How FBS Is Bringing AI Directly Into the MLS Experience with Michael Wurzer

Innovation is changing the way real estate professionals work with MLS data. As AI becomes part of everyday business, agents are finding new ways to access information faster, work more efficiently, and spend less time on repetitive tasks.

Matt Fowler and Allan Nielsen sat down with Michael Wurzer, President and CEO of FBS, to talk about how AI is making its way into the MLS experience. Michael shared why FBS invested in Model Context Protocol (MCP) technology and how it creates a more natural way for agents to interact with MLS data.

The conversation touched on real-world use cases, the importance of security and privacy, and how AI can help simplify common workflows. They also discussed the API infrastructure that makes these advancements possible, along with a preview of Scout, FBS's upcoming AI assistant.

As AI becomes more deeply integrated into MLS platforms, agents will have more tools to access information, manage listings, and get work done faster. Tune in to hear Matt, Allan, and Michael discuss what these changes mean for real estate professionals and where MLS technology is headed next.

Specifically, this episode highlights the following themes:

• Connecting AI tools directly to MLS data through MCP technology
• Balancing innovation with security, privacy, and data integrity
• Streamlining listing management with AI-assisted workflows and automation

Links from this episode:
Know more about Michael Wurzer: https://www.linkedin.com/in/michael-wurzer-16b7a41
Learn more about FBS: https://wearefbs.com
Learn more about Doorify: https://doorifymls.com

1ae5b43598204883b524f061d4880e6d10fca88c (for podfollow.com)

Transcript

Matt Fowler [00:00:08]:
Welcome, everybody. Thanks for joining us today. We have about a half an hour to let FBS tell us what's coming down the line and what FBS subscribers should be excited about. I think they're first to market here with enterprise level AI data access. We're super excited about it. Alan and I have been using it for a few months now and it's a real game changer. So, Alan, take it away and tell us who we've got today.

Allan Nielsen [00:00:33]:
Yeah, thank you. Thank you, Matt. And yeah, we are very excited. Today we have Michael Warser with FPS. So welcome, Michael. Glad to have you on. Thanks, Alan. And like Matt was saying, I think this is really our first podcast where we really get to talk about McP servers and AI in the context of one of our platforms.

Allan Nielsen [00:00:55]:
But before we jump into all of that, Michael, introduce yourself and let everybody know who you are and let's start there.

Michael Wurzer [00:01:05]:
Yeah, thanks so much for having me. This is a good conversation and we're pleased to be invited. I'm Mike Werzer. I'm the president and CEO of FBS, which makes the Flex, MLS and Spark platforms. We service about 160 MLS mostly in the US but a little bit internationally as well now. And we've been doing it for a very long time. So we've been in this industry for around 40 years through lots of technological changes. But this is definitely the most exciting technological change that we've been a part of.

Michael Wurzer [00:01:40]:
And it's going to be a fun conversation.

Allan Nielsen [00:01:44]:
It's exciting to some. It's a little scary every now and then, right? But yeah, so like you said, you know, FBS and FlexMLS, all of our subscribers have access to FlexMLS through listing managers. And then we have a good amount of subscribers who are also using FlexMLs for their front end of choice. And I think as we go through the conversation today, I think our subscribers will find new reasons to subscribe to FlexMLS. But let's jump in and start talking about MCP servers. So tell us. We just launched this with Dorify. It's now available.

Allan Nielsen [00:02:23]:
It sits on the Dorify dashboard. It is under our Labs icon and easily accessible with instructions. Can also access it directly from FlexMLs. But tell us a little bit about what drove you guys to invest in the MCP and bring that to the market.

Michael Wurzer [00:02:41]:
Yeah, so it really goes back about a year and Matt referenced it. We had our CEO. Every year we have a meeting of our customers that are thinking strategically. And about a year ago we had that annual session, Anthropic and other companies had just put forward this standard called mcp, which stands for model context protocol. And so, you know, those are foreign words, probably outside of AI land, but really all it is is a standard way to connect a data source, like an MLS system, to these AI models in a way that doesn't require those models to know everything about how to connect to that particular data source. So it's just as. It's like a standard plugin in a way. When I saw that, I was like, hey, that could be really cool.

Michael Wurzer [00:03:39]:
For accessing our APIs, we built Spark, which is our set of APIs over the last 15 years or so, primarily to power mobile applications. Back when the iPhone came out, restful APIs were an important part of how to make those, the data show up on that phone. And so the same APIs now can make that data show up in your AI application. And that's essentially what MCP does. And so do what. I mean, the pace, as Matt said, the pace at which all this stuff moves is crazy because they Anthropic, which makes the Claude set of AI tools and models, they were the first to kind of publish this specification. But then quickly thereafter, OpenAI adopted it and many, many, many other companies. And that's always a sign that you got to look for, is this going to be a standard that lasts and how is it going to evolve? But it was pretty clear very early on that that was going to be something worth investing in.

Michael Wurzer [00:04:43]:
And so we did. And by July of last summer, we had essentially a proof of concept up. We could connect it to whether it was ChatGPT or Claude or any of the other chat bots and see what it could do. And we started having people playing with the customers, playing with it throughout the fall. And at that point we were really committed to investing in it fully. And so that led us to the launch state where we are today. And it is definitely part of a much bigger vision and we've learned a lot in that time period. But it's exciting to have stuff out in the wild because we'll start to see how people, you know, what are the questions people have that they really want to ask and how can that MLS data be tuned to really answer those questions with accuracy and, and concreteness? I have a lot more I could say, but I want to make sure you're letting you ask questions.

Allan Nielsen [00:05:40]:
No, it's perfect. It's perfect. And you know, there are so many terms that are being used in this space now, right? MCP API restful, all of those things. Right. So for our typical subscriber who is trying to get into AI, and this obviously opens up a brand new way for them to speak to the mls, right, and obtain data. But how would you describe what this will do for an agent? What's the use case here?

Michael Wurzer [00:06:15]:
Yeah, no, and that's a really good question. Let's just Talk about the APIs for a second because I think that the easiest way to understand the API is by thinking about the day to day use of the MLS system. And so when you log into the MLS system, it knows who you are first of all, right. Because you've authenticated. And so then it knows what, what your listings are, which ones are yours, which ones can you edit, which ones are in your office, in your company. All of that permissioning is something that has to be built into these APIs, so that when you're logging in and getting user interface, whether it's on your phone or in a browser, it knows what, what to show you, including all your things like your settings and your preferences, what you know, what do I like my dashboard to look like? You know, all of that stuff is really served by these APIs. And so if you think about what can I do with it, it should be lots of things that you can do inside the MLS system. I think that's the easiest way to think about it.

Michael Wurzer [00:07:19]:
Now, when we first put the MCP server out for public, we have not turned on any of the APIs that allow rights yet. We do intend to do that for sure, but just from a testing and security perspective, making sure that we can do that securely. We know the APIs do it securely, but we really want to make sure that these LLMs can use the APIs in a secure way.

Allan Nielsen [00:07:50]:
And Michael, just for clarity, so what you're talking about here is the ability to add data or modify data through the API or through mcp.

Michael Wurzer [00:07:58]:
Yeah, yeah. So if you go in as a user on a day to day basis, one of the things you're doing is creating searches, right? And as you create those searches, the system saves them, you know, as a draft, or you can save them as a, as a search that you want to rerun later. And that's writing to it, or you create a contact, or you update your listing, or you add a new listing. All of those things are writing data back through the APIs. And so all of that is capable of being done. When we did our beta testing, we had all the early on, all the right stuff opened up and it was actually working really quite well. So we definitely intend to do that. But this first real launch to broad sets of users, we're just letting it read information so it can read all your contacts.

Michael Wurzer [00:08:45]:
I think one of the coolest use cases is, hey, tell me who are my most recent contacts or most active contacts and give me advice on how to interact with them. Or write a plain for how I should interact and touch base with my clients, what are they going to be most interested in? And the LLM will read your contacts, look at their searches, look at what they're favoriting, and it will analyze all of that for you. So I think that's a really cool example. Others are market analytics. One of the APIs we have is a market statistics API. And so it can go and compare different markets, draw charts and graphs and all sorts of things. And all you have to do is just ask it, you know, hey, I want to compare this market market to that market or other queries you can ask. So I think statistics and, and certainly things like your listings, you can say, hey, what's the listing with the most days on market that I have right now? And give me advice about how I can, you know, improve that listing.

Michael Wurzer [00:09:46]:
And the LLM will do that whether it makes sense or not. Who knows? You know, I mean these things are super creative and super inventive and they will always give you an answer. But it is, I think an important thing for everybody to keep in mind. You must be vigilant about what it's telling you and that's again why we're being a little careful about what we're exposing. And I should have said actually this is an important distinction. With this initial launch of just the MCP server, we actually had to pull back the contacts one. So I need to take that back a bit. And the reason is this, that has personally identifiable information in it or pii.

Michael Wurzer [00:10:29]:
And so for privacy reasons we decided for now not to put that into the MCP server because it's being connected to a third party product like ChatGPT that we don't control. So then the data is flowing out of there to these other sources and we don't control that. And when it has something like privacy involved, that's a potential concern. So we don't have that currently, but we are going to be turning that on. We have a new product coming that's a full chatbot and all of these tools will be available in there because we control end to end the experience and can more guarantee the security of that information.

Allan Nielsen [00:11:11]:
I Appreciate you bringing up the privacy aspect of this because that's obviously something that for one Dorify is concerned with and any responsible uses of AI or any tools, that's something they're concerned with as well. So it's good to hear that you guys are being really careful there. So I want to go back to the writing aspect of this and the ability to write in. So does this mean that eventually we can see listing changes, price list changes, those type things being done through the LLM?

Michael Wurzer [00:11:43]:
Yeah, for sure, 100% Yep. I think that that will there. We already have APIs that allow for price changes and open house information and other things to be pictures and documents and other things like that to be up loaded through the APIs so we can turn that stuff on today. Again we're waiting on all rights through the third party chat GPTs of the world for a little bit but all of that's there and we fully anticipate having a full addit API that these LLMs can use so that an agent can do that from whatever device they ultimately want. But for sure within the Flex MLS environment they'll be able to, if they want to add a listing through that, they'll be able to do that. But most importantly I think edit a listing. Chat boxes are not great interfaces for adding a listing today. I think they'll get better and better as more auto population occurs and all of these things happen.

Michael Wurzer [00:12:43]:
There's a lot lot that can be done there but every field needs to be edited at some point in time and being able to do that on the fly with your voice or otherwise, I think is is a very important use case for all users.

Matt Fowler [00:12:57]:
I want to just jump in real quick, this is Matt and give the subscribers a quick path of how they can use this. Alan, maybe you can cover that if you're an FPS user. It's there already in Settings. You have to go turn it on. There's a little toggle if you're not user and we've got about 10,000 Paragon users and about 2,500 FPS users currently. And the 10,000 Paragon users are mostly still using Paragon because we used it in the beginning before we offered a choice. But they are using it because that's where their saved searches are and where all their people are and where all of the all their contacts and it's a lot people who have dozens or hundreds of those. It's a lot to move over.

Matt Fowler [00:13:42]:
But they don't have to just view the new Spark MCP Server. So tell us a little bit about that and the cost.

Allan Nielsen [00:13:49]:
Yeah, so with the front end of choice here, it doesn't mean that you have to just pick one. You have the flexibility to add additional front ends. And for anybody who wants to start working with the MCP server but remain on their front end of choice, so to speak, it is as simple as to go in and start their subscription on the FlexMLS platform. And all that is handled through our dashboard in order to turn it on. Like I said, currently we have information about the MCP server sitting down in our labs under the Labs tile, but we are within the next, I would think probably 30 days, Dorify will be launching a bigger campaign that will heavily touch on the FPS or Flex MLS MCP server. And we will also talk about what Dorify has coming out in terms of NCP server and the data that will be accessible through that, such like engagement metrics and, and what have you. So we are really putting a lot of focus on this. I think it makes sense.

Allan Nielsen [00:14:54]:
This is the direction that everybody is going and like Matt was saying, good for you guys at FPS to really push this and basically come out first. So we are very excited about that.

Michael [00:15:06]:
The first thing that jumped into my mind, Michael, and thank you for having us here. But the MCP is definitely a new terminology to most folks, but they are familiar with the P in there for the protocol. And everyone who's used a web browser, which probably everyone is familiar with HTTP, and I think that we all learned at some point that the HTTPs, that that S is kind of important at the end there. And for folks who don't know, that's just the secure portion of the hypertext transfer protocol that allows web browsers to talk to each other. It's the part that handles encryption and security and whatnot. Is there an S for mcp? Does that make sense in that world? Is there a secure version of the MCP protocol that is coming down the pipe, or is security baked into MCP for transport? Yeah, tell us about that.

Michael Wurzer [00:15:52]:
Yeah, absolutely. So I think the most important thing to know about the Flex MLS MCP server and the Spark APIs, which is what the MCP server connects to, is that they all require authentication. So it isn't just anybody that can access them. It has to be somebody who is an MLS member who has permission, who you know by the MLS to do that. And the plus side of that too is that then when you are accessing it, it knows who you are, as I outlined before, because of that authentication. So everything. And it, and it does run over secure channels as well in terms of how the data is transported back and forth. But most importantly, again, it won't let you edit anything that you can't.

Michael Wurzer [00:16:40]:
Don't have permission to edit. It won't let you read anything that you don't have permission. So a good example are contacts. The contacts are all scoped to an agent, just like in Fleximos. When you log in and look at your contacts, you're the only one that can see those contacts. And so that's the same with this MCP server. Everything is scoped based upon the authentication. And when you do log in, that's a lot of what you're doing, right? When you connect it, it says, do you want to.

Michael Wurzer [00:17:06]:
You have to log in to access this and then you put in your username and password and just like you would be logging into the MLS system. So it's all secure in that way. And the other thing that the reason why we have these little toggles to turn it on in the first place, it isn't so much about toggling on that user, it's about making sure they see the terms, you know, because it is an AI thing and we just want to make sure everybody knows these answers that come back are AI generated. And they do have to be responsible for checking the information, double checking information, all of these things. So it's this wonderful new world that's super exciting. But being careful and cautious at the same time is really important.

Michael [00:17:49]:
So I'll follow up briefly there. So that's for your subscribers and the one to one authentication makes a lot of sense there. So it's kind of a walled garden for subscribers. But Flex is a little bit unique in their environment in that they've created this powerful set of APIs that also then integrate with a marketplace that MLS providers can use and vendors can connect and submit their applications to that marketplace. Is there a world in which some of those services will start using MCP as the data transport as opposed to traditional API? And it doesn't make sense for all sorts of applications or services. But what are your thoughts on that?

Michael Wurzer [00:18:25]:
I do believe that will be coming. We currently are not putting the MCP server forward in that particular way for people who want to enter. Most of the people that are interacting with our APIs today are getting credentials and they are software developers and so they can just access the APIs as they're intended programmatically and they'll probably use AI to develop the software that accesses the APIs, but they're doing that directly without the MCP as an intermediary, but certainly I think a great use case. And again, we haven't really set up the credentials for this for the developer accounts, but being able to understand the metadata and querying the MCP server to just get back the field information, you know, what fields does this MLS have? Which ones are standardized, which ones are cut local? There's a lot of information available through the APIs that you could then ask questions to in your coding agent and say, hey, tell me about this. Why is this query not working like

Michael [00:19:31]:
a quick start layer for developers?

Michael Wurzer [00:19:33]:
So we definitely see that right now this, all of the MCP stuff is focused on actual end users rather than developers.

Matt Fowler [00:19:42]:
Let me expand on that just a little bit, Michael. And Michael, because I think it's important. And then as we get towards the end of our time together, I wanted to see if FBS could do a little future casting. And I don't know if you're ready to talk about Scout Michael, and what's coming down the line for fbs, but whatever you want to share with us, I think before we move to that, I think it's important to kind of just underscore what you just said about the API. I've been describing the MCP servers as a novel new way for liberal arts majors like me to use the APIs because they were just opaque for me before. I'm not a developer, but I have data questions that I want to ask, and I now have the ability to do that through the MCP servers because the access is just, I don't know, it's more usable to the bulk of our subscribers than APIs have been. But my point I'm attempting to make is that it's not really more than that. The API is the reason that all that permissioning works.

Matt Fowler [00:20:42]:
The API is the reason that we have that metadata. The API is the reason that we were able to stand up MCP access to this resource in such a comprehensive and deep way. When I think much of the industry is struggling to just organize their data, to put it into a format that would be publishable in this protocol. Because they haven't made the decade long, you know, who knows what it. What has been spent on Spark so far. But this is when, in my opinion, as CEO, has hired FBS as a vendor. I think this is where you guys really start to see the returns from that investment. The MCP server is so extraordinary because of the API.

Michael Wurzer [00:21:28]:
Yeah, I totally agree with that. And I think it's a really important point and like I said, when a user logs into any MLS system, but for sure, Flex MLS in this case, what you're doing is powered by APIs. Today, everything that moves, the data around your interface, that you're used to, the search, all of that is powered by APIs. And so because of that, now you have this entirely new vehicle where you can ask it questions in plain English and get back plain English responses. And that is a really empowering set of tooling, I think, for people to engage with the MLS system in just new ways. And we know like this whole conversation is pretty geeky, right, The MCP and mb, but all that stuff, most users, you know, they don't have to learn that, they shouldn't have to worry about that, they don't have to think about all that. It's the early adopters that are going to go try this MCP stuff and that's awesome. Everybody can and should try it.

Michael Wurzer [00:22:32]:
But you know, you're busy selling real estate and so we respect that. And that's where Scout comes in. Scout is a chat interface that we'll be introducing to Flex MLS and making it available so that you can just ask questions. You'll just be there and you'll be able to ask questions. You don't have to think about connecting an MCP server or re authenticating because it'll automatically authenticate you since you're logged into the MLS system already. I mean, so all of that will just sort of be taken care of. You won't have to know about models or anything else. You'll just start asking questions and hopefully getting really good answers.

Michael Wurzer [00:23:09]:
And so that's, that's hopefully coming July ish is where we're aiming at for sure. This summer.

Matt Fowler [00:23:16]:
I thought it was later than that. So really soon then that's just.

Michael Wurzer [00:23:19]:
Yeah, yeah, we're just going into beta now with a couple of customers. You know, there's a lot, lot to be learned about how, especially on how these LLMs process large sets of fields like in an MLS system. So we're learning a ton and we'll keep learning, but. But we do expect to have a product available this summer, probably towards the end of the summer.

Matt Fowler [00:23:44]:
That's amazing.

Allan Nielsen [00:23:45]:
I'm going to take it way down here or way up, I suppose we can say. So what if I'm an agent that used to run my search in my platform and I exported it and I copied it all and I put it into Claude or ChatGPT and got restored the Results that I was asking for. Whatever. Tell us why that's a really bad idea to do it that way and why this solution is so much better.

Michael Wurzer [00:24:14]:
Yeah, I think the first thing is that now you know that it's using the MLS data, right? Whereas if you're just generally in ChatGPT or something, it'll go search all sorts of places that you don't, you don't know about. And so I think having it be directly connected to the MLS system just gives it that more certainty that you're getting what you want. In this day and age, when listing information is available from so many places, having a trusted source is really key because you don't know how often that other source has updated the information. Even if they're getting it from the mls. Well, they might be less vigilant or maybe they don't have all of the information or whatever the case might be. That's where doing those kinds of searches, I think in, you know, knowing that you're hitting that, that source of truth is really valuable and important.

Allan Nielsen [00:25:09]:
Michael, anything else you want to share about mcp? I do want to touch on a different topic because today, which is June 2nd, where we're recording this, we are actually going live with the Auto Assistant. So I think it would be great to just touch on that. But before we make that shift, anything else you want to add about the MCP that we haven't talked about?

Michael Wurzer [00:25:31]:
No, just say, yeah, try it out if you're interested. But if you want to wait for Scout, that'll be another great use. You'll also be using MCP then. So it's just a new fun thing to try and it's going to be very powerful now and long into the future. So this is just an intro and think about it like that. These things do move fast, so it can be scary, but this is a time to just be fun and experimental and try things out. But yeah, autos, the new listing entry module you guys are launching today, hopefully that goes swimmingly. New launches are always a little scary.

Michael Wurzer [00:26:10]:
So I'll knock on wood and hope everything goes great. It should be smoother faster for people to enter listings. There's a lot more auto population happening from the very beginning just simply by entering the address. Even parts of the address, if you get house number in the first part of the street name, typically it will auto suggest an address to complete that. And then it automatically looks up public records, takes photos and incorporates those and will suggest detail fields, which I can give you all a lot of props on. Because when you set up your schema, your database in the first instance with us, you were very vigilant about being compliant with the RISO data dictionary. And that's really going to, that's paying off with this auto population from photos because that's also targeting the RISO data dictionary. And so there's a lot more matches as a result on your data.

Michael Wurzer [00:27:05]:
And so again it's just really trying to streamline the listing entry to put as much data from, from whether it's current listings or copied listings from others that you have permission to copy. And every MLS has their own rules about that to auto populating from public records to geo information to the photo AI and public remarks, you can use AI to auto populate those too. So lots of tools to streamline listing entry. That's the main focus of these new tools.

Allan Nielsen [00:27:41]:
Fantastic. Yeah. And these are features that all Dorify subscribers will have access to, including the photo auto populating from photos. So we can't wait to see the adoption on that and have I would say significant expectations to see improvements on the listing quality because of that. But I think it is important to say that when listings are auto populated there's a responsibility still for the listing agent to validate that that data is accurate. And like we said, we're launching this Today and again June 2nd and trainings on this new features will take place during the month here. So definitely for those who wants to attend those sign up for them. Yeah.

Allan Nielsen [00:28:25]:
And I think unless there are any other questions from anybody, Matt or Michael. Any other questions?

Matt Fowler [00:28:31]:
Alan? I think I'll point out something real quick and let's Michael consider if he has a question. But the auto pop is something that you know, we just turned on today. There are people that really resist using that brokerage offices that you know, tell their people to limit the use of it. And I want to share with everyone that the, the kinds of problems that we have with autofill are not really about the data not being accurate, it's about not being complete. And it's, it's and constant for somebody to autofill property. I ran across one the other day that was two different parcels being sold in one MLS number. And when they autofilled the tax information, only one tax fee came through. So they got less than half of the actual taxes on the total land for sale because it was made up of more than one parcel.

Matt Fowler [00:29:22]:
So the listing agent looked at it, saw that it was correct, it was correct, it was just not complete. So we would really encourage everyone to Lean into the autofill and then to do a double check check with the seller and everything you can do to make sure that the data you're publishing is accurate. We're of course at dorify, we believe in measuring absolutely everything. So we're going to be measuring to see if we do get more fields for today and if they are more accurate, we have the ability to make those comparisons. So we really want people to really lean into that and try it. All of the new ad listing options that are coming out really lean into pulling data from other authoritative sources into the listing workflow just to make it easier and faster and more accurate for everybody. Yeah, so that's what I wanted to point out.

Michael Wurzer [00:30:09]:
I would echo everything you said, Matt, and just and say that the system is designed to essentially be a learning loop. So the more feedback we get from people about that, we track it in the system. If for example, as the auto pop occurs, do they edit that information or not? Do they have to correct it or not? Those things are all there so that if we see that there's chronic problems with certain types of public records, for example, well, we can work on trying to get better public records or better gis or better whatever the case might be. So usage is really, really important. If people just bypass those features, we're not going to. The system will have less ability to learn. So getting that feedback both through the system usage, but also through verbal and emailed and other reports, super valuable to improving it over time. Real estate data is, you know, it's incredible how difficult it is to bring systems together, but that's getting more and more powerful all the time with different ID systems and things like that to tie these together.

Michael Wurzer [00:31:18]:
And that will just make agents jobs easier as time goes by. So we really want to, as you said, lean into that to ultimately make it so that the agent can focus on selling and hopefully the system and your processes that the MLS can worry about accuracy and other things like that more and more and more. Agent's always responsible at the end of the day, but we all want to make provide accurate data to help speed your job along.

Allan Nielsen [00:31:45]:
And also I think with the new UI that is coming out, modifying your listing on your mobile is now a completely different experience as well. It's always been, well, I'm going to wait till I get in front of my laptop to do this right. But now the interface actually makes that much more doable directly from mobile.

Michael Wurzer [00:32:02]:
Yep, yep. All 100% and every field, every part of the listing fully editable on your phone.

Allan Nielsen [00:32:07]:
Wonderful. All right. Well, thank you for your time, Michael. Thank you for telling us about MCP and APIs and all of that stuff. And also on the, the on the auto assistant as well. So. And thank you, Matt and Michael and Dusty who's on here as well. Thank you, everybody.

Allan Nielsen [00:32:26]:
And thank you for everybody listening in. Links to our social media and other things will be provided in the description somewhere. So thank you, everybody.

Michael [00:32:36]:
Thanks.

Michael Wurzer [00:32:37]:
Have a good day.