid utilities — and yours is one of the last to do it — then adding RUBS actually doesn't push you above market. It brings you in line with it. In that case, the full $42,000–$47,000 recovery is legitimate upside.

The due diligence question that changes everything: Are competing properties in this market charging utilities separately, or are they all-inclusive? If the comps are all-inclusive, RUBS is a retention risk. If the comps have already shifted to tenant-paid, RUBS is a free $470,000 in value creation.

I underwrote this deal conservatively — stabilized NOI of $221,857 with no RUBS. The RUBS upside is real, but I'm treating it as a bonus if the market supports it, not a lever I'm counting on to make the deal work.

Offer Strategy

Because the property is at 74.5% occupancy, conventional/agency financing isn't available yet. The right vehicle is a bridge loan (I modeled 65% LTV at 7.0% IO), with a plan to refinance to conventional debt after stabilization.

	Opening	Target	Walk-Away
Price	$1,350,000	$1,475,000	$1,600,000
Price/Unit	$28,723	$31,383	$34,043
Cap Rate (Norm NOI)	10.89%	9.97%	9.19%
Cash-on-Cash (Yr 1)	12.9%	11.3%	9.8%
DSCR	2.39×	2.19×	2.02×
Total Equity Required	$664,500	$710,750	$757,000

5-Year Projection (at Target Offer):

Year 3 refinance returns ~$597,000 of original $710,750 equity invested
5-year total distributions + exit: ~$1,886,000
Equity multiple: 2.65×
Estimated IRR: ~29–31%

The Verdict

This is a compelling value-add deal — but not without real risks.

Green lights: Strong market fundamentals, excellent safety profile, below-market entry price, multiple independent upside levers, strong DSCR even at current occupancy.

Red flags to resolve in due diligence:

Why did management fees go to $0 in H2 2025? Is this self-managed and operationally stressed?
Plumbing — is it galvanized? That answer alone could change the deal.
Studio vacancy — is it condition-related, pricing-related, or demand-related?

The deal works at $1,475,000 if the inspection comes back clean and the studio vacancy is a management problem (not a demand problem). If the plumbing is galvanized, you're looking at a $50,000–$100,000 budget revision and a renegotiated price.

Some of you may already be using AI to assist you in your deal analysis, if not though, I would encourage you to give it a try. Also, I mainly use Claude Code, its seems to keep getting upgraded each month to where it is smarter and faster than the month before. As a side note, AI isn’t perfect (at least not right now), but it’s helped me save a lot of time vetting deals that on their face value look promising. One thing to note about this deal, is the suggested purchase price seems very low in my view, but the AI is looking at it from a black and white view, they are not thinking "thats really low per door", the AI is just going off the facts of the numbers and giving a suggestion based upon those facts, no emotions involved whatsoever.

What would you want to know before pulling the trigger on a deal like this? Drop your questions below.

@Carl Mcknight One thing that stood out to me was the relatively low occupancy rate despite the landlord covering utilities. You mentioned this as well, but properties with utilities included usually see stronger occupancy, even with smaller unit sizes.

It might be worth looking into whether the studios could easily be converted into one bedroom units. I’d also check whether the rents are priced too high for the market or if the units may need some updating. Those factors could be contributing to the lower occupancy but with workforce housing if clean, functional, and priced right there shouldn't be any problems with keeping units filled.

Turning the units into one beds may be something I look into. One thing I noticed on the T12 was the lack of a property management expense from July-Dec. I checked in on it, and it looks like it went back to self management, that may have something to do with it, maybe the owner is having his kid manage it who knows nothing about leasing up units, but unkown right now. Thanks for your inquiry James.

Tough to say but sure you will figure it out by digging deep enough and asking questions. Something weird going on there to me though. If nothing else just hire a Realtor to lease them for you.

REI James w/ eXp Realty54 Reviews

Website View Page

0 Reply

Michael Eskenasy

Investor · Pacific Northwest · Member since 2026 · 538 posts · 314 votes

7mo

@Carl Mcknight -

Alex — this is a really interesting breakdown and a great example of how AI can speed up the first pass on a deal. Being able to upload an OM, T12, and rent roll and get structured analysis back is a huge time saver.

One thing I spend a lot of time thinking about with AI systems in general is how sensitive they are to inputs and context. Two runs of the same analysis can produce different conclusions depending on things like:

• how the prompt is structured

• what assumptions the model is inferring vs. what is explicitly defined

• what context or documents are included in the session

• model/version changes over time

That doesn’t make the analysis wrong — it just means the process needs to be repeatable and explainable so you can trace exactly how the output was produced.

AI is incredibly useful for surfacing patterns and speeding up analysis, but the real value comes when the underlying assumptions and variables are clearly controlled.

Curious how structured your workflow is when you run these — are you using a fixed prompt/template for deals or more of a conversational approach?

1 Reply

Nate Marshall
Real Estate Consultant · Evergreen, CO · Member since 2018 · 1k+ posts · 740 votes
7mo
Quote from @Michael Eskenasy:

@Carl Mcknight -

Alex — this is a really interesting breakdown and a great example of how AI can speed up the first pass on a deal. Being able to upload an OM, T12, and rent roll and get structured analysis back is a huge time saver.

One thing I spend a lot of time thinking about with AI systems in general is how sensitive they are to inputs and context. Two runs of the same analysis can produce different conclusions depending on things like:

• how the prompt is structured

• what assumptions the model is inferring vs. what is explicitly defined

• what context or documents are included in the session

• model/version changes over time

That doesn’t make the analysis wrong — it just means the process needs to be repeatable and explainable so you can trace exactly how the output was produced.

AI is incredibly useful for surfacing patterns and speeding up analysis, but the real value comes when the underlying assumptions and variables are clearly controlled.

Curious how structured your workflow is when you run these — are you using a fixed prompt/template for deals or more of a conversational approach?

That happens when people don't have a system prebuilt that manages all of this. RCIC uses GPT and has a lnumber of PDF based instructions plus the GPT instructions itself behind the scenes so to speak. Users are seeing numbers within 3% of manual underwriting. People trying to just input T12's, OM's and Rent Rolls etc without anything behind will still see garbage in garbage out results.
1 Reply
Michael Eskenasy
Investor · Pacific Northwest · Member since 2026 · 538 posts · 314 votes
7mo
@Nate Marshall - having structure around the inputs definitely helps. The reason systems become necessary though isn’t just document organization, it’s the nature of the models themselves. AI doesn’t have durable memory and it doesn’t reason deterministically, so the same analysis can follow slightly different paths depending on context, session state, or how earlier information is weighted in the window. That’s why simply feeding documents into a model isn’t enough — you need a framework that governs how assumptions, inputs, and outputs are handled so the analysis stays consistent and traceable. Once that layer exists, the model becomes much more useful as an analytical engine rather than just a conversational tool.
0 Reply
Nate Marshall
Real Estate Consultant · Evergreen, CO · Member since 2018 · 1k+ posts · 740 votes
7mo
Quote from @Michael Eskenasy:

@Nate Marshall - having structure around the inputs definitely helps. The reason systems become necessary though isn’t just document organization, it’s the nature of the models themselves. AI doesn’t have durable memory and it doesn’t reason deterministically, so the same analysis can follow slightly different paths depending on context, session state, or how earlier information is weighted in the window. That’s why simply feeding documents into a model isn’t enough — you need a framework that governs how assumptions, inputs, and outputs are handled so the analysis stays consistent and traceable. Once that layer exists, the model becomes much more useful as an analytical engine rather than just a conversational tool.

100% wrong on that. It's generally operator error and we have too many basement dwelling incels with 7 screens in their "command centers" in the ai industry.
0 Reply
Michael Eskenasy
Investor · Pacific Northwest · Member since 2026 · 538 posts · 314 votes
7mo
Quote from @Nate Marshall:

Quote from @Michael Eskenasy:

@Nate Marshall - having structure around the inputs definitely helps. The reason systems become necessary though isn’t just document organization, it’s the nature of the models themselves. AI doesn’t have durable memory and it doesn’t reason deterministically, so the same analysis can follow slightly different paths depending on context, session state, or how earlier information is weighted in the window. That’s why simply feeding documents into a model isn’t enough — you need a framework that governs how assumptions, inputs, and outputs are handled so the analysis stays consistent and traceable. Once that layer exists, the model becomes much more useful as an analytical engine rather than just a conversational tool.

100% wrong on that. It's generally operator error and we have too many basement dwelling incels with 7 screens in their "command centers" in the ai industry.

Nate, I think you’re describing the typical pattern where someone throws an LLM on top of a workflow and hopes the model figures things out. That criticism is valid for a lot of “AI automation” out there. But that isn’t the architecture I’m building.

What we’re running is a deterministic system where signals enter through defined ingress points and are normalized into a structured payload before any reasoning layer touches them.

At a high level the system works like this:

Signals enter from multiple sources — email replies, scheduling responses, inbound forms, CRM updates, listing engagement data, and other interaction surfaces. Those signals are not interpreted immediately by an AI model. They first pass through an ingestion layer that extracts structured fields and attaches metadata such as workspace ID, message thread, timestamp, and origin.

Once ingested, those signals are normalized into a canonical schema. Every event becomes a structured object that can be routed deterministically through the system. This eliminates ambiguity about what the system is acting on.

From there the signal moves through a set of explicit stages.

Stage one is classification. Here models may assist with narrow tasks like determining intent categories — for example whether a message represents meeting readiness, a request for information, an objection, or a negative signal. The model does not decide what happens next; it only labels the signal.

Stage two is rule evaluation. Deterministic logic maps those labeled signals to state transitions inside the workflow. For example, a meeting-ready signal routes to scheduling orchestration, while a needs-information signal routes to the knowledge response layer.

Stage three is orchestration. The system executes predefined actions such as generating a response draft, updating conversation state, triggering follow-up workflows, or scheduling availability checks. Every action is executed through explicit steps rather than autonomous decision making.

Every step produces a run log entry. The system records inputs, outputs, timestamps, and state transitions so that every workflow execution can be inspected later. Nothing is hidden inside opaque reasoning loops.

The key design principle is containment.

The AI layer does not run the system. It performs tightly scoped functions inside a governed pipeline. The workflow engine controls the order of operations, the rules for routing, and the actions that can occur.

That structure exists specifically to reduce operator error and eliminate the black-box behavior that people worry about with agent-driven systems.

So the model is not sitting in a “command center” making decisions. It’s functioning more like a specialized classifier or interpreter inside a deterministic process.

If you see a flaw in that architecture, I’m genuinely interested in hearing where you think it breaks down. But the system you’re criticizing isn’t the one we’re actually running.

here you are bud, since you felt the need to say 100% wrong…feel free to go through and come back and tell everyone where I’m wrong.
https://docs.google.com/document/d/1L8iiyTrJILLxWybHQJIsp1Pb...
0 Reply
Michael Eskenasy
Investor · Pacific Northwest · Member since 2026 · 538 posts · 314 votes
7mo
@Dominic Mazzarella you should catch up too
0 Reply
Michael Eskenasy
Investor · Pacific Northwest · Member since 2026 · 538 posts · 314 votes
7mo
@Nate Marshall

What we’re actually launching next week is not “AI telling people what property to buy.”

It’s a structured outbound conversation system.

The purpose of that system is to create real conversations with property owners and operators before a property becomes a listing, before a broker blast goes out, and before the same opportunity is being looked at by everyone else.

The way it works is this:

First, the system defines a target universe. That means it starts with a specific market, a specific ownership profile, and a specific set of conditions we care about. That could be based on hold period, geography, operator type, ownership concentration, or any other factors that suggest a conversation is worth starting.

Second, it layers in the context around that target. That includes whatever signals are relevant to the market and to the kind of owner being contacted. Sometimes that is ownership duration. Sometimes it is local operating pressure. Sometimes it is changing market conditions, financing pressure, insurance pressure, or simple market positioning. The point is that the outbound is not random. It is context-aware before the first message is ever written.

Third, the system builds the outbound message itself. And this is the part people keep misunderstanding. It is not just “generate an email.” The message is built from a structured compose process that takes the known information, the positioning rules, the market context, and the purpose of the outreach and turns that into a draft that is designed to start a conversation. Not close a deal in one shot. Not blast a template. Start a conversation that makes sense to the person receiving it.

Fourth, that message gets written into the outbound channel in a controlled way. We’re not talking about random copy pasted text floating around in an inbox. The system creates the draft, preserves the structure, and puts it into the outbound workflow so it can actually be sent and tracked consistently.

Fifth, once the message is sent, the system does not stop. This is where it becomes different from normal outbound. It tracks the conversation state. If somebody replies, the reply is not treated like a random inbox event. The system resolves who that person is, what thread this belongs to, what has already been said, what the tone is, what kind of response it is, and what category of follow-up makes sense.

That means the system can distinguish between:
- genuine interest
- soft curiosity
- a brush-off
- a bounce
- an out-of-office reply
- a timing objection
- a request for more information
- a future opportunity that is not ready yet
That matters because most outbound dies right there. Someone replies, the whole thing becomes manual again, and all of the structure disappears. In this model, the reply becomes part of the system.

Sixth, the system generates the next step based on that conversation state. If the right move is a soft follow-up, it drafts that. If the right move is clarification, it drafts that. If the right move is to leave the person alone for a while and follow up later, it can schedule that too. So instead of outbound being one email and then chaos, it becomes a managed conversation pipeline.

Seventh, every step is logged. That means the system keeps track of what was sent, when it was sent, what was received back, what changed in the conversation, what assumptions were present, and what happened next. So over time it is not just sending emails. It is building memory around the outreach process.

Eighth, that memory starts to matter. Once you have enough outbound and enough replies, you can start to see patterns in who responds, what kinds of pressure points create engagement, what kinds of messages work in which markets, what kinds of owners are open to discussions, and how those conversations evolve over time. At that point outbound stops being a guessing game and starts becoming an intelligence layer.

That is why I keep saying this is not “AI replaces experience.”

The system is doing a very specific job:
- selecting who is worth talking to
- building context before the first message
- drafting the outreach in a structured way
- tracking replies as stateful conversations
- generating coherent follow-up
- logging every step
- preserving the decision trail
- and turning the entire process into a repeatable operating system instead of one-off hustle
So when people say “AI can’t buy property” or “AI can’t replace operator judgment,” I agree. That’s not the claim.

The claim is that most people are still doing outbound like it’s 2009:

manual research, manual drafting, manual follow-up, manual memory, manual judgment, and a ton of dropped conversations.

This system changes that.

It creates a structured outbound environment where conversations can start earlier, be tracked more coherently, and compound into real opportunity over time.

That’s the piece going live next week.

And because a few people here seem interested in how that translates economically, I attached the case study that walks through the cost structure behind a thoughtful outreach email and why outbound done properly is much more expensive cognitively than most people think.

https://drive.google.com/file/d/10vCDviPmc5NmfJL-J3f1UfmEKsuefMrO/view?usp=drivesdk
1 Reply
Nate Marshall
Real Estate Consultant · Evergreen, CO · Member since 2018 · 1k+ posts · 740 votes
7mo
Quote from @Michael Eskenasy:
@Nate Marshall

What we’re actually launching next week is not “AI telling people what property to buy.”

It’s a structured outbound conversation system.

The purpose of that system is to create real conversations with property owners and operators before a property becomes a listing, before a broker blast goes out, and before the same opportunity is being looked at by everyone else.

The way it works is this:

First, the system defines a target universe. That means it starts with a specific market, a specific ownership profile, and a specific set of conditions we care about. That could be based on hold period, geography, operator type, ownership concentration, or any other factors that suggest a conversation is worth starting.

Second, it layers in the context around that target. That includes whatever signals are relevant to the market and to the kind of owner being contacted. Sometimes that is ownership duration. Sometimes it is local operating pressure. Sometimes it is changing market conditions, financing pressure, insurance pressure, or simple market positioning. The point is that the outbound is not random. It is context-aware before the first message is ever written.

Third, the system builds the outbound message itself. And this is the part people keep misunderstanding. It is not just “generate an email.” The message is built from a structured compose process that takes the known information, the positioning rules, the market context, and the purpose of the outreach and turns that into a draft that is designed to start a conversation. Not close a deal in one shot. Not blast a template. Start a conversation that makes sense to the person receiving it.

Fourth, that message gets written into the outbound channel in a controlled way. We’re not talking about random copy pasted text floating around in an inbox. The system creates the draft, preserves the structure, and puts it into the outbound workflow so it can actually be sent and tracked consistently.

Fifth, once the message is sent, the system does not stop. This is where it becomes different from normal outbound. It tracks the conversation state. If somebody replies, the reply is not treated like a random inbox event. The system resolves who that person is, what thread this belongs to, what has already been said, what the tone is, what kind of response it is, and what category of follow-up makes sense.

That means the system can distinguish between:

genuine interest

soft curiosity

a brush-off

a bounce

an out-of-office reply

a timing objection

a request for more information

a future opportunity that is not ready yet

That matters because most outbound dies right there. Someone replies, the whole thing becomes manual again, and all of the structure disappears. In this model, the reply becomes part of the system.

Sixth, the system generates the next step based on that conversation state. If the right move is a soft follow-up, it drafts that. If the right move is clarification, it drafts that. If the right move is to leave the person alone for a while and follow up later, it can schedule that too. So instead of outbound being one email and then chaos, it becomes a managed conversation pipeline.

Seventh, every step is logged. That means the system keeps track of what was sent, when it was sent, what was received back, what changed in the conversation, what assumptions were present, and what happened next. So over time it is not just sending emails. It is building memory around the outreach process.

Eighth, that memory starts to matter. Once you have enough outbound and enough replies, you can start to see patterns in who responds, what kinds of pressure points create engagement, what kinds of messages work in which markets, what kinds of owners are open to discussions, and how those conversations evolve over time. At that point outbound stops being a guessing game and starts becoming an intelligence layer.

That is why I keep saying this is not “AI replaces experience.”

The system is doing a very specific job:

selecting who is worth talking to

building context before the first message

drafting the outreach in a structured way

tracking replies as stateful conversations

generating coherent follow-up

logging every step

preserving the decision trail

and turning the entire process into a repeatable operating system instead of one-off hustle

So when people say “AI can’t buy property” or “AI can’t replace operator judgment,” I agree. That’s not the claim.

The claim is that most people are still doing outbound like it’s 2009:

manual research, manual drafting, manual follow-up, manual memory, manual judgment, and a ton of dropped conversations.

This system changes that.

It creates a structured outbound environment where conversations can start earlier, be tracked more coherently, and compound into real opportunity over time.

That’s the piece going live next week.

And because a few people here seem interested in how that translates economically, I attached the case study that walks through the cost structure behind a thoughtful outreach email and why outbound done properly is much more expensive cognitively than most people think.

https://drive.google.com/file/d/10vCDviPmc5NmfJL-J3f1UfmEKsuefMrO/view?usp=drivesdk
This is already out. It's called RCIC and you're late to ta party you were never invited to, Cry harder boi!
0 Reply
Michael Eskenasy
Investor · Pacific Northwest · Member since 2026 · 538 posts · 314 votes
7mo
@Nate Marshall seriously Nate...you are embarrassing your self.

Nate — you should probably slow down and actually read the document before announcing that something “already exists.” Right now you’re arguing against something you clearly didn’t bother to understand.

RCIC is a tool stack for investors. What you’re describing is a bundle of prompts, underwriting templates, and AI utilities that help someone evaluate a deal faster. That’s a productivity layer. It sits on top of the work a human investor is already doing.

What I’m describing is an architectural layer beneath that entire workflow.

There’s a fundamental difference between a tool and infrastructure.

RCIC appears to operate at the application layer:
- a user inputs deal variables
- the system runs some analysis
- it outputs underwriting or guidance
- the human investor decides what to do next
That’s a classic decision-support tool.

The system I’m describing operates at the cognitive infrastructure layer. Instead of helping a human evaluate one deal, it models the information environment itself.

The architecture looks more like this:
1. Signal Ingestion
2. Normalization + Intent Modeling
3. Deterministic Decision Graph
4. Stateful System Memory
5. Autonomous Workflow Execution
In other words, the system is not answering the question:

“Is this a good deal?”

It is modeling:

“What is happening in the market and who should act next?”

That is a completely different category of system.

RCIC is closer to something like:
- a calculator
- a spreadsheet
- or an AI copilot for underwriting.
What I’m describing is closer to a coordination engine that governs the entire information flow around a market.

Saying “RCIC already exists” in response to that is like responding to someone describing an operating system by saying:

“Yeah, Excel already does spreadsheets.”

Those two things live at entirely different layers of the stack.

So before you declare that something already exists, you should probably read the material that was posted. The document you replied to literally walks through the system architecture, the signal layer, the orchestration layer, and the cost structure of the workflow.

Right now you’re arguing with a strawman that exists only because you didn’t take five minutes to understand what you were responding to.
0 Reply

Carl Mcknight

OP

Rental Property Investor · TN · Member since 2013 · 54 posts · 50 votes

7mo

Hey Michael, thanks for your inquiry. What I've started playing with are basic inputs, such as: Vacancy rate in the metro, average rents in the area, unemployment rate, population growth since so and so year, major employers, crime stats in the specific city and area. Those are some of the basics I tell the agent when I am creating it. The next step is to tell it exactly what it is, such as, "you are a apartment acquisition specialist that works for a large private equity firm", a prompt like this causes the agent to research exactly what a acquisition agent does for a private equity firm. Once the agent understands what exactly their responsibilities are going to be, it starts writing the code for itself and then creates an agent that will do those specific tasks. Your agent then is existing, and all you do is upload the data you have collected and it begins to process it, its not perfect, but it does a pretty good job. Plus, you can do the aforementioned over and over, it just continues to do the work you ask it. You can converse with the agent at anytime before, while, and after it has completed the analysis. Lets say you notice something in the analysis and your need greater clarification on it, you just go in the chat box and start asking it the question. If the response is too technical, you tell to it to explain it as if you are a 6th grader and use plain simple english. Hope I answered your questions.

0 Reply

Michael Eskenasy
Investor · Pacific Northwest · Member since 2026 · 538 posts · 314 votes
7mo
@Carl Mcknight

Carl — first off, credit to you for even experimenting with this. Most people I talk to about AI in real estate stop at “upload the documents and ask questions,” so the fact that you’re thinking about roles, inputs, and structured analysis already puts you ahead of where most people are approaching it.

The part that starts to get interesting with these models is that the challenge usually isn’t the data itself — it’s the structural behavior of the model. Because the context window is limited and the model isn’t deterministic, the reasoning path can shift depending on how earlier information is interpreted during the session.

So even if you feed it the exact same inputs, the analysis can quietly follow a slightly different path the next time it runs.

That’s not really a flaw in what you’re doing — it’s just how these systems behave today.

Where things start to change is when you build a framework around the model that stabilizes how inputs, assumptions, and reasoning paths are handled so the analysis becomes traceable and repeatable instead of conversational.

Once that layer exists, the model behaves very differently — it stops acting like a chat assistant and starts acting more like an analytical engine.

Either way, it’s cool to see someone else pushing into this area. Not many people are exploring it yet.
1 Reply
Carl Mcknight
OP
Rental Property Investor · TN · Member since 2013 · 54 posts · 50 votes
7mo
Quote from @Michael Eskenasy:

@Carl Mcknight

Carl — first off, credit to you for even experimenting with this. Most people I talk to about AI in real estate stop at “upload the documents and ask questions,” so the fact that you’re thinking about roles, inputs, and structured analysis already puts you ahead of where most people are approaching it.

The part that starts to get interesting with these models is that the challenge usually isn’t the data itself — it’s the structural behavior of the model. Because the context window is limited and the model isn’t deterministic, the reasoning path can shift depending on how earlier information is interpreted during the session.

So even if you feed it the exact same inputs, the analysis can quietly follow a slightly different path the next time it runs.

That’s not really a flaw in what you’re doing — it’s just how these systems behave today.

Where things start to change is when you build a framework around the model that stabilizes how inputs, assumptions, and reasoning paths are handled so the analysis becomes traceable and repeatable instead of conversational.

Once that layer exists, the model behaves very differently — it stops acting like a chat assistant and starts acting more like an analytical engine.

Either way, it’s cool to see someone else pushing into this area. Not many people are exploring it yet.
Great points you have brought up Michael. Thats what I am trying to accomplish now, having the agent get as close as possible to an analytical engine. I'm testing the aforementioned out with a stock trading agent I have created, I'm not a day trader, just have been curious if an agent could run off certain parameters (momentum movements with certain stocks, stop losses, etc) and then run autonomously and over a month time frame, and self correct where it learns what worked successfully and what didn't. It's already working on a Paper Account (monopoly money) with Alpaca, I'll see how it does. Thanks again for your input and thoughts, I appreciate it.
0 Reply
Michael Eskenasy
Investor · Pacific Northwest · Member since 2026 · 538 posts · 314 votes
7mo
@Carl Mcknight that's awesome bud. You wanna compare notes? Hit me up!
0 Reply

Bill B.

Investor · Las Vegas, NV · Member since 2013 · 8k+ posts · 10k+ votes

7mo

It is nice of the seller to give you a $700k discount rather than pay someone to fill 10 apartments. You’d lose over 75% of your upside if they had a property manager or handed out $10k bonuses for filling each of those 10 units.

Using your numbers I don’t understand you’d walk away at $1.6M. You aren’t willing to do 3 months of work to make $800k? Especially when 75% of it is simply filling vacancies and the other 25% is hiring someone. I know lots of people who’d do 3 months of work for $300k. I can’t believe you won’t be outbid by them. I assume they didn’t see the listing, know more, or are truly too busy/lazy. Good luck, let us know next week if you got the deal.

Ps. I didn’t see an adjustment in value for dropping rent to make up for utilities no longer being included. Maybe I missed it. But obviously you’ll have to drop rent by as much as utilities cost or it would be cheaper and faster for you just to increase rent the amount of utilities if you believe they will pay the higher total. Not that it really matters as that’s only 25% of the gain. 75% is the seller being so rich and lazy they won’t fill those 10 units for $700k. Good luck and let us know what you end up paying. Personally if you trust your numbers $2M is certainly still a good deal. Take the $400k gain and sell if money’s tight.

0 Reply

Nate Marshall

Real Estate Consultant · Evergreen, CO · Member since 2018 · 1k+ posts · 740 votes

7mo

RCIC Operators are able to analyze any deal (all asset classes and strategies) in under 60 seconds and reporting 97% accuracy. It is using OpenAI. It is exceeding expectations. I also have 1500+ hours into it and have uploaded 12,000+ documents including all of my libraries. Plus 19,000 lending, family office hedge fund and fund manager contacts.

So AI not only does this but exceeds.

0 Reply

Ryan Stomel

Property Manager · Calabasas, CA · Member since 2026 · 141 posts · 67 votes

7mo

Good breakdown, and your note on the RUBS problem is exactly right — it's one of the most consistently mismodeled line items I see in multifamily underwriting. The AI did well to flag it. I come at this from the commercial side (NNN leases, retail/industrial, not multifamily), but the same issue shows up constantly when brokers throw RUBS or CAM pass-through assumptions into pro formas without doing the market comp work first. On the AI workflow question — I've had similar results using it to do first-pass reviews of T12s and lease abstracts. The value isn't in replacing judgment, it's in making sure you don't miss something obvious when you're looking at 20 deals a week. The management fee normalization is a good example: that's exactly the kind of thing that should be flagged automatically, and it was. Where I've found the limits are anything that requires local market texture — understanding what a comparable NNN rent actually is in a specific submarket, or whether a tenant's use clause is going to create problems down the road. The AI can prompt you to ask the question, but you still need someone with feet on the ground to answer it.

0 Reply

David Olusola

Member since 2026 · 11 posts · 3 votes

7mo

Carl this is one of the most practical AI breakdowns I've seen on here. The management fee normalization catch alone justifies the whole workflow. Most people reviewing that T12 manually would have missed it or caught it too late. The point about AI flagging what to ask rather than answering everything is exactly right local market texture still needs a human.

0 Reply

Ryan Stomel

Property Manager · Calabasas, CA · Member since 2026 · 141 posts · 67 votes

7mo

@David Olusola — agreed on the management fee catch. That's a normalization that should be standard but rarely is, and the fact that the AI flagged it without being prompted is exactly the right use of the tool. The "flagging what to ask" framing is how I think about it too — it's a forcing function for due diligence, not a replacement for the judgment calls that come after.

1 Reply

Using Claude Cowork to Underwrite a Deal in 10 mins.

Using Claude Cowork to Underwrite a Deal in 10 mins.

Most Popular Reply

21 Replies