In this episode, a new AI workflow tackles due diligence on a three parcel infill lot in a hot Midwest market, priced at $250K for roughly three acres (three separate one acre parcels, one carrying an old storage barn/silo). The base case grosses $390K to $400K, but the deal stress test surfaces utility red flags severe enough to threaten viability against a seller demanding a nine day close.
Key Takeaways:
- Build Skills From Winning Threads Turn a successful Claude conversation into a repeatable skill, then generalize it past the original deal type (a triple net coffee shop submission became an all purpose deal stress test).
- Water Tables Can Kill Infill Lots Soil borings hitting water at two to four feet make two of the three lots unsuitable or prohibitively expensive for septic, turning written sewer confirmation into the dealbreaker.
- Pressure Test Vendor Cost Quotes The operator’s $10K to $15K septic estimate for the best lot likely runs two to three times higher given regional water table conditions.
- Always Validate the AI’s Source Forcing the model to show its evidence exposed that its wetlands sketch (drawn from blurry 2019 reports already past the five year local validity) could not be confirmed, and that only six of eight expected borings were delivered.
- Stack Skills Into a Pipeline The stress test memo feeds the Chris Duff voice skill to draft the follow up email, reducing the work to minor trims rather than stripping out AI tells.
Listen to the full episode to see exactly how stacking AI skills converts a fast moving, complex, out of wheelhouse deal into a quick validated go or no go call.
(Podcast transcript below)
Welcome to Get Serious, where at Serious Land Capital, we have funded over six and a half million dollars worth of land deals with industry leading 41% operating margins. So today I wanted to go over a new AI build that’s assisting our due diligence for real estate properties, specifically land. I mean, that is our focus here. so like it comes without saying at this point, you if if any of you have been
Tuning into you know any of our content, either the podcast, newsletter, what have you, like AI is paired in pretty much everything that we do now. so that is just again, what what we think our highest leverage point is, you know, besides our existing capital and our judgment that we bring to to real estate deals, plus some of the relationships. I mean, it it’s a whole grab bag that that we’ve been trying to build over over the years for a business here. but
you know, always trying to build in more leverage points and how do I make this business simpler to manage with less necessity for you know subjectivity or reliance on individual contributors, whether it’s myself or somebody else, and you know, that just makes processes within the business more prone to errors.
naturally. and you know, I I think I’ve talked a lot about more recently, you know, we’ve been building out this SLC chat software that’s like taking all of our underwriting knowledge and building it into an AI backend to effectively have you know all of serious land capitals direct knowledge for underwriting land deals that you can just plug and play and
You know, the current state of that is trying to build this cowork plugin that can handle a lot of the comping and you know automatically pulling the correct comps at least for a human review afterward. so that’s still in progress. but you know, so many of these other deals, you know, especially the higher value ones being thrown our way routinely. there’s just so many other parts of due diligence required.
know, especially when it comes to utilities or you know, existing soil tests or you know, other engineering studies and so forth. I know I went over more recently a larger entitlement deal that we were doing in the Western US that just had tons of documentation that we plugged into AI and really helped us suss out. Okay, what are the follow-on due diligence questions that we need to be asking here.
And so what I wanted to do is like, okay, and and this is always the question with AI. It’s like, okay, if you build out a cowork solution or a back and forth conversation with an LLM and it gets you a solid result, it’s like, okay, do I want to have to go through this entire process again? or can I take the learnings that we did from this particular process and build it into a repeatable
you Claude Skill, for instance, or potentially even a a project if it requires certain you know pre-existing knowledge base items. I won’t get too heavy into the AI jargon side here. I really want to focus on the the on the process and the thinking behind it. so like for example for some of these deals that have been trickier to look at are like somebody sent us a you know triple net retail coffee shop.
land deal, which yeah, like in the past, we might just like throw that out right away because it’s just, you know, it’s out outside of our wheelhouse, a lot that we’d be, you know, a huge learning curve that we’d have to jump up, especially if the operator hasn’t done one of these before. And so we have to rely more on our diligence to understand, okay, what are the missing pieces here? so in circumstances like that, I was able to take, you know, all that
DD that the operator brought us, what should I be asking for in a deal like this? Can we look at the back end numbers and start to validate them? And that was extraordinarily helpful because now it’s like, okay, I instead of just you know getting overwhelmed or throwing out a deal that’s outside of our wheelhouse, it’s like now I can just assess this quicker to see, is there anything here for us to dive in deeper? And if not, okay, fine, but now we at least have an idea on how we might approach a deal like this in the future.
and so after I’d gone through that exercise again, it’s like, all right, I I want to be able to repeat this going forward here. So generally what I do if I have a solid claude thread that worked really well and got me a solid output, you know, I’m gonna be like, okay, I want to repeat this skill going forward here. And you might have to wrap claude to you know, orient a little bit more, like if there was just
Too many specifics within a particular thread that you had with Claude, like to make it more general. Hey, I I don’t want it just to focus on like triple net lease deals here. I want it to be more widely applicable to like any type of land deal that might be centerway, or you could even make it more broadly applicable to any type of real estate deal. and basically have a
set of guidelines where okay, I can create a subfolder, throw all of the due diligence items in, all the engineering ports and you know, pro forma numbers, even, you know, the PDF of all the email exchanges that we’ve had. You don’t even need to do a PDF, you could just have an MCP connection to find whatever deal you’re you’re working working in that might be a little bit slower but it could also pull like multiple email threads if you needed to. I
really prefer to just have like a centralized database. You know, it’s really easy to get PDF of email threads and so forth. Like I I want all of that more easily referenced by the A AI model. and so I really know like the full set of knowledge that that I’m working with. So I’ll throw that all into a folder and then had Claude create a skill, which I call the what did I call this one? Just the the the deal stress test skill. Usually Claude kind of name names its skills on on its own here. So
You can always go back and name them, but generally it’s going to default to whatever it feels like the general gist is for a particular skill. and again, the guidelines are like, hey, any type of deal that I want to review within a land space, you attach this certain folder with the info related to it. Again, the more info, the better that you can that you can collect.
And ha have it go to town on that. And sometimes like it can be a bit rougher. Like sometimes on these, like, hey, I just really want to figure out what what what the heck do these soil tests even mean? I barely read these things. what should I be asking for here? And I don’t even include all like the purchase agreements and prices and all that. and you know, credit to this the quad scale name. Hey, like what we don’t even know what the price is, like what what’s going on here? but then I just refocus it, just ignore that for now. I just need to figure out what these soil tests mean.
And then we can figure out whether there’s even any potential to go in deeper on this. So again, you can always direct the AI to you know what you specifically want to get done individually at any given time. and you can always adapt the the skills that you build to whatever you know current current problem set that you’re working with.
And so that that’s kind of the the guideline there. And so, you know, if I have a folder filled with items, I’ll, you know, just have this deal stress test really built with like our underwriting and you know, also taking on, you know, the role of any expert within that particular field, like somebody within like the triple net lease perspective or you know, infill lots or rollouts, whatever. and so it’s always looking to really, really push.
Any potential red flag, again, try to be as conservative of underwriting as you know we try to do internally at serious land capital, to really poke as many holes as possible and build memos for follow-up questions and what to request going forward. So that was the thinking behind building this skill. and now that I’ve been testing it for a couple of weeks, just to again run through.
what how how we’ve been doing this. And I th this is again like, you know, AI is so iterative. you build a piece here that works pretty well. And it’s like, okay, I can actually take this a step further and a step further and build into a a full package, which is like, you know, the skill that I’m building here would just be added into the Claude Cowork skill, which again, comping and pulling live from Redfin or Zillow and what have you. Like it’s just a more complex
Process to do that properly, which is just why it’s taking a bit longer. Whereas like, okay, throwing a bunch of files into a folder and trying to assess some red flags, it’s just an obviously simpler build, but extremely useful if you do it properly. And that’s like an additive function to any deal that you would be reviewing. So like I I’m just, you know, teaching myself along the way, learning along the way as well, too, on how best to utilize these AI tools so it can just reduce the mental burden of
you know, potentially considering some of these really complex or high cost projects. so you know to to give you another concrete example one that’s super fresh we’ve looked over in the past few days here it’s this reparcel basically infill lot within a hot Midwest market. Again, you know, you’ve heard me talk
number of times and it’s no secret even the mainstream media is like more on top of this with you know the Midwest and Northeast being kind of the primary sellers markets at the moment. still some warning signs, especially from an L overvaluation perspective, you know, be careful for real estate hyperlocal, always do your own research. I always have to throw those disclaimers in there. But yeah, this is a particular market, you know, using reventure that’s pretty encouraging. It’s also, you know, near
you know, where some of my wife’s family lives and where you know I grew up in in the Midwest as well, too. It’s like, okay, I’ve some familiarity with with with the area here. and again, like a three parcel infill out, what does that even mean? So it’s like roughly three acres that that are contiguous to each other, but i it’s three separate parcels that are like roughly one acre each. And one of the parcels has have an older
storage barn slash silo on it. questionable whether it should be
The yeah, what what’s the phrase that my new dad brain is creeping in here? like demolished, yeah. That that that that’s what I’m looking for. you know, did do we clear that off? potentially, who knows? But you know, really nice a area of town, a luxury condo being built across the street, which apparently had interest in this particular property. So really, really interesting. you know, but a high cost, right? 250 grand for three acres.
and you know, in Phil Lots, we’ve had our run of bad experiences in in the past with them. So our threshold is always higher with those. decently high variance in the market. And it’s one of those deals where at least looking from the comps, the indication is that like there would be very low likelihood for losing money. Again, that’s like always our first threshold. and you know, this particular property was sourced by an operator who lives locally and the
The guy selling the seller selling the lot is not marketing this widely, you know, putting up a sign and but you know, not an MLS or anything like that. So you know, pretty intriguing from that perspective, of of just kind of finding, you know, what’s basically a solid off-market opportunity. Those are always going to be fairly interesting to us.
But the chance of having a 2x gross conservative margin on it is probably not there. And so usually that’s what we’re gonna be looking for in a lot of these, but yeah, the downside protection is solid. So it it would be likely as a base case to gross at least 390 to 400k on a 250k purchase across three parcels.
So that like that makes it very borderline for us for having in going after something like that. But again, like the deck, like there was looking at the market, it was fast moving, one of the best markets in the country, and seemed to have near 0% chance of losing money on the deal based on on the com set that we can that we can figure out. so pretty interesting. But you know, also the seller was playing hardball and wanted to close in like less than two weeks. I guess it’s a lot of money.
when we don’t have all the answers. Like it’s not an immediately clear, hey, just buy this and figure it out later. That’s almost like never the case nowadays. you you get you gotta figure out all all all your answers here, but like okay, well let’s take a look at this one further. so you know, some of our questions here.
know, especially on infill lots, like a lot of it comes back to utilities here. Like we doesn’t have sewer access, is it septic? Like, okay, what what what are the soil tests, anything like that, look like here? Are there are there significant risks from a utility perspective? Because those can really make or break infill lots specifically and drastically change the you know underlying value of the lot or how much work the
and buyer might have to do from a value add perspective or adding septic systems, especially engineered ones or aerobic or what have you, and wells potentially if if there’s not public water.
So all of that had to be considered and the operating team seemed to have a lot of these these answers here and they mentioned, hey, yeah, we have some soil report reports, some wetlands alineation. studies here can be sent over.
and so they were all sent over, but it’s like you know, 15, 16 pages, and a lot of it was like kind of poor resolution or you know, handwritten reports. I mean, many of you who’ve been in this business have seen, you know, some of these soil reports, and like it’s hard to parse out unless you’re you’re in the engineering space. like you really have to dive in and figure out, yeah, what exactly does any of this even mean? and so what I did, you know, to come back to the start of this conversation is I created a folder, threw it into the deal stress test, and I’m like, okay.
Here’s the base situation with this deal. We got a fast moving timeline. I really need to understand what are the flags here. What are the flags from the utility perspective? What else should we be be asking? and what could really go wrong with this particular particular deal if we, if we are missing on something here. so again, like pretty basic off the bat, not all the info that we would have been asking for still would have to comp it out further and speak with the broker and all that, but I needed to figure out the utilities. I knew that.
So I could start that deal stress test right away. and you know, very fortunately, again, some of these like new quad models Opus 4.8 using the max effort one, like it’s just it’s remarkable how just in-depth it goes into like at every possible question and you know file that you throw at at it. just huge context window. and
so it it it assessed this opportunity and is like, yeah, there’s definitely some concerns here. And the indication was like two out of the three lots had access access to sewer per the operator who brought us the deal, which was apparently by way of the city. But a huge flag in regard to this was those two lots that supposedly had sewer, or at least, you know, verbal confirmation of it.
probably would would be not suitable for septic or like would be really costly to install heavily engineered septic systems because there were, you know, soil boring tests done to where it was hitting water tables at like two to four feet roughly. and in in Midwest, like, you know, a lot of houses have would have basements, like that’s probably not gonna make it viable. and so like if sewer was not available for these, like the the cost would be
very significant if it if not unbuildable for these particular lots. And then the third lot, interestingly, that we thought was going to be septic, actually had the best soil test results. Like the water table was the lowest. So probably the you know cheapest septic system could be built there. But still like Claude was pointing out, hey, like these guys are quoting, it might cost 10 to 15K, but based on what we can figure out regionally and based on this type of water table, like might be double to even triple that.
so if you want to check in on that. so it it flagged that, plus it flagged, hey, like you gotta get the confirmation for the like get written confirmation for the sewer for these particular lots. And it’s indicate like it did all its own web search and was like, okay, yeah, there’s a sewer main going down this road, but it’s unclear whether there’s lateral connections that are available and what the cost would be and it if you can even hook up along this road. So you’ve got to figure this out. If you can’t get in writing, like this probably kills the deal.
So it just listed out, you know, some of these flags, what we got to follow follow up with. and then, you know, one of those, okay, checking in on on what that anticipated septic cost would be for that third lot. there was also supposed to be like eight borings done within the soil. We only got six. Like I wouldn’t have really known that unless I was super in in in depth looking over those 15, 16 pages. AI points that out right away. It’s like, okay, can we get the follow-on
information there. you know the wetland delineation study that was sent to us was done in 2019. but you know Claude was able to figure out hey locally in this area like any delineation study that’s beyond five years is actually out of date. So you would need a new one anyway. so okay what’s the cost of that? What’s the timeline to get one of these? and
Like some of these files sent over to us were a bit blurrier, like trying to figure out, hey, where exactly is the wetlands from this? And so I even asked Claude, like, can’t hey, can you take the aerial that I gave you of these three lots and just map where the wetlands delineation studies seem to indicate the wetlands were? So I can’t see anything on LAN ID. It’s not showing me any wetlands there. So where is it from the delineation study? And so it actually built out a sketch, not perfect, but you know, solid of where it it could interpret the delineation study to be.
But a critical point of that too, and like this is where a lot of people mess up with AI, is like they they just take its word much quicker than they should. Like you always have to validate. and you know, some of that’s not less like reading line after line after line, but more, you know, so e even like skimming to make sure, okay, it it is there really a chance of hallucination? Like, can we double check this work? And so, you know, especially for a sketch that was done from AI.
and like me glancing over the dots. It’s not like I didn’t look at them, but you know, I I wanted to see, okay, what’s the core source of truth that AI is pulling from? And me looking at the delineation reports, I’m like, I I really, you know, I I I I can’t eyeball this and see where the wetlands are coming from. So like I just pushed Claude A. How exactly did you figure this out? and you know, it it it went back and was like, yeah, you know, it’s a good call out there. And then it, you know, actually shows the evidence on.
How it determined that. And it’s like, yeah, it’s not quite positive because it’s just, you know, again, like fairly blurry images and can’t quite figure it out. You’re going to need a new delineation study anyway. But from what we can tell, that that this is roughly where things are. You still want to understand, okay, are there buffer limits on you know where to build on properties that might have some wetlands? We’ve dealt with a property in New Jersey before that had something like this. You couldn’t build within like 100 some odd feet.
Something like that. So especially on infill lots, you better make sure that you you have you know room to build with correct setbacks from wetlands areas. So like that was another question that we had to figure out one way or another. plus like okay, what’s the minimum lot size for you know, probably less so for sewer, but it’s still worth asking versus septic. you know, general rule of thumb is if thing, you know, lots are over a half acre.
Some areas of the country over an acre are required for septic. So like we got to figure that out, especially if one of them needs it for for certain. And because it like it was the same owner for this particular lot, we’ve seen this before too, where okay, it’s three separate APNs, but was it subdivided legally before? that has been an issue. So our understanding for this particular lot was title had cleared it. Yeah, okay. If title insurance is on it, great. that that clears up a lot for us.
But that’s still another flag that should be pointed pointed out on. so you know, we we put we put all of that together and like while I I was gonna get all these you know questions back to the operator who brought us the deal. again, if you just ask any, hey, like prep me a follow up, it’s all gonna sound like AI, right? And a bunch of dashes and so forth. So, you know, that this is where you just
overlap skill upon skill upon skill is you know I had the deal stress test prepped me my memo and questions I’m like okay then need the follow-up email here. So I have the Christoph voice skill that’s built on like all the email threads and and you know newsletter writing and everything like it understands you know my approach for for speaking especially in like you know client communications and so forth.
And so it’s like, okay, transform those follow-up items into my voice so it’s ready to be sent as an email here. Again, still not perfect. I I would say that that that’s an area that AI will, you know, still need to be you know, additional work on, but it was much easier for me to polish up and you know, shorten here and there. Again, a AI is probably more wordy on on average, so it just making things a bit more concise. is usually how you can
reorient it but still makes it a lot easier compared to like having to clean up m-dashes and stuff that is does not sound like me at all. And then I could fire off that email. Okay. And was also noting okay like there’s if you want us to close in nine days on this like without answers to these, like there’s just no way we’re gonna risk a quarter million dollars, especially on infil lots. so like go back, you know, if you can go back to the cell I mean if you can figure all this out great but like we need this.
in order for us to consider this in much more detail. and you know, you can always mention to the seller, like even if they’re playing hardball here, like, and and this is what we always tell people, is like, you know, like any serious buyer is gonna have these questions anyway. And so, yeah, if you don’t want to move forward with us, like it might take, you know, what what what’s the cost of another extension, another like 30 to 45 days? Whereas like, okay, the only way we were marketing this thing was like not even on MLS and so forth, like
Probably going to take you a lot longer to find another buyer and potentially even at a better price. Like we’re ready to move at 250K in cash. it’s a solid deal for for a lot of folks out there. so you know, that then you have to kind of reorient. and I I thought, okay, yeah, maybe the operator’s just gonna like walk and move, move, like seeing all these requests. but they got back to us like, yeah, what we’ll figure all this out. We’ll push the seller to get a an extension.
Because like if if you show, hey, we’re we’re taking the time to really review this thing and like here’s key concerns, like it’s showing your expertise in an area and and that you are legit in in you know knowing what to ask for and and what to look for and you know obviously have the money to close on a deal like this. that’s just gonna build build up trust in that direction as well. And you know, folks are gonna try to find a way to deliver, at least the the serious ones are. so.
you know TVD on where that that’s heading for this particular deal. but I just wanted to mention that to you as an example to where you know some of these deals again with like a lot of complexity or you know new areas that that you haven’t worked in before it’s like how how do I really quickly get an assessment on key questions that need to be answered, especially if something’s fast moving potentially, on whether we should dive in further or or not on this particular
Deal. So again, like these tools are just super, super helpful to do. Like again, so long as you’re validating what the AI is giving back to you, like reading initial soil reports and all that, I would take AI, especially the bet the solid models, like the you know, the opuses and the you know, the top end plot ones, the top end ChatGPT ones, even Gemini, like any day over a human. Like it’s it’s just gonna be way better at just consolidating and understanding all of the information versus like going page by page by page.
And and trying to figure out you know what where is this directly heading? Like unless you’re in that space, like yay, you have you know several years working as a civil engineer and you can just read these reports in your sleep. I don’t know many in investors who also have that background knowledge, like a handful will, but but not many, not many folks. and so you know, to have have a tool that can assist.
In in doing that. And so then okay, you can validate the important parts there and then figure out okay, get these questions answered. Let’s keep keep moving forward. So this gives you an idea on how to stack the skills, how to improve the underwriting process, and you know, how we’re thinking about it internally in our business. Okay, I’m solving this piece of the puzzle here. We’re building the cowork plugin to solve the comping piece as well, too, and just like get continuing to create more workflows, more automations within our business.
That’s only going to allow us to, you know, first and foremost, but build more freedom. And you know, I got like the a big forcing function is I’ve got, you know, two young kids at home, including a three week old. Like I have minimal time. and so d it’s just okay, how do I make this as simple as possible? Like that that that that’s usually a good indicator on how you can build qualified workflows for for your business here. So that that that’s what I have on my plate here and and why like it
It’s again forcing me to figure these pieces out. and again, like we’re just love the impact that that AI can bring to businesses. So hopefully this one is helpful today. If you have a deal for us to look at, this one’s a really exciting one. We’ll we’ll see. We we’ve been moving on some solid deals here recently. serious land.capital to submit your deal. and subscribe and share, everybody. Hopefully you enjoyed this one. Take care. I will talk to you all next time. Bye.
