Marc Rutzen is the CEO of HelloData.ai, a platform that automates the due diligence process for real estate investors. In this episode, Marc discusses the disruption potential AI has in the real estate industry and how HelloData.ai works to mitigate the mundane and manual tasks associated with the due diligence process of real estate investing.
Marc Rutzen | Real Estate Background
- CEO of HelloData.ai
- Based in: Chicago
- Say hi to him at:
- Best Ever Book: The Innovation Stack by Jim McKelvey
- Greatest Lesson: Don't try to do too much; narrow your focus, and hone in on one specific area of expertise.
Click here to learn more about our sponsors:
Slocomb Reed: Best Ever listeners, welcome to the best real estate investing advice ever show. I'm Slocomb Reed and today I'm joined by Mark Rutzen. Mark is joining us from Chicago; he is the CEO of HelloData.ai. It is an artificial intelligence company that automates the due diligence process for real estate investors. Mark, can you tell us a little bit more about your background and about HelloData.AI?
Mark Rutzen: Yes. My background is actually in real estate. I got my masters in real estate development and worked in the industry for five years as a developer, a consultant and broker, before launching Enodo, which was a company whose mission was to automate the underwriting process. Now, we built Enodo into a really powerful platform and sold it to Walker and Dunlop in 2019. Walker and Dunlop is a commercial real estate lender. From there, I rose from SVP of IT to Chief Product Officer, and built an internal product suite that added substantial efficiencies to the underwriting process, and worked with all the different departments through the company. In doing this, going from real estate to technology to working with a publicly-traded company to make their systems more efficient, I realized that automating the underwriting process is great, but people really like underwriting. They like tweaking assumptions, getting creative to make deals work.
So this time around, with HelloData we're really focused on the parts that people don't like to do. In the due diligence process, it's a lot of stare and compare; you've got documents on one screen, and you've got to compare it to documents on the other screen, make sure that the address matches, make sure the year built, and number of units. Uderwriters spend a lot of time doing this to prepare for submitting a loan. So our thought and our vision with this company is "Take all the documents that are related to the due diligence process, and take all the publicly available data that we can collect on the property that is the subject of the transaction, extract the data from the documents, cross-compare it, and then compare it to publicly-available data to make sure there are no anomalies, there are no things that need to be looked into. And if there are, identify them and draw people's attention to them, so they can more quickly go through the underwriting process."
Slocomb Reed: Tell us about the anomalies that you've seen HelloData come up with.
Mark Rutzen: Well, at this point, we're pretty early, so we're focused more on the extraction piece right now... But the vision is to extract from all these different documents simultaneously, and identify those anomalies. So things that commonly come up - if you have a right of first refusal in a commercial lease, is it actually subordinated to three other rights of first refusal? That's something that wouldn't be readily evident from a quick read, but we can pull that data out of the document and draw your attention to the fact that it's something you've got to look into in more detail. Or more simply, just looking at the appraisal versus the property condition assessment. Do they have the same year built number of units, the same property address? Very commonly, these things are different across documents, and that becomes a nightmare for an underwriter when they're trying to get something through at the last minute.
Slocomb Reed: I'm curious, Mark, what property types are supported by our platform, but also on which property types have you seen the most concerns be raised by your platform?
Mark Rutzen: We're focused primarily on multifamily... And as I've learned about the underwriting process from doing it for the past seven years, or building tech, I should say, for the past seven years, things like rent roll or an operating statement not being signed; or having the wrong date. Fannie and Freddie look at this and they will dock lenders; they have a score for each lender who submits loans to them. It negatively impacts the score if a rent roll was not actually signed, or if the date on that rent roll is too far in the past. It has to be within the last 30 days, I believe.
Little things like that take a ton of time to verify and go through all these documents and find any issues. And if an issue is found at the last minute, that could delay the approval of the loan. So that is a huge use case for the platform. But even more simply, just getting the data out of the documents in the first place is huge. How much time is spent going through an appraisal and looking at the comps, looking at the appraised values? The income approach, the sales comps approach, and just putting that into a spreadsheet to continue your analysis? That takes a lot of time. To do the same with a commercial lease - that's an analyst spending three, four hours going through a lease and trying to summarize, get to a lease abstract; or you can pay an attorney several hundred dollars. So to be able to do that in a matter of minutes - that's a huge value.
And one of the big things that we do that I think will get crazy adoption in the market - we have this [unintelligible 00:06:44.01] HelloData.ai email, that if people send a document to it, it will recognize the document type, extract the data from it, and email it back to you. Just that by itself -- even if we didn't do any of the due diligence analysis, that by itself is extremely valuable, because people spend crazy amounts of time with a document on one screen, a platform or another document type, or Excel on the other screen, just looking at the data, copying and pasting it over... I'm sure you've done it countless times. It's a lot of manual work, and it's the work that people hate to do, and that's why we want to automate it.
Slocomb Reed: I'm curious, with this platform where I'm just emailing HelloData and it's emailing me back the data points, as an investor acquiring property, but also as an owner-operator and a property manager, I do spend time looking at the lease agreements in place in properties where I'm either taking over management, or I'm buying and therefore taking over management... One of the things that I'm looking for in those leases though is some of the quirky terms that other operators or other landlords are putting in their lease, that I will be legally bound to for the life of that agreement if I take over management or ownership of the property. It's the kind of stuff that doesn't fit neatly into fields of a spreadsheet, like how much is the rent, when does the lease end, things like that. It's the weird stuff that I'm really looking for, because I can go pass off to a virtual assistant "How much is the rent? When does it expire? What are the late fees?" type things. What is HelloData doing to find the weird stuff?
Mark Rutzen: On hard variables, as you mentioned, the lease terms, base rent - that stuff we just extract. For the clauses that are more impactful and harder to understand, we summarize. So the algorithm will go through and look at that right of first refusal, it'll look at that permitted use clause, and just broadly summarize "You can do this, this and this with the property", or "These are the parameters of your right of first refusal." We use large language models to do that.
So everyone's familiar I think at this point with ChatGPT. We use OpenAI, but we also use other large language models, depending on the use. So those models are very well suited to summarize things, as I'm sure you know if you played around with it a little bit; you can summarize large volumes of text, and it comes up with a pretty concise summary.
I think the art behind it, and what we do is we're able to comb through the document, extract the text and the tables with very high accuracy, and then group together the parts that refer to a particular topic. So the topic is right of first refusal; it may be mentioned, or things related to that topic may be mentioned at various places in the document. We'll comb through entire text, surface all the places that refer to that clause, and then feed that to the large language model and let it summarize the terms for you.
So it's getting to a lease abstract as safely as possible, but there's an artful way to get there. You can't just go to ChatGPT and be like "Here's my commercial lease. Tell me everything I should look out for." There's a lot of real estate logic that has to be built on top of it for that to work.
Slocomb Reed: I'd like to transition the conversation Mark, but I could see myself and a lot of our listeners being interested in this, especially on a more case by case, document by document basis, because it's not all that often right now, like this month, that I have a lot of these documents that I need to review, but then all of a sudden, I have a bunch, or all of a sudden I have one, and I would love to save my time. What does setting that up with HelloData look like?
Mark Rutzen: It's pretty straightforward to get signed up. You can go sign up on our website and use the docs [at] HelloData.ai email, try that out with a few different document types. All you do is send your document, it will extract it, it will send it back usually in one to two minutes. So that is how you try it out. Now, if you want to add different support for other document types, we're adding new documents all the time. We focus first on appraisals and commercial leases, but we're now starting to tackle Purchase and Sale agreements... We did offering memos, we're refining that a little bit more to make it so people can screen deals more effectively... But to sign up, it's very simple. Go to the website, put in your information, and then you can send documents to that email.
Slocomb Reed: Mark, I'm poking around on your website right now, and I have to say - I don't know if this is a call out or not, but it doesn't look like your pricing is readily available on the website just from browsing. So if I wanted to do something like this, how much would it cost me?
Mark Rutzen: It's farther down on the front page. I think we need to break that into a separate tab there, but its price per page for the doc extraction, and the various products we have. We basically built a suite of simple minimum viable products that do one thing well, and then put each of those APIs available on the site. But now that vision of automating due diligence is bringing them all together, so that not only documents, but there's also floor plans, right? If you're looking at a commercial lease or a development, there are floor plans associated with it; we extract data from floor plans, too. If you're trying to cross compare data from documents to what's publicly available, that's where we have our [unintelligible 00:12:13.12] product. Each of these is valuable in and of itself, but the goal is to bring them all together into one automated due diligence platform.
Slocomb Reed: Looking through the list on your website, one thing I don't think we've discussed yet that I do find interesting... "LiquidRent.ai, providing price recommendations for apartments using supply and demand data." Now, for our larger multifamily listeners, when you have 100 plus unit properties and they're all identical, your property manager should have a really solid expectation of exactly what those apartments rent for, because their leasing activity is constant, and the product is fairly uniform. However, when you get into smaller properties and older properties and quirky properties, that can be difficult. So tell us what your platform does to make rent recommendations for those kinds of quirky units.
Mark Rutzen: Once you get below a certain size, because we're using supply and demand data, it's not going to work for every property. If we try to combine the rent source and liquid rent, which we may in the future, grabbing all the rent from all the different listing sites and pooling it together, and Liquid Rent is doing the revenue optimization - they serve different parts of the market. Liquid Rents - you probably need to be above 100 units; I think that's a cut-off for that to actually work, because we needed enough supply and demand data to actually drive the model.
But how it works... So we use a very similar algorithm to Google ads, how they forecast what the price should be as a function of the demand, the search volume for particular topics. So we've tried to make Liquid Rent the most transparent revenue management software. At the top of the funnel, the demand for each different type of floor plan helps forecast how those things should be priced. And based on the applications that convert to leases, that convert to renewals on the back end, we're looking at the whole pipeline and determining where you should set your price. But also, if you want to adjust that price upward and downward, we'll tell you what the impact of that price adjustment would be. So say we say you should charge $1,500 for this unit on this particular day, and you say "I think we should push the pricing to $1,550." We'll actually tell you "Okay, you can expect this many applications, this many leases signed." So prospects to applications to leases signed. We'll tell you how that will change as a function of where you set the price, and then you can have those pricing discussions with a lot more transparency. It's kind of like a pricing sensitivity.
Slocomb Reed: I do want to transition the conversation and get a bit more broad, actually, given your expertise. Stepping away from the AI that's designed to help me, I want to ask, given your expertise with AI, about a lot of people in a lot of industries feeling concern or feeling fear that their work will be replaced or simplified to the point that it's not going to provide gainful employment or gainful opportunities to earn for them. I have not personally played around with ChatGPT at the time of this recording, in part - and I want your response to this, Mark. I believe I'm in an industry where ChatGPT doesn't have any capacity to usurp my income earning potential. I don't know any AI that can own real estate yet, and collect those rents directly... But also, as a property manager - yes, there are lots of things that I do that could be automated, but as a construction manager, ChatGPT can't paint the apartment or replace the air conditioner. And so all of the things that I'm focused on right now professionally as a real estate investor and an operator of real estate-based companies, I haven't had any concern about what AI is doing in my industry the way that I know a lot of my friends have. Mark, tell us, going beyond real estate investors can use AI to make their operations more efficient and more systematized and save them time - beyond that, what disruption potential do you see for AI within real estate investing?
Mark Rutzen: I think a lot of the process of listing a property and bidding on a property - it's a lot of translation of data into a document format that is not really easy to ingest and work with. And then, a real summarization of the data from that document, where basically you can use generative AI to create parts of offering memos, for example. And you could use generative AI to summarize parts of offering memos on the other end.
So I think over time, where real estate people put together these documents that summarize the deals, they're gonna use generative AI on this side to do it and generative AI on this side to understand it, to the point where they're probably going to realize it doesn't make any sense to do it in a PDF format. At the end of the day, it's just two AIs talking to each other. "Give me the key deal terms in a digital format that's easy to digest, and stop converting it to document types that are really hard to work with."
It's crazy, because it's not just in producing the OM, but it's in the lending process - there's a loan narrative that's put together with the key components of the deal, and then that is given to Fannie and Freddie, and they put their narrative together of the key components of the deal... And then from there, I'm pretty sure there's other layers of players that are summarizing that deal over and over again.
So I think the big disruption that will come from this is we'll probably realize it doesn't make sense to do that deal summarization over and over again. Let's just come to a uniform data standard. I know [unintelligible 00:20:13.24] has been trying to do this for maybe 30 years at this point, to get a data standard that permeates all of real estate... But maybe generative AI gets us to that point. Everyone can put the data in whatever format they want, and they can translate to everyone else's format, to the point where you don't need to do OMs, [unintelligible 00:20:31.24] that sort of thing anymore.
Slocomb Reed: Intellectually speaking, when your starting point is "What can AI do?", there are a lot of applications in real estate, as you were saying, especially when you're being transactional, when you're buying or selling. There are a lot of places where AI can simplify, clarify, make things a lot more time-efficient. I'd like to come at this question from another perspective, Mark. Let's talk about me, or a hypothetical Slocomb here, whose primary source of income profit, lifestyle is the ownership of assets that produce cash flow, or that appreciate over time and I can sell for a profit. Speaking specifically to my ownership of assets, do you see AI doing anything to disrupt the way that people own assets for cash flow and appreciation?
Mark Rutzen: No, not in the immediate future. Hard to predict where things will go, but AI is not going to be able to do physical things like operating a property. It can make it much more efficient to do the paperwork aspect of it, but AI is not going to go buy real estate tomorrow. I'm saying this now, and then the singularity happens and --
Slocomb Reed: I know... The question is partially on behalf of my listeners, mostly on behalf of myself. I'm asking defensive questions. Like, where can AI come get me? And I just don't know where the answer is in that regard, when it comes specifically to the ownership of assets that produce cash flow. There are a lot of peripheral advantages to using AI, and other technologies. I can see where blockchain technology would be very beneficial for tracking the ownership of assets, but I don't see a blockchain actually owning anything. It's a weird question to ask... I did want to see whether or not you had an answer I wasn't aware of... And you've already touched on this second question, but for those of us whose business is predicated on being physically present, in order to be able to diagnose a physical or mechanical issue, or repair drywall, and paint, to diagnose and then address those kinds of physical issues, it seems to me that blue collar work is pretty safe, all things considered. In fact, I think there are a lot of ways that the economy has demonstrated to us over the last few years that blue collar work and blue collar workers are safe from the shifts happening in the economy and in technology.
Aside from some practical applications, like if you can plug into an analytics system to diagnose, AI could help with that. Are you seeing any other ways that AI could disrupt the way that blue collar work is done in America today?
Mark Rutzen: No, I think you hit it with the diagnostic piece. I could imagine -- even though contractors tend to not be the most tech-forward people out there, I can imagine that it's as simple as taking a picture with your phone, and having an AI actually say "That's a leaky pipe behind the wall. Here are the signs, here's why." "Okay, we can get to that." But the homeowner, if they don't know how to repair a pipe and drywall, they're not going to go do it themselves, and they probably shouldn't; don't touch plumbing, electrical. It's fraught with peril there, right? So on the diagnostic side - yes. Otherwise, until we get humanoid robots that are leveraging AI to think like humans, I don't think blue collar jobs are in any danger. But when we get there, I think we'll probably have many bigger problems to deal with than the state of employment in that type of job, if we got robots walking around everywhere, right?
Slocomb Reed: Mark, I definitely want to give you credit for bringing up an interesting point here, because it's really easy to just say, "No, Slocomb, you're right; those jobs are safe. No, Slocomb, you're right. Those assets are safe." But one thing that I hadn't considered was how valuable it could be to be able to use a piece of equipment to, quote/unquote, look inside of a wall and see what's happening in there without having to open the wall. Because just as a homeowner - I know you're a homeowner; we were talking about that before the interview... Just as a homeowner and an investment property owner, I'm thinking about the cost of opening the wall, I'm thinking about the cost of closing the wall as well, as I am making those repairs. Understanding that this is not your expertise, Mark, are you seeing advances being made in that direction?
Mark Rutzen: The AI world is kind of focused on things like healthcare diagnostics, or reading and understanding legal documents. On the healthcare diagnostics front, I've read something a few weeks ago where a diagnosis was made by a doctor that was actually wrong; it could have been one of three things, and the doctor actually diagnosed the wrong thing. And the person went to ChatGPT and said, "What could this be? Here are my symptoms", and they kind of went through it, and it actually picked out the right diagnosis... Which is kind of crazy. And then they went to a different doctor, got a second opinion, and actually fixed the physical issue that they had. And there's tech now that on the medical diagnostic font is getting more and more robust, as they're tailoring these large language models for it.
I think it's relatively easier to diagnose problems with real estate than it is people, and certainly less risky to use, like a combination of infrared and AI to look at what's going on behind a wall... So I don't think it'll take long before tech like that starts to hit the marketplace. I haven't seen any specifically focused on that, but just in talking here, we can think about all the applications of it. Someone else is working on it right now, I can almost guarantee that. And with the pace of AI advancements, we should see something hit the market soon. It's not going to come from us, but there's a lot of companies that are doing similar things out in the market today.
Slocomb Reed: Mark, I'd like to summarize the conversation that we've had here, taking the perspective - well, my perspective, but hopefully also the perspective of our real estate investor listener base. Frankly, when it comes to artificial intelligence in general, my emotions are pretty much just excitement. There are a lot of things that I do now that can be simplified, relatively affordably, all things considered, as AI enters into the space where I make a living. But the core principles of how I make a living are basically going to be unaffected by AI. I'm not under any threat, because the vast majority of what I do does not begin with a text input, or some other style of data input that generative AI requires in order to begin its operations.
That said, specific to HelloData, it sounds like there are some things that you all do that would save a lot of time and make a lot of - especially due diligence more efficient for a lot of real estate investors. I'm glad we had the opportunity to have this conversation, Mark, and for me to ask some really weird questions, trying to get some equally weird answers out of you. Generally though, I, and I believe our listeners as well, are optimistic about what it is that AI is going to do for us in our industry.
Mark Rutzen: Yeah.
Slocomb Reed: Last question, Mark... Where can people get in touch with you?
Mark Rutzen: Go to HelloData.ai. Check out our product offerings. And if you want to set up a demo, go to the Contact form, and submit, and we'll get something set up.
Slocomb Reed: That link is in the show notes. Mark, thank you. Best Ever listeners, thank you as well for tuning in. If you've gained value from this episode, please do subscribe to our show, leave us a five star review and share this episode with a friend you know we can add value to through our conversation today. Thank you, and have a Best Ever day.
This website, including the podcasts and other content herein, are made available by Joesta PF LLC solely for informational purposes. The information, statements, comments, views and opinions expressed in this website do not constitute and should not be construed as an offer to buy or sell any securities or to make or consider any investment or course of action. Neither Joe Fairless nor Joesta PF LLC are providing or undertaking to provide any financial, economic, legal, accounting, tax or other advice in or by virtue of this website. The information, statements, comments, views and opinions provided in this website are general in nature, and such information, statements, comments, views and opinions are not intended to be and should not be construed as the provision of investment advice by Joe Fairless or Joesta PF LLC to that listener or generally, and do not result in any listener being considered a client or customer of Joe Fairless or Joesta PF LLC.
The information, statements, comments, views, and opinions expressed or provided in this website (including by speakers who are not officers, employees, or agents of Joe Fairless or Joesta PF LLC) are not necessarily those of Joe Fairless or Joesta PF LLC, and may not be current. Neither Joe Fairless nor Joesta PF LLC make any representation or warranty as to the accuracy or completeness of any of the information, statements, comments, views or opinions contained in this website, and any liability therefor (including in respect of direct, indirect or consequential loss or damage of any kind whatsoever) is expressly disclaimed. Neither Joe Fairless nor Joesta PF LLC undertake any obligation whatsoever to provide any form of update, amendment, change or correction to any of the information, statements, comments, views or opinions set forth in this podcast.
No part of this podcast may, without Joesta PF LLC’s prior written consent, be reproduced, redistributed, published, copied or duplicated in any form, by any means.
Joe Fairless serves as director of investor relations with Ashcroft Capital, a real estate investment firm. Ashcroft Capital is not affiliated with Joesta PF LLC or this website, and is not responsible for any of the content herein.
The views and opinions expressed in this podcast are provided for informational purposes only, and should not be construed as an offer to buy or sell any securities or to make or consider any investment or course of action. For more information, go to www.bestevershow.com.