How We Use AI to Replace 100 Hours of Real Estate Work a Week

Original automatic captions. Names, numbers, and punctuation may contain transcription errors.

0:00 Hello and welcome back to the Owner
0:01 Meeting podcast hosted by Multif Family
0:02 Strategy. I'm Christian, your channel
0:04 host. Guys, we're going to have a ton of
0:06 new episodes that are in production
0:08 right now coming out. I'm super excited.
0:10 However, today we have one of my
0:12 favorite topics with one of our favorite
0:13 guests, Caleb Hmel. Caleb, welcome back
0:15 to the channel.
0:16 Thanks for having me,
0:17 dude. This is going to be a fun one. So,
0:18 AI and real estate. I've done some
0:20 videos on this. We use it a ton, but
0:22 this is going beyond the basics. You're
0:24 using this at a very high level. I've
0:26 developed some AI projects that are like
0:29 real time working in the mentorship
0:31 group making a huge difference in the
0:34 way that we look at real estate. I think
0:36 that you are at the front end of some of
0:38 the things that you can do in real
0:39 estate. We're talking capital raising,
0:41 relationship management, some really
0:44 cool tools that you are using right now
0:46 that I want to share with other people
0:49 just so they even just start thinking
0:51 about how to use AI differently. Caleb,
0:54 what projects are you currently working
0:56 on that you were most excited about?
0:59 Yeah. Gosh. I mean, they all have to do
1:00 with AI. Literally, everything I'm
1:02 excited about when my head hits the
1:03 pillow at night, last thing I think
1:04 about is AI. Uh, first thing right now
1:06 is first thing I wake up is AI. It's
1:08 didn't realize a lot of the monotonous
1:09 stuff we do. Um, as real estate
1:11 investors, real estate's kind of
1:12 obviously an older industry, especially
1:14 property management. There hasn't been
1:16 much integration for a long time besides
1:18 certain softwares. So, being able to
1:20 come in uh even a PM, something like
1:22 expansion, hey, how do you go find old
1:24 properties that maybe need new
1:25 management? Hey, how do you go find a
1:26 property management company, bad
1:28 website, maybe looking to sell, that
1:30 kind of thing, and really organizing
1:31 capital lists, going back through old
1:33 capital leads that may have slipped
1:34 through the cracks, generating follow-up
1:36 messages for these kinds of people
1:37 automatically with AI are some of the
1:39 main things we're working on on the real
1:40 estate side of this.
1:41 Oh, dude, I and I'm absolutely loving
1:43 it. How how are you finding as real
1:45 estate has been a relationship game and
1:47 both of our model has been build
1:49 relationships with brokers with owners.
1:51 How is AI changing the relationship game
1:55 for you? And is real estate still a
1:58 relationshipbased game?
1:59 Yeah, I don't think that's ever going to
2:00 change. I think no matter how much AI
2:02 gets involved, whatever, it's just going
2:04 to increase the load of scamminess and
2:07 really I think increase how much
2:08 relationships really matter. Like you
2:10 see the guys sending out mailers and all
2:12 this stuff. Okay, now they can automate
2:13 it where they don't even have to go to
2:15 the mailbox. Somebody's already had that
2:17 down, but now they don't even have to do
2:18 prospecting. Everything's dialed in for
2:20 them. And at the same time, all these
2:21 old owners who you actually want to
2:23 build relationships with are now going
2:24 to be even more turned off to all the
2:27 mass marketing and really wanting to go
2:28 on the relationship based. So, while AI
2:30 is a great way to get your foot in the
2:32 door, it doesn't replace building a
2:33 relationship by any stretch of the
2:34 imagination.
2:35 So, talk to me about the projects that
2:36 you are using right now that are on the,
2:38 you know, the cutting edge of AI and
2:40 real estate. what what things you know
2:43 so so we're going to assume like okay
2:44 we're still meeting with brokers still
2:46 meeting with owners we're still finding
2:47 deals the oldfashioned way right what
2:50 things does AI replace that has
2:54 significant time savings or competitive
2:56 advantage
2:57 oh I think it's the like the grunt work
2:59 and the prospecting are the number one
3:02 thing like even something on like cm
3:04 like maintainment or going through like
3:06 projects like making leads fall through
3:08 like leads that have fallen through the
3:09 cracks in the past to use AI instead of
3:12 praying that chat GPT instead of being a
3:14 yesman will give you a good answer. You
3:16 can actually use AI to sync with whether
3:18 it's your email, your SMS, whatever, to
3:20 actually go back through your messages,
3:22 pull these leads out, extract these
3:24 leads out that you may have dropped the
3:25 ball on, put them in a spreadsheet. You
3:27 could even program it to send automated
3:28 emails out for you. If you have certain
3:30 SMS software, you could send out
3:31 automated messages that are actually
3:33 tailored to the person. So, it's not,
3:35 "Hey, Caleb, just following up on this."
3:37 is hey Caleb last time we spoke you were
3:39 working on XYZ did you end up finishing
3:41 that project or how is that going way
3:43 more powerful on a follow-up basis as
3:45 well
3:46 this is really interesting so you said
3:48 instead of chat GBT what are you using
3:49 instead of Chad GBT that you
3:51 um I only use pretty much just claude in
3:53 anthropic at this point and I'm not some
3:55 AI soant by any stretch of the
3:57 imagination but how much progress I've
3:59 even been able to make with that stuff
4:01 it's been night and day versus any other
4:03 product we've used
4:04 this is what I love about AI AI will
4:07 tell you how to prompt it. You can start
4:09 with the end. So, you go, hey, here's my
4:11 end goal. What questions should I be
4:13 asking or how would I build this? And it
4:14 will actually teach you how to use it
4:16 better. It's the it's the beautiful
4:17 thing about AI is that you actually
4:19 don't need to be an AI genius. However,
4:20 some of the stuff that you're doing does
4:22 feel pretty genius. Can you give me some
4:24 depth into So, you mentioned you can
4:26 look for messages that you've missed or
4:28 people who, you know, talking like
4:30 investor capital. Let's use that as our
4:31 case study.
4:31 Yeah. So, we have a lot of people over
4:33 the last what six years now who have
4:36 reached out to me, reach out to you.
4:37 Hey, I want to invest in a deal. I want
4:39 to learn more. I want to I want to
4:40 participate. We now have a deal come up.
4:43 Instead of us looking through every
4:45 single past lead or every single past
4:48 form beyond just copying and pasting a
4:50 list into AI, how is it actually going
4:52 through and helping you sort this and
4:55 automate it?
4:55 That's a really good question. So, it
4:56 depends on the platform. It depends if
4:57 you're in your CRM, your iMes, or your
4:59 email. If it's on your iMes on anthropic
5:02 or claude, you can literally go into a
5:04 certain thing on into the co-work
5:05 section, enable something for SMS, you
5:09 give it the right permissions. This
5:10 thing can literally go through like
5:12 today I did this on my Mac for some
5:13 stuff and looking for some certain leads
5:14 in the past fall into the cracks.
5:17 Prompted it correctly said, "Hey, at the
5:18 end, create me a spreadsheet with all
5:19 the information X, Y, and Z." It went 5
5:22 minutes later. It has 597 people in
5:24 there already in that spreadsheet pulled
5:26 up with context on each and every single
5:29 person ready to go. And I can follow up
5:30 with those people. Now, I can either
5:32 have it go in and actually send these
5:34 messages out for me or I can go plug
5:36 this into a CRM or send those, hey, like
5:38 if they have their emails or we have
5:39 their emails anywhere. Hey, email this
5:41 person based off our last text thread
5:42 and follow up. Hey, I have an
5:44 opportunity because you said X, Y, and
5:46 Z.
5:46 Oh, this is really interesting. So, if
5:47 you're doing this as like a 506c, so if
5:49 you're doing like a syndicated thing
5:50 where you're allowed to market, you can
5:52 literally just write a awesome marketing
5:54 campaign and send it to everyone who's
5:56 expressed direct interest in investing
5:58 with you. And you how long does that
6:01 take AI to pull? I'm assuming this is
6:02 pretty much instant
6:03 like going through all the messages with
6:05 all the approvals. Five minutes tops.
6:08 So, in five minutes, so you're getting
6:09 an average of like a 100red leads a
6:11 minute going through
6:14 not found. If I had all the permissions
6:16 enabled, it would have been like a
6:17 minute or less. So yeah, it's pretty
6:18 much instantaneous on actually going
6:20 through this stuff, let alone even like
6:22 talk about underwriting on a deal or
6:23 talking about, hey, broker sends you a
6:25 deal. You create something with cloud
6:27 code that actually plugs straight into
6:28 your calculator. It runs all the things
6:30 for you. It spits out with underwriting
6:32 taking you hours. That takes the AI what
6:35 30 minutes or 30 seconds to a minute
6:37 maybe tops and just breezes through
6:39 everything.
6:39 See, and that that is absolutely
6:41 amazing. It's a really fun way to do
6:43 this. So, um, you know, again, you can
6:45 use that to market. The other thing you
6:47 can do is you can take that list,
6:48 everyone who's expressed interest in
6:49 investing. Uh, let's assume that you're
6:51 not allowed to market directly to them.
6:52 You can host events. You can rekindle
6:55 relationships. You you can put together
6:56 a list of like, hey, who do I have an
6:58 existing relationship with? I've had
7:00 multiple points of contact who I could
7:03 reasonably reach out to. And you can put
7:05 together a actual hit list. And you can
7:06 talk to these people and you can go
7:08 through the list and say, "Hey, do they
7:10 qualify? Do they not?" like you you can
7:12 make this instead of looking through the
7:14 years roughly how many contacts do we
7:17 have in our in our backend system for
7:19 people who have reached out. We've had
7:21 85 85 or 90. Okay.
7:23 Okay. So, so around 100,000 people. We
7:26 can bring that list down to the top 500
7:28 people that we need to speak to. And now
7:30 I can look through a list of 500 people
7:32 and say, "Hey, who's appropriate with
7:33 the regulations of this deal and who's
7:35 not?" I can go and make informed
7:37 decisions on a list of 500 people
7:40 instead of informed decisions on a list
7:42 of 90 to 100,000 people. That's that's
7:45 an insane difference.
7:46 Yeah. And maybe you spend $10 like like
7:48 you could use there's different ones
7:49 with APIs because that integrates with
7:51 those too. Grock, you could use OpenAI,
7:54 Anthropic, whatever in the back end.
7:56 Like the cheapest one, Grock, probably
7:57 cost for that many people. I I don't
8:00 have a ballpark, but I assume anywhere
8:01 from under $100 to screen 90,000
8:04 contacts. Like it's like every time like
8:07 like I was looking at something I had
8:08 ran recently, it was 1.1 million tokens
8:11 and that equated to 28 cents. So it's
8:13 like if you could actually get this down
8:15 where you have your system down. I don't
8:17 have a finite number of how much it
8:18 would cost. depends how much data and
8:20 like how much into the weeds you're
8:21 going with that so to speak. But I mean
8:23 you're talking like okay you spend let's
8:25 say even a hundred bucks in API tokens
8:27 and it saves you what three weeks worth
8:30 of work.
8:30 Well and I remember what we used to do
8:32 one or two years ago together when we
8:34 were you and I would both we we've gone
8:35 through a few but usually we'd have two
8:37 VAS on staff. You'd have one iPad one
8:39 for our various tasks and it's usually
8:41 like $800 to $1,000 a month. So you're
8:43 paying them like $200 a week. All of
8:45 these functions now take way less time
8:48 and are way better. I mean, it's kind of
8:50 way more accurate virtual assistant.
8:52 It's a it's really kind of probably
8:53 going to make that job a lot harder. But
8:56 as a entrepreneur doing massive amounts
8:59 of tasks for 28 cents,
9:01 crazy. Why wouldn't you? It's like the
9:02 time savings and efficiencies there. Or
9:05 you can literally build something like
9:07 I'm working on this right now. Like
9:08 literally goes through my email, goes
9:09 through like I mean you get more emails
9:10 than I do. You can literally build a bot
9:12 that will just go through your entire
9:13 email and put the top 10 most important
9:15 or urgent emails that are actually from
9:16 people and not just spams at the top of
9:18 your to-do list. And then instead of you
9:20 having to go sort through hundreds of
9:21 thousands of messages, be like, "Okay,
9:23 cool. I need to bang these out this
9:24 evening. Awesome. I'll get it done." And
9:26 you could literally just have exactly
9:27 what you need to do, you could have it
9:29 respond. Or another thing we do is on
9:31 the property management front, on the
9:32 leasing front, we have a lot of stuff on
9:34 Facebook Marketplace. You could
9:35 literally create a bot even on cloud
9:37 co-work. You just program it, tell it
9:38 what to do, it'll go on Facebook
9:39 marketplace and just respond to the
9:41 leads. I have a project I need you to
9:43 help me with after this call. That
9:44 [laughter] is so I need some help on
9:47 some Washington state properties. Um, if
9:49 you can build that bot for me on that,
9:51 that is that that that would uh I would
9:52 love to see how you're doing that.
9:54 But these are the ideas. Like if you
9:56 think it for the most part, you can
9:57 prompt AI and like, hey, how would I
9:58 build this and what have other people
10:00 done similar projects? What should I be
10:02 asking? Like you can write down your
10:04 questions, your end goal, and have it
10:05 tell you what to do to try to optimize
10:07 it. You follow these strategies. It just
10:09 works really well. One thing that I mean
10:11 this is insane. Uh mentorship, my goal,
10:14 every single person who signs up for the
10:16 mentorship buys property. Nothing else
10:17 really at the end of the day matters.
10:19 You can have the best courses, the
10:20 funnest community, the biggest
10:21 community. If you don't have active
10:23 investors who are getting their desired
10:25 result, like if you sign up, you need to
10:26 buy real estate. So I want you to feel
10:28 supported. for us to better support
10:30 people. We've actually rolled out some
10:32 AI things that have been awesome. A
10:33 custom GPT, which I just asked Chat GPT
10:36 to help me make. Took me 2 days of just
10:38 like non-stop prompting, 8,000 lines of
10:40 prompt, going through our methods,
10:42 everything I've filmed on YouTube,
10:44 everything that I've filmed in the
10:45 course. It downloaded the transcripts of
10:47 all of this, based its personality off
10:50 of that, asked a ton of questions and
10:53 situational things. It now recommends
10:55 deals exa analyze them exactly the way I
10:58 would except it has built-in
10:59 calculators. So if you're a student, of
11:02 course you have one-on-one support. Of
11:04 course you have coaching. In addition to
11:06 that, you can instantly you have this
11:09 big math problem and trying to figure
11:10 this out. You can take that while you're
11:12 waiting to book a call with me or you're
11:14 just like, "Hey, I want the answer in
11:16 two seconds." You go to GBT, you can
11:18 upload the actual OM for your deal and
11:21 prompt it and it will actually tell you
11:22 what to prompt. They'll say, "Hey,
11:23 Christian recommends these parameters."
11:26 A bank's probably going to look at the
11:27 deal like this based on your books. Your
11:29 offer price range is here to here. Now,
11:33 you have inputs and you can test all of
11:35 these theories in the calculator that's
11:37 provided in the course. By the time my
11:40 students hop on a call with me, the
11:42 difference is not that they get less
11:44 coaching. It's when they hop on the same
11:46 call they were scheduled for, they're 10
11:49 questions farther. Their deal is almost
11:51 ready to go. And instead of me filling
11:53 in like, hey, this is what you need to
11:55 factor for for landscaping or hey, you
11:57 forgot taxes are going to reassess after
11:58 you close, they're coming in with a
12:00 fully underwritten deal and I'm checking
12:04 their assumptions and we're working on
12:06 writing the offer. The difference in
12:08 just in the coaching space, it's yes,
12:12 it's saving me time, but more
12:13 importantly, the same amount of actions
12:16 with students have 10 times the output.
12:18 We just came out with a app. It's it's
12:20 it's my virtual mind. And it literally
12:22 it's like a you can call it it actually
12:24 has my voice and you can go through
12:27 downloaded I think it's 150 million
12:30 words that I've said on YouTube on the
12:33 course
12:34 that's a few
12:35 on other people's podcasts I actually
12:37 plugged in like other podcasts too is
12:39 everything I've said over five years um
12:41 and same thing multiple days of
12:44 interviewing me it got down to details
12:46 of like hey when you bought the Robin
12:47 Hood Village Resort you mentioned uh
12:49 that you have a leasing strategy of 9595
12:52 in property management for multif
12:53 family. Are there similar parameters
12:56 that you look for in hospitality? Like I
12:58 mean it went down to like the level of
13:00 like
13:02 specific questions my mentees if I'm on
13:05 a call with another mentee you can call
13:07 basically my mind and ask it any
13:10 operating question.
13:12 Yeah that's powerful.
13:13 Those both took two full days. I mean I
13:16 had to take the day off. So we're
13:18 talking four days of my life. My mind is
13:20 now in chat GPT on a math and analytics
13:24 perspective. You can ask technical
13:25 questions and combine it with the search
13:28 power of GPT or you can ask it like
13:31 functional or operating questions that
13:33 only I would have the answer to and
13:35 those are also downloaded in another.
13:38 They're both available in our course.
13:39 You can talk to me, you can message, you
13:42 can calculate, you can get deal advice.
13:44 It's lifechanging.
13:46 four days. Four days of work. That would
13:49 take
13:50 years to train an employee to be able to
13:53 do that.
13:55 Unbelievable the time savings and value
13:57 that it's able to create.
13:59 This isn't an ad for the for the
14:00 mentorship. This is just a straight up
14:01 plug for like, hey, if you're not
14:02 thinking about these things for how
14:05 you're associate, like how you're
14:06 communicating with your clients, how
14:08 you're building your real estate and b
14:10 business. I'm looking at building a very
14:12 similar AI for my employees of multiple
14:15 companies where they can go in and like,
14:17 hey, when you're stuck on this,
14:18 sometimes I'm coaching, sometimes I'm in
14:20 a meeting, sometimes Caleb and I are in
14:22 a closing, we're signing docs. How cool
14:25 to be as a property management company
14:26 to have all of the data of our
14:28 operations, our SOPs, our best
14:31 practices, and you can literally come in
14:33 and just talk to your boss when they're
14:36 busy.
14:37 Well, talk about saving like employee
14:38 costs as well. Like you have all these
14:40 employees like like executive
14:41 assistants. I'm like, how do you guys
14:43 see any future in the jobs you guys are
14:45 at the moment? Not saying our own people
14:47 internally, just generally. It's like,
14:48 okay, you could build an AI bot that
14:50 does absolutely all of this. And
14:51 granted, you want to put all this stuff
14:52 out there everywhere on the internet
14:53 with all your pass keys to your back
14:54 end, your root passwords. But at the
14:56 same time, you can create something that
14:57 runs locally on your machine that is
15:00 password protected and everything that
15:01 can literally just do all the tasks you
15:03 want it to. It's customuilt for you. You
15:04 don't need to go try to, hey, let me
15:06 grab this software for $20 a month. Let
15:07 me get this one for 200. Let me grab
15:09 this one for 300. Hopefully we can spice
15:10 it together and be an employee. You can
15:12 just build it. Granted, it takes some
15:13 finite work and there going to be some
15:15 errors and some troubleshooting, but you
15:16 can basically build an employee for the
15:18 most part.
15:19 Yeah. Well, one, you can build a lot of
15:20 employees. And what I've done is, you
15:22 know, for like a media editing, for
15:23 example, AI has a lot of software that
15:25 can do it really quick, but it still
15:27 doesn't quite replace having a person
15:29 and a creative director who can come in
15:30 and organize it, come up with the
15:32 strategy, implement the strategy, get
15:34 high quality edits, or just the person
15:36 to interface with me of like, hey, can
15:38 you film this, this, and this? I need
15:40 this clip. It doesn't quite replace uh
15:43 media coordinator. What it does do is it
15:46 replaces every single employee
15:48 underneath them. I can build out an
15:50 entire media company
15:52 with one person. Our production I can
15:55 get. And now I don't want to blow up my
15:57 audience, so I don't do this, but I can
15:59 easily produce 10 short form pieces of
16:01 content for every platform every single
16:03 day. Rip it out there. Do long form
16:06 YouTube.
16:08 We can streamline all of these things.
16:10 But your entire production company, you
16:13 need someone who is a skilled
16:14 photographer and a good creative mind.
16:17 And if you have those two things and a
16:18 decent editor, if they have those three
16:20 skills, AI can do every other piece of
16:23 that thing. You can have a 10person
16:24 company with one employee.
16:26 Yeah, it's it's life-changing as far as
16:28 expansion, as far as systems tracking. I
16:30 mean, even like I think there's a cloud
16:32 extension or I don't know, I think it's
16:33 cloud. I haven't used it, but just plug
16:35 straight into your Excel sheets. Like
16:37 you can actually use something where it
16:38 integrates with Excel and like whatever
16:39 you needed to do run whatever data you
16:41 can pull like you can create custom web
16:43 hooks like literally you could create
16:44 your own web hook based off your own
16:46 domain and use a web hook function
16:48 whether it's AP or something else to
16:49 actually integrate with your own website
16:50 and send web hooks. It's like there's so
16:52 much you can do as far as actually with
16:54 AI and what that looks like building it
16:56 out. Oh so cool. Well and uh so so one
16:59 thing that I don't quite understand yet
17:00 you're farther than me on this AI stuff.
17:02 Uh you mentioned having like a co-worker
17:04 like co-working in Claude and then I've
17:07 been playing with AI agents. Is
17:09 co-working and agents are those similar?
17:10 Are those completely different?
17:12 What are those two things and how are
17:13 you using them?
17:15 Yeah. So co-working is like tasks. So
17:17 like co-working on cloud like let's say
17:19 for that Facebook thing this morning
17:20 with all the on all the Facebook leads.
17:22 You're like hey I need you to go on
17:23 Facebook and I need you to respond to
17:24 these people in this way with all the
17:25 unread messages and if there's already a
17:27 conversation go pick up where it left
17:28 off. You can basically do that and
17:30 there's an integration where it can go
17:31 on Google Chrome and from Google Chrome
17:33 it can then lo if you have the Facebook
17:35 login up it'll log into your Facebook go
17:37 on your messages tab on Facebook
17:39 marketplace and tackle all those. So
17:40 co-working is more of an assistant but
17:42 you have to constantly prompt it where
17:43 claude code is different is you're
17:45 actually building websites you're
17:46 actually building domains you're
17:47 actually building backend infrastructure
17:48 and it's going through your code and
17:49 optimizing it and fixing things. So
17:51 there's chat with claude, there's claude
17:53 co-work and there's claude code are the
17:55 three different ones. It basically wipes
17:56 out the whole open claw thing a few
17:58 weeks ago. That basically has taken a
18:00 backseat cuz cloud basically came out
18:01 with its own better version.
18:03 Interesting. So does that completely
18:04 replace the AI agents that that do
18:06 thing? You create your own agents who
18:08 are doing all these things. Is the
18:08 co-working basically its own agent?
18:11 Yes and no. It it depends like if you
18:13 want to have an agent doing it like
18:14 automatically on their own. Maybe you
18:15 create it like with notebook LM and you
18:17 actually create an agent that way.
18:19 But if you want to like and just say I
18:20 need you to go do this task like XYZ
18:22 update these people's data in XYZ
18:24 spreadsheet. It can just go do that for
18:26 you. And you can actually generate
18:27 another co-working session where you
18:29 let's say you have, hey, this person's
18:30 doing this, this person's on Facebook,
18:32 this person's on data, this person's on
18:34 expansion. You can have three things
18:35 doing different prompts. It'll let you
18:37 know when you need to go access
18:38 permissions for something, but you can
18:40 have three different things firing at
18:41 the same exact time, all doing different
18:43 tasks.
18:44 So, you can basically build your
18:45 orchard. So, if I'm building this as a
18:47 company, and by the way, you you people
18:48 who are really good at AI are probably
18:49 like, Christian, you're you're you're an
18:51 idiot. We know all this stuff.
18:52 Oh, same here. Like there's some people
18:53 who would so it like so far.
18:55 I I'm I'm super curious though. So can
18:57 if if I'm doing this the way that I
18:58 build a company is I I'd actually have
19:00 an org chart. Can I create an agent who
19:02 manages or or or a co-worker, however
19:05 this works. Like can I create a bot
19:07 essentially create an employee who I
19:09 instruct and then they instruct the
19:12 employees underneath them? Like can I
19:13 interface with one
19:14 feeas feasibly? Yes. It should be
19:16 doable. I haven't gone that far yet is
19:18 actually build like a psych. That would
19:20 probably be on how I would think about
19:22 it and I'm probably wrong. It's probably
19:23 some AI nerd in this like Caleb such an
19:25 idiot, but you'd probably create like a
19:27 headbot actually with notebook LM is
19:29 what I is the best one to use and then
19:31 actually create sub ones under them and
19:32 then ob you make your own domain so you
19:34 can manage and actually see progress
19:35 through all of them is probably how
19:37 you'd feasibly do that.
19:38 So you have a high school degree, you
19:41 have 400 rentals today. Are you four or
19:43 500? I forget. Did the 80 units count
19:45 425?
19:46 Yeah, I'm 413 right now.
19:48 413. Okay. 413. You have 413 rental
19:52 units. You run the property management
19:54 company. Well, you have been helping
19:56 with Facebook, but it actually might not
19:57 even be Caleb anymore. And by the end of
19:59 next [laughter] week, how long did it
20:00 take you to learn this to the level that
20:02 you're at? Like from the time that
20:03 you're like, "Hey, I'm going to learn
20:05 how to implement AI in my business to
20:08 now you're 23 years old again, high
20:10 school education,
20:12 very busy individual, insanely busy
20:14 individual. You have a full-time job and
20:16 you run multiple companies. How long did
20:18 this take you with all of that on your
20:19 plate?
20:20 I would be lying if it said it was if I
20:21 said it was instant. There's definitely
20:23 like the premise of AI was fairly quick,
20:25 but actually learning how to build these
20:26 things efficiently has taken time and
20:27 I'm still not where I want to be.
20:29 Obviously, like there's some things
20:30 where today and yesterday has pretty
20:32 much been troubleshooting this one
20:33 domain I built to analyze old
20:35 conversations, score them on a scale of
20:37 1 to 10, then send a custom follow-up
20:38 message.
20:39 Like for something like that, it takes
20:41 extended time, but overall it didn't
20:42 take that long at all. like best guess
20:45 when did you we obviously been using AI
20:47 for I mean it's been about like a year
20:49 it hasn't even been that long when did
20:50 you really start building in AI beyond
20:54 you know beyond hey I'm using chat GBT
20:57 to ask questions like when did we
20:58 actually start implementing AI in the
21:01 business maybe a week and a half two
21:03 weeks or so like to actually going in
21:05 depth on it
21:07 I just want to point out how crazy that
21:09 is you're talking about building
21:10 basically an AI company of AI employees.
21:15 You've done coding. You've built apps
21:16 for me. Uh we've done multiple apps for
21:19 my company for integrations and this is
21:23 like a few weekek project.
21:25 Uh this is amazing how fast uh learning
21:27 happens. Uh really fun to see where it's
21:30 going. But this is uh the goal of
21:31 today's podcast is talking like it's the
21:33 advanced part of AI and real estate. Can
21:36 you create agents to take out task? What
21:38 things should you outsource and what
21:41 things should stay as relations? The way
21:44 that I have been doing this and I I
21:46 think this works uh really well. This is
21:48 a friend of mine actually back when we
21:50 were doing the VA strategy before all
21:52 this AI stuff was available. One thing
21:55 they would do every day and I think this
21:56 is a practical drill. Write down every
21:58 single task you do for a few days. And
21:59 is it is a pain like every time you do
22:01 stuff just write it down. Then you're
22:03 going to color code it as things that
22:05 like absolutely do not need to be you.
22:07 Things that sometimes you should have
22:09 some input on and things that like
22:10 always need to just be you. If you have
22:13 a color-coded list, everything we'll
22:15 we'll call red our magic color for like,
22:18 hey, I shouldn't be doing this task. And
22:20 if you're on the fence, ask yourself,
22:21 does this task actually make money? If
22:23 you're an entrepreneur, that's pretty
22:25 much your main.
22:26 Yeah, that's pretty much
22:28 doing this task make more money now.
22:31 Does it make more money later and I
22:33 probably should have input or is this a
22:35 just like is this just a non-money task?
22:38 Those non-money tasks, those things that
22:40 you have to get done but probably don't
22:42 need your input, every single one of
22:44 those should be offloaded either to an
22:46 employee or an AI bot. And now the
22:48 question in the past it was like does it
22:50 need to be a US employee or can this be
22:53 a virtual assistant? today. The question
22:55 is, do I need an employee or can we
22:59 build an AI bot to do this with this set
23:02 objective by the end of today? The the
23:05 biggest thing that Caleb does, and this
23:06 is I think this is most of the success
23:08 that Caleb and I have had in our
23:10 business, instant implementation of
23:12 ideas. Someone has an idea on a call, by
23:15 the end of the day, it's a thing. Hey,
23:17 we should build a AI bot for this. By
23:20 around 9:30 tonight, we'll probably have
23:21 an AI bot for it. Hey, I have a great
23:24 marketing idea. We should try this ad
23:25 campaign. It'll be filmed, edited, in
23:27 production, and we'll have beta tested
23:29 four different hooks by the end of the
23:32 day. You can instantly analyze this. You
23:35 could listen to this podcast and you can
23:36 put this in your business by the end of
23:38 the week in multiple levels.
23:40 Absolutely.
23:42 Super exciting stuff. Caleb, any closing
23:44 thoughts on AI and real estate and
23:45 technology? Anything that you have
23:46 learned that people absolutely need to
23:49 be doing right now?
23:50 Um, there's some part of it actually
23:53 building this stuff. It's kind of like
23:54 real estate usually takes like not more
23:56 expensive because I mean it's very
23:57 minimal on like the actual costs of
24:00 this. Closing thoughts. I've I've
24:01 started doing this making sure you
24:03 actually ask like, "Hey, does this cost
24:05 tokens? Does this cost money to your
24:07 AI?" Because there's some people I've
24:08 seen on the internet that rack up like
24:10 if you're using Quad without a paid
24:12 monthly plan, you're going to rack up a
24:14 very hefty bill. If you're using paper,
24:16 like pay per use, just get a regular
24:18 client, but actually make sure you're
24:19 tracking this stuff because there's a
24:20 lot like you're like, how does somebody
24:21 make money off this? How does this
24:22 actually run? How does this fire? Like,
24:24 why does this make sense? And actually
24:25 looking at that stuff's pretty damn
24:27 important. So, that's the big one. And
24:29 building this out, you're it's probably
24:30 going to take a lot of debugging
24:31 eventually. You're going to tell it to
24:32 do something. You're going to get it
24:33 there and then you're on the finish
24:35 line. You're like, I completely missed
24:36 how to do that part. I just need and you
24:38 have to go back and rebug and redo the
24:39 whole thing. Doesn't take forever, but
24:41 it's going to take longer, especially to
24:42 start.
24:42 Oh, that's great advice. Fantastic.
24:44 ending point on this too. Yeah. It's
24:46 like, hey, if you're doing this, uh, you
24:48 know what can tell you the most cost
24:49 effective way to run it? AI. Verify once
24:51 you build the system.
24:52 Absolutely.
24:53 Did did I build this optimately?
24:55 Optimally. Was there a better way to run
24:56 this? I actually ran an app recently
24:59 where in that app I was like I'd
25:01 finished the whole thing. I'm like, this
25:03 is actually pretty cool. Like is there a
25:04 better way to run this? And it's like,
25:05 wait a second. This code is really
25:07 inefficient. You should add this and
25:09 this. And it literally it it tried to
25:11 patch it. It kept crashing itself and
25:14 then it said, "Hey, I'm going to run my
25:16 leave your computer on. I'm going to
25:19 recreate this entire thing in an optimal
25:22 way over the next couple hours." And
25:24 when I woke up in the morning, it had
25:26 already not just built, but it had
25:28 tested everything that I'd previously
25:30 tested, found the errors, restarted, and
25:33 it's like, "Hey, this is your optimal
25:34 amount of code for the most efficient
25:37 version of the app that delivers
25:39 everything that you wanted." And I've
25:42 already tested every feature to make
25:45 sure that you didn't have the app
25:46 crashing the way that it was uh prior.
25:49 Literally wrote itself over the
25:52 beautiful thing. So even when you're
25:54 checking these things, all you have to
25:56 do is just ask your app you designed it
25:58 on the question, did I build this right?
26:02 Go in and tell you,
26:04 you did a good job. However, I'm going
26:06 to build it for you again one more time
26:07 and build it better. Yeah. Or you can
26:09 ask, "Hey, can I build it cheaper? Can I
26:10 do this? Can you You can just ask it
26:12 questions. It'll spit back at you. It
26:15 makes it a lot easier."
26:16 Yeah. Well, I hope this was helpful. If
26:18 you guys are listening to this, you're
26:19 bored out of your mind, then, you know,
26:20 pick literally any other episode about
26:22 real estate on this channel or business.
26:25 Uh, updates in the field. Again, we are
26:26 going to bring a ton of new speakers
26:30 onto the podcast. We took a little
26:31 hiatus as we opened up Abalene. We've
26:33 been building out the PM company. We
26:35 started this year crazy crazy crazy
26:39 heavy on acquisition. So buckle in. This
26:42 podcast is about to get a ton of new
26:44 episodes, ton of content. So save this,
26:47 comment below, give us a fivestar rating
26:49 on whatever app that you're on. Help us
26:51 spread the word and we appreciate you
26:53 listening. This is the Owner Meeting
26:54 podcast hosted by Multif Family
26:55 Strategy. I'm Christian. This is Caleb
26:57 Hmel. I'll see you on the next episode.