PROPTECH PANEL:AI and Machine Learning in Proptech

Watch Proptech Australia President Kylie Davis and a panel of four of the industry's foremost leaders for a pivotal discussion on the future of AI and Machine Learning. In this webinar, we explore the latest technological innovations, current adoption rates, and the significant emerging opportunities shaping the Australian market.

Expert Panelists:

• Damian Merchan – CEO of Sonar AI

• Leann Jones – CEO of Nimo Industries

• Aonghus Stevens – CEO of Asseti

• David Howell – CEO of Dynamic Methods

On the Agenda:

The Current State of AI & Machine Learning: An in-depth look at trends and adoption rates across residential, commercial, and property management, with a comparison to global leaders.

Emerging Technologies: Insights into what's on the horizon for AI/ML in proptech, including Generative AI and agentic AI.

Human-AI Collaboration: Exploring how AI augments human expertise in real estate and the importance of human oversight.

Investment & Innovation: Identifying key investment opportunities and the evolving landscape for proptech AI in Australia.

Real-World Use Cases: Practical applications of AI in property valuation, market forecasting, tenant experience, predictive maintenance, and more.

Overcoming Adoption Barriers: A frank discussion on the challenges to AI/ML adoption and actionable strategies to overcome them.

Data, Privacy, & Ethics: A critical conversation on data quality, navigating Australian privacy regulations, and ensuring fairness and transparency in AI.

Don't miss this essential opportunity to gain clarity and strategic advantage in the evolving world of Proptech!

Transcript:

Kylie:

Hello, everyone! My name is Kylie Davis, and founder and president of Proptech, Australia, and it is great to see you all joining us here today for our Proptech panel on AI and machine learning. And I'm really looking forward to a really deep dive conversation. Today, I am here today on land of the UN. Nation, home of the Budgela Manji people on the very beautiful south coast of New South Wales. It is an area that covers from Lake Conjola up in the north near Aladulla, and stretches all the way down to just above Naruma. No, I think it comes down to our mates down at Powerhouse at. So we're about half an hour past Bateman's Bay, where I am. And so, before I begin, in the spirit of a reconciliation project. Australia acknowledges the traditional custodians of the country throughout Australia and their connections to land, sea, and community. We pay our respect to the elders, past, present emerging, and extend that respect to all aboriginal and Torres Strait Islanders joining us here today and I'd also especially like to thank the support of our sponsors, who make all of our Proptech panels possible.

Leann:

And if you don't know stone and chalk, it was founded as a not-for-profit in Sydney in 2015 to help Fintech startups, commercialize and grow, and from 40 startups in 2015 it now has well over 200 startups in Sydney, Melbourne and Adelaide, covering all areas of emerging Tech, including Proptech. And I think at last count, they had around 2030 Proptechs that call stone and chalk home and have recently moved up to the new tech central.

Kylie:

So now, is there any hotter topic in technology at all at the moment than AI from the launch of ChatGPT in November 2022. The adoption of AI has captured the collective imagination and become one of the most popular features used to promote technology. Sorry, I got myself lost to promote technology and Proptech is absolutely no exception. And in the recent Proptech awards a whopping 75% of our entries featured AI from its initial launch in helping agents, writing real estate copy to finding leads and writing better emails. AI is now automating workflows, analyzing property and client data, assessing risk, predicting risk, asset maintenance, requirements and modelling and the fastest way to property ownership scenarios just to name a few. But there are anxieties around using AI.

So what are the important things that all clients of Proptech, whether they're real estate agents, property owners, or facility managers? What do they need to know about the AI operating in the back end of their systems? What are the questions they need to be asking to manage AI safely? And what guidelines should we, as Proptechs, be following to come up with consistent and trusted AI experience across all of our clients, because no one wants a skyline.

My panel of experts in this area are recent winners from the Proptech awards, who all boasted AI in their entries. Damian, merchant from AI analytics platform Sonar, hey, Damian.

Damian:

Hi! How are you?

Kylie:

Great Leanne Jones from modular banking platform, Nemo industries.

Leann:

Hello! Everyone.

Kylie:

And facility management. Expert Aonghus Stevens from Assetti, Hi. Aonghus.

Aonghus:

Hi, Kylie! How are you?

Kylie:

And they are joined by the creator of the 1st MCP protocol. And yes, I realize that protocol is a double up when I say MCP protocols, but sorry, because who knows what the Tlas mean? David Howell, from dynamic methods and forms live? Hey, David, welcome back.

David:

Hello! Thank you for having me.

Kylie:

Yeah, so look welcome everyone. So because we have so much to dive into. What I'm going to do is ask everyone just to sort of introduce their Proptech and the problems that they solve and why AI is ideally designed to support the solutions in your space and then we're going to open it up to a broader conversation, and for those following along. Like, if you've got any questions, please pop them into the chat. It's easier for us to just see them in one spot rather than questions here and chat over there. But if you've got any comments, we will kick off. So as the newest Proptech in the space Damian, have you launched yet? Because I know you were very pre-launch when you entered the awards. But let's kick off with you. Tell us what sonar does, and for who and why AI is well suited in your space.

Damian:

Yeah, thanks, Kylie, so we are doing a soft launch in this. So we've got up to yeah. 20 clients. Now, 20 different agencies are using it so soon as a performance platform that's built for residential real estate leaders. We unified data from sales. Li, listing performance data, you know, coming from Crms finance data, market data and bringing that into a unified view for the leaders of that business. So they could say, you know who's on track, what conversations need to be had, which listings are performing and which aren't those sort of insights we're really trying to give the right signal at the right time. So, yeah. And I think you know, AI is very good at that. It's taking the data sets, finding, you know, signals in the noise, so to speak. So yeah, that's what we're doing.

Kylie:

And so, could you have done this 5 years ago, or is it only something that is now possible.

Damian:

Yeah, no, definitely, not you could have done it 5 years ago, but it would have been hell of a lot more expensive. So you know the availability of tools in the space. It's just, you know, exploded. And so you know, the cost of computers has come right down, and the tools available. You know, you could have done it with big teams. But now you can just go a lot further with less resources. So yeah.

Kylie:

Awesome. Okay? So, Leanne, let's talk about Nemo. What? What does Nemo do? Exactly. And who are your main customers? And why is AI required in your solution?

Leann:

Thanks, Kylie, much like Damian. Actually, we unify data. So we are a digital online banking solution. We're white label products that means that our main customers are banks and lenders, Adis and non-adis, and really in turn, then the end customer might be Sally on the farm who wants to buy property or refinance her farm. She's our end user. So what we offer is that complete digital experience from origination, application, assessment contracts, consents, and all the way through to settlement and fund disbursement of funds and then we still provide that one unified platform to manage and service that particular that particular loan, and there's portals there for the broker, for the lender, and for the end customer. So it's a much more seamless experience from both the end customers. But equally for the user, which are our banks and lenders, and importantly, for our broker.

So what it means in real terms, like Sally might be wanting to refinance or equity release to access funds for some yellow goods. I mean, she can do it all digitally through her portal. So, rather than using different sources, different offerings. For example, digital ID. You know, taking time to go and see a broker or a branch manager to get those funds in terms of AI. Probably a little bit like Damian, like we've got a really comprehensive platform, sophisticated workflows and rule sets. And it's really quite comprehensive. This we call one platform to master them all. But there are things that we're doing lately with AI that we probably were unable to do 8 years ago, to be honest. And we're able to offer things with more resources, more tools, but more economically. And I think that's a really important part for some of the pain points that we're solving.

If I think about a key pay point with lenders and their core banking functionality, a lot of our in the ecosystem of property. Once you take out that property, it's got to be managed that loan's got to be managed, and current lenders are reliant on certain legacy systems, or they have to build their own. They just don't have the funds. So when they're using Nemo to use AI to do a lot of complex thinking. We can do it way more economically, smarter, a better user and a customer experience. And what that means in turn for the property ecosystem we can support more new starters, more offerings in the market for lenders to use, whether it might be an equity, release product, or a particular view on how you might want to access your lending and your funds. We enable a lot of those sort of t tier, 2 tier, 3 tier, 4 lenders to be on the market that consumers can then access.

The second really important use case for us is in the broker segment. We know that 90% of the brokers of our 17,000 brokers in the Australian market use the system next Gen. So we use AI in a very clever way to populate extract data to make that a seamless experience rather than having information that gets mismatched, erroneous data, etcetera. So staff to Daniel's point that we couldn't have done years ago, or we could have, but it would have been way more expensive, and to be able to offer that to our lenders.

Kylie:

Just so, just to make sure that I'm clear on how the AI is working in your system. So your legacy industry has always been like lots of different silos of, you know, accounts and styles of lending, and all of that sort of stuff. Are you sitting on top of that, or have you like it? Is the AI, which is where the AI kind of pulls all of its magic together, or are you? Have you? Have you rebuilt all of those structures into one big thing?

Leann:

We're our proprietary platform. So if you like, we, our platform, our developers, have built it from end to end, and are a bit like a toaster. In that journey there are different off ramps and on ramps in which we work with other 3rd party suppliers. Whether it's digital ID, statements, verification of income, etcetera. But the way in which we treat the data we use agents and tools to be able to treat that and streamline it in our data, in our data pool, which makes the data safer. And you know, very clever what we can do with the data to make that seamless experience for the lender and for the customer data is a really good example of being able to offer different property valuation options at a more economical level for brokers and for banks is one of the things that we can do. And we do that because of AI.

Kylie:

Awesome. Okay, cool! Aonghus, congratulations because you won quite a few awards on the night. So tell us a little bit about a Seti and the problems that you guys are solving. And the role of AI in that.

Aonghus:

Yeah, absolutely. We did have fun on the night with Carla. So thank you for that. In terms of a city we used property portfolios, distributed property portfolios to predict, plan, and operate those assets. Basically. So our ideal customer, or the sort of key use of our platform as sites across many areas, many geographies, and the challenges that come with that we focus primarily at the start on condition data so building like asset hierarchies and defects, and these sort of things defect registers. We've moved beyond that now into more of the general operations function that these companies have but in terms of how AI sits in there and sort of this, this.

We'll probably touch on this bit later, too. But the transition of system of record to system of work to agentic systems is sort of like the way AI is generally considered in these products, anyway, in products generally. So it's like, okay, how can we? In the system of record sense, we're building source data for these customers to use. And that's still a really cool part of what we do. And so that's where more machine learning on very large sort of repeat data sets that come in to enable us to generate this data really quickly, and deliver, you know, not quite an insight to a customer, but, you know, generate some sort of rows in a in a table, for what a better description, then, system of action stuff is.

Okay, how can we use AI in a product sense? So this is more customer facing this sort of stuff, and how can we help the customer surface and automate some of their action? But they would still do that in their system. And then the agentic stuff which is coming out in the product is almost like hands off. So it just does the whole thing for them. And they're sort of the ways that we use AI. So we use AI to more rapidly generate the data that's critical to our system. But a customer may not actually see that or interface that. AI. And then there's the actual customer facing product sense as well.

Kylie:

And so from a customer facing product. Guys, I'm just if you're not speaking, can I get you to pop onto mute? So that would be great. So I guess from a customer facing thing. What is the AI doing from their point of view?

Aonghus:

Yeah. So for a customer, it's obviously originating data for them to see. So they see that data still, but very quickly. Then there's, you know, the sort of what we call the “Assetti agent", but more like a Chatbot functionality that sits in there to help them really understand what's happening in their data? And let them interrogate that to a level of depth that they can, because customers that we deal with might have like a hundred 1 million records in their asset hierarchy or something.

So there's like a ton of data in there that they're really sort of working with and then it's also, you know, how can they say they're using our asset? OS functionality to, you know, take a defect, generate a work order rather than them setting that workflow up. It's like, Hey, you've done this 10 times in the past. Would you like to automate this little section that's sort of like where they would use it today.

And then where it's heading, I guess, in the agentic piece is okay. How does this all happen? And that's where I think agentic is sort of seen as this Nirvana. But really it's AI in the sense of how we look at it. And you know how I think the issue is looking at is like, how can we deliver an outcome for the customer as quickly as that outcome needs to be reasonably achieved. So in the sense that we generate, you know, we take a data signal of some sort.

We ostensibly want to deliver the outcome for that as instantly as possible, because the value of that insight or that signal starts to degrade. And so that's really what AI is to a customer. How quickly can we deliver the outcome that the customer needs for as little effort as possible for that customer while they understand it.

Kylie:

Yeah, got it. Fantastic! Okay and so, David, tell me a little bit about what forms live does and the role of AI in your workflows.

David:

Okay. So forms live is a saas platform which has all the real estate forms and agreements. So everything from residential sales, property management, commercial, rural things like that. And I guess we're doing things a little bit differently in terms of what we're doing with AI rather than using AI internally, which we are, you know, Vibe, coding and and things like that. We're actually building tools on top of our platform specifically for AI. And you said earlier that we've developed an MCP server which is a protocol to help the agentic AI or any AI really use things in the real world, because ultimately most AI platforms are very enclosed, and you can't do things like tell that you can't ask what the time was, you know, in the early days.

You couldn't ask it to send a form you couldn't check to see if forms were out for signing unless you had all these tools and things like that. So we're actually building the tools for AI rather than building AI tools within the platform itself. Yeah, so it's a little bit of a different tactic, but it aligns heavily with what Aonghus was saying. So we're kind of the other end of it. And I think Aonghus, is this end, you know, processing hundreds of millions of records is perfect. Your AI case because it's kind of self contained. And it's very good at data analytics and NLP and things like that.

But then you want to be able to do those real world things after you've done all that analysis. And that's where MCP. Servers can come in and the Llms. Can actually use that information and do something out in the real world. So that's what we're doing.

Kylie:

Cool, and it sounds like Aonghus. You're starting to do that as well, too, as part of your experience that people are having with you.

Aonghus:

Yeah. And I think that that's where absolutely, we deliver. You know the outcome for the customer, I think an interesting piece around the relevancy for the customers and this is our, you know, this is my view and our view on this business. But we can help our customer achieve the quickest and best outcome by the proprietary data we hold.

So inherently, they're giving us the stuff that we understand, because that's the value of what we provide. But certainly, we see the outcome like where we've had the worst outcome, where we've had the least value. Sort of provided to a customer is where we're disconnected from the outcome. And so that's really our focus is, how can we?

How can we retain the whole life cycle of a decision? Because as soon as that decision gets closed out that actually provides us the information as well. And so you know this? It. It's definitely a journey, absolutely. But you know, I think, yeah, staying up to date with AI in A in a product. Sense for us as well, is really critical, like in a background, and otherwise.

Kylie:

There's a really big Venn diagram here, though, isn't there where data and absolutely overlap because you can't AI without data. And I think what I'm hearing in all of these examples is that AI has this capability to chew through huge amounts of data in ways that human brains just simply couldn't do previously, and that has taken us weeks, months, years, even of manual processing but and that sounds awesome! But where some of the either risk or complication comes from is that the AI works better when you add your data sets to it.

But where does your data end and start once? It's sort of sitting inside the AI. So what you've just outlined there, Aonghus, is this whole idea that if you try to keep them separate it won't work as well, but when you overlap them, what you know from dealing with whopping, great data sets sitting inside facilities management all the time means that you, almost as a customer. What makes sense? Check your data against that, and then have the AI start to give you the insights based against your data set and your own.

Aonghus:

Yes, like an example, there would be and absolutely like the tailoring of it at a customer level is sort of critical. So we think about it like it at a product level that is regional. So the way someone in the US does something to Australia is very different. So we sort of have it at the regional level. Then we'd have it at the workspace level, the customer level. Then we'd have it at the actual user level. So they like the tailoring of the product or the AI for a customer. But the way you know, one customer fixes leaves in their gutters for an example, would be vastly different to someone else, and that directly impacts the sort of outcome of a guttering component across its life cycle.

So for us, if we can see into their sort of CMS data and see historically how they've fixed things, we may actually make a decision earlier for the customer, like we might increase the criticality level of a defect, repair because we know it takes them longer to repair it as an example. And so they're the sort of ways where, like actually actually knowing what a customer, how a company is operated is really critical as well for us to provide them the best outcome as quickly as possible.

David:

We're doing something very similar. In the form space. Because and and you couldn't do this without the AI. Well, same as before. You could, but very expensive in terms of machine learning. But simple things like analyzing how your forms are coming in and how quickly they're being electronically signed and things like that. So you might see, you know, the AI might see that there's a pattern, or there's, you know, for tenancy agreements coming in and getting signed, and there might be one or 2 which are not quite completed or executed, and rather than you finding out later and having to, you know, reactively, fix something.

The AI can actually analyze that and suggest that perhaps it might be worth giving Kylie a call this morning because she's not quite yet finished her tenancy. Agreement and just really simple things like that could save you 5, 1015 min, or could save you a lot of effort down the road. And there's, you know, it's kind of AI working for you on those types of scenarios.

Aonghus:

And I think the magic in that sort of scenario is like you think to say, Damian only and sort of use case, and it's like a can. They see the correlation between a delay in this form being signed, and an actual at risk, line or –

David:

Yeah, you start, join them together.

Aonghus:

It's like, How can we just like, you know? So in the Damian sense, the customer, by the sounds of it is, you know, is the agent or the office, you know, and sort of maximizing the gross commissions effectively for that business. How can you know? How can. How can the AI deliver the best outcome for that customer, you know, and that's the piece where I think this becomes really interesting, just like, continue. And that's why it's not a destination, you know. It's just like it's just a continual evolution, which I think is the exciting part about it.

Leann:

Can I add to your point, Aonghus? It's about how we best serve our customers and the banks that serve the customers. But I think you've made the point before. It's about knowing who you're dealing with and knowing what happens to the data. And we know of examples in our industry and with other players where you might think the AI is being processed or it's happening onshore.

And we know of examples where it's happening offshore. We talk about data, sovereignty, and one of those risks I know that's what you're keen to explore and what people are perhaps anxious or concerned about when we think about Australian banking, there are really tight regulations around what we do with our data and where we do it.

So if I was to think fast forward into 40 min from now. What I'd leave people with is to consider who your partners are and who you're doing business with and like, where and how do they keep that data? Where do you know? Where does it happen? You know it's almost like the analogy. Lift up the hood of the car. What do you see under the car?

Kylie:

Yes, yes, so. So one of the you know that consumer and finance data Leanne in is often seen as highly. Well, you know it's private, and it's kind of quite personal. How do you guys protect that? Isn't that where people get anxious around AI? What are your protections in that place?

Leann:

Yeah, look, I think anxiety, absolutely. People feel what they don't know. And we use the analogy a lot in business. It's the clear box versus the black box, and we've always been using AI up until now, I mean 2023. It was. Everyone was kind of talking about it. 2024 people are dipping their toe. We saw lots of random acts of AI, and I think we're seeing fitted acts of applications of AI.

Leann:

And I think that's where we're at in terms of advice, I'd you know, be considering making sure there's considered use of AI so when we think about the use case in Nemo and how we keep it protected. We make sure who we partner with. We ensure data sovereignty. You know, we were dealing with data and processing it onshore. Of course, there's, you know, guidelines that have been introduced in the country, and of course we've got privacy, legislation, etc, that protects us. But I guess how we protect what we do.

It's all proprietary built, so we're very careful who we do business with, and how. But naturally, of course, consumers are really, really concerned about it. When I think about our lenders, though, it's a really interesting mix when we come across decision makers, whether it's board members or C-suites, or heads of or whoever they are.

They're really excited like AI is the buzzword. And it's like this, you know it's a balancing act between really, really excited. They want to know that you're innovating. But are you safe and secure and stable? And is my data protected, and that's our job to be able to answer that question for them and give them that safety and that answer.

Damian:

Yeah, I like, sorry just to jump in. But man, clear box, the black box. Analogy is a good one. Because, yeah, I find that something our customers are always wanting to avoid is that black box scenario? If you're giving them an insight, they want to know. How did you arrive at that. Where is the source data coming from? How is that being calculated? Those sort of fundamental questions? I just don't want to be in the dark around that. So.

Leann:

And the risks are too high. We can't be.

Damian:

Yeah.

Kylie:

So one of the things that AI is letting us do, I guess at the moment is iterate new tech or new? You know new versions of our platforms or code really quickly, and things like that, and that, I guess, compares against the anxiety and also against the slowness of adoption. Do we think that AI from our customers do we think that AI is going to kind of create a bigger gap between kind of customer adoption, and where the tech can go? Or do we think that actually, the sexiness of AI is going to basically drag everyone forward a bit faster.

David:

Well, I see, I think, AI, ultimately, the biggest value of AI is that it's an easy user experience environment. So you know, you think about your mobile phone. And you know, if you're using Openai or or Claude, and you can have a conversation with your computer effectively, you know, and then you add the MCP server tools on top and then you've almost got a mini assistant right in your phone and you can just talk to the to that assistant rather than having to type and be on a desktop, or, you know, be on a tablet or anything like that.

So to me, the natural language passing is the NLP. It’s probably the biggest advantage we've got right now, I think long term, or down the track. The value around the power of AI will emerge. But I think we're not quite there yet. So things like context, rotten and memory and all those types of issues are really the biggest barriers at the moment, but ultimately.

Kylie:

Context, rot.

David:

Context. Rot.

David:

Yeah. So you know, the more the more you feed the AI over time it kinda starts to get a bit confused. You know I'm using, you know, hallucinating. And all those types of things.

Leann:

We did. AI restores. And yeah.

David:

Yeah, so prompting. And all those types of tools. But you know, and the other one is to rag around, you know, storage of this information, making it easy for AI to use and utilize, I think, is really the barriers we're coming up against at the moment. That's where MCP servers can help because for example, in New South Wales, we've created a termination notice, Wizard because the latest legislation changes made it quite more comprehensive in terms of having to get a compliant termination. Notice, and that includes gathering evidence from the landlord and things like that. So what we do is we build your regular finite state machine type workflow.

Then, we bring it into AI through MCP servers and the tools around that. So then the AI can use that workflow without having to learn how to do it, if that makes sense. So whereas if it was a person you'd have to, you have to run through the details and and you know how. Get them to understand the workflow. But through MCP tools you can actually guide the Llm. Through the workflow programmatically. And that's probably the biggest advantage that we've got around our MCP service. So, David, is that creating sort of one of the like a risk mitigation in that.

Kylie:

If you ask the AI to identify. If Damian is – sorry, Damian, we're picking on you. But if Damian is a good tenant, or if he's not, and if maybe Damian had an argument with a landlord in the past, and therefore that's kind of been out there. But it got resolved. And if it's still hanging around, are you able to – are the tools able to say, but you know but disregard xyz or –

David:

We're more around the compliance side of things. So, making sure as a business, the customer is doing everything by the book, so less about managing that. But it's the way I see. AI now is almost like a junior or an apprentice, you know, so you shouldn't give the tools to the, to the office, the keys to the office, you know. We've all heard that I don't. Well, in the development community, the software development community you hear of people have hooked AI up to their production database. And it's decided that it's going to remove a whole bunch of records and put a whole bunch of mock information in there, and you know, so you can't. You can't give it everything. And again.

Aonghus:

That deserves that to happen, David.

David:

Exactly. That is true, but that's the thing. You wouldn't give your apprentice, you know, production access to your database, or let alone right right access to your database. So yeah, they get what they deserve. But they're the things people are trying to do, our focus is on security and privacy so that's the advantage. It should be the advantage of MCP. Servers if they're done correctly, that you can't get yourself in trouble like that, you know. So.

Kylie:

Indeed, layers of access.

David:

Right.

Kylie:

If you're a real estate client or a Proptech Proptech client in any way, shape or form. One of the 1st questions on your list should be. Do you let the AI iterate off your.

Aonghus:

I will try. Well, yeah, and think about like agentic stuff. So in the sense of an agentic system, where it is a system of action. You actually need it to be able to create, read, update, delete. It has.

David:

You do? Yeah.

Aonghus:

And so.

David:

You can limit it. So you know, we can say, look, we can create forms, finalize them, read them, analyze them. But you couldn't send an email to a customer, or a client, or a tenant, or a thing like that, so we can.

Aonghus:

Around like, how does it actually like the productization of AI, then is this piece? Where? How do you actually, you know? And this is a big piece of like, I think, a bit of a training. We'll just put AI in and see what happens but –

Kylie:

Well, you can code it on your apple phone, and you can learn how to do that, you know, in 60 days.

Aonghus:

But it's like this divergence of the product. And it's like, who is the customer and what outcome are you selling them? And how does AI deliver that more coherently? Whereas, I think, like, in the sense of people putting on an Llm. You know, over the top of their product. Nothing wrong with that necessarily, if that's like, if that helps people. But you know, if all of a sudden it's then taking, you know, insights from Google search and feeding those in. And it's like, Okay, well, what is your customer actually doing in your product that this delivers faster or better for? And it could result in an actually a more diluted and less coherent experience.

It often does for the customer, and I think that's where the promise of AI has failed in that. Everyone has just put AI in. And the 80% of people, you know, products that haven't actually focused on the outcome of the customer like, I can tell you, not a single one of our customers gives a crap about AI. Well, they couldn't care about AI, and if a customer says, like, you know, I'm really interested.

They're 12 months away from buying. So it's, you know, for us, it's okay. How do they use it? Everything a customer interfaces within our system is driven by AI. You know there's people across it. There's Qa, etcetera, that sits there, but every single aspect of it. But you won't find almost a single piece of our marketing that has AI in it, because that doesn't matter to our community or our cohort of customers.

Leann:

I actually agree with David's comment. I feel like in terms of our ecosystem. We're toddlers. I've heard that a few times, and I completely agree. I feel like we're just learning to walk where there are some missteps that we see, and we had just the other day is actually chatting to one of our developers who had heard from one of his cohorts, and they were looking at someone's code on a new app, and there were exposed apps and vulnerabilities everywhere that, you know, we just we need to be really clever and know what we're looking for, and again know where those vulnerabilities and look with a closer eye.

So I agree with David and Aonghus. I also agree with you. I've seen too many examples of. We're starting with AI, and it's not the, it's not the starting point. The customer is going to be the starting point. If AI is part of the solution, that's great. But there might be other ways that we can solve things. It just happens that we're able to extend, perhaps some of those solutions a bit more with AI. I heard about one of the conferences. I was an emcee a few months ago. One of our T. 2 banks in one of the Australian T.

2 banks have got an AI 1st strategy and AI 1st policy. And I'm like, I thought that was really I was curious, because to me that implied everything will be AI orientated. It was a bank, we all know. So I thought that was really curious. And I think we need to remember it's going to start the customer. The customer does not care if it's AI. They want to know what the experience is going to be like, and if we can make that smoother for them, or better or more intelligent. That's what we've got to stay focused on.

Kylie:

Doesn't mean we haven't heard from you for a while, so –

Damian:

Yeah.

Kylie:

Realized, are real estate principles. Are they keen to get into AI? Are they? Where where's their head? At.

Damian:

I think there's cautious optimism. But you know there's a there. There is a lot of fear mongering out there around AI like people. Oh, is this going to replace me? Or you know fear of missing out, or it? You know

Damian:

I mean real estate. Still, such a fundamental people orientated industry, you know agents are thinking about. Where's the next buyer for their listing coming from, or where's their next listing coming from? Principles are thinking about, you know.

How do I keep my great staff and attract other great staff, and you know up my market share, and, you know, create a point of difference, you know, I feel like sometimes we in the Proptech community can maybe think they're thinking too much about us. But really, what all these guys are saying is, map whatever you're doing to the problem that you're trying to solve for your customer. 

AI is full. It's a co-pilot. It's not a decision maker. It's if you're AI first, I think you're in danger of actually becoming just producing gimmicky stuff that amplifies noise. You know some. We've learned a lot, you know, from starting where it's like, you know, we're taking so much data in. And then it's like, Well, how do you cut through and actually isolate and get to the real problem and the metrics.

That really matters, you know, that's some of that AI like we're using it. We're pretty lightweight in ML, we started off, being pretty heavy and sort of, you know, unsupervised learning, and we went pull back, and had to really put the guardrails in these large language models that can hallucinate. So you know, we had to put some really stringent guardrails on that and vectorized stuff. So it's there. But you know, I kind of feel like we just never forget what we're doing. We're building solutions to our customers' problems.

AI is just a part of a toolkit and how to book it. Sorry.

Kylie:

So. Damian, I just want to go back to you and just a little bit about it because I think we kind of started going down the Agentic AI path with some of the things that Aonghus was talking about but with the MCP protocol.

David:

Server, yes.

Kylie:

The Agentic Pro. What are the most interesting things that you guys have learned so far?

David:

We've learned a lot. It's still very early days. And it's pretty much what Damian's talking about. And Aonghus, you know, like it. It's very early. The biggest thing we've realized is the MCP spec is changing daily and expanding daily. And I think it's because we're all kind of working out how to improve it. And it's been great from a community perspective, because a lot of the AI community is on board with the specification which makes it.

It's kind of snowballing into something now. But what I'm really seeing is how the MCP server can help direct the agentic AI into making the right decision and you know through things what they call elicitations and and other types of tools, through the spec. Where at the moment, not all Llms are the same, you know. If you have an on device. Llm. It may not interpret the tool usage as well as it could, as if you were using Claude, or or sonnet, or ChatGPT where they will, they'll just use it instantly.

They know they know what to do. So what we're building is, I think the ability to bring in those guardrails that Damon is talking about just just in what we're doing, which is forms and agreements. So we've got these like, the spec will help kind of guide the Llm. And then we start to talk about what Aonghus is talking about where you've got fully autonomous agentic AI with these tools from multiple servers. And it can start to make these decisions without any help from the outside. But I still think we're quite away from it.

David:

Yeah.

Aonghus:

I think a bit with that too though. It’s like, it doesn't need to sit all in one bucket, and that's where, like, we will never have a fully agentic system for our customer, you know. Some, I'd say some of the roles that a customer has will absolutely, you know, become more, almost a hundred percent agent in the way I engage with our system, there'll be others that that aren't. But it's that sort of letting the customer push into the bucket.

They need to sort of meld the combination that works for them, and I think that that you know, in the sense of being AI first, st I think that's always that's a buzzword, too. But you know, from our sense we repositioned all our code at the start of the year, so it could be consistently understood by AI. That's just purely internal purposes. But you know, in the sense of the way we write our where our code is, you know, sort of been re-architected in a sense that's AI first, st you know, in the way that it can be used. But it also benefits people, you know.

You've got dozens of different people building something. Of course you're going to diverge in the way it's written. And so it's just tidying up some of that. And I think that that's where, you know, using AI to deliver the best possible outcome for the customer that starts internally with us that still starts with the customer, using it in some of their platform facing aspects. It starts with the, you know, the data sets they use in, you know, in the future, we're probably going to use it, or we're very close to actually using it to write integrations right?

Like parts of our integration library, where like through web hook libraries, like giving people some turnkey stuff in essence that we haven't even had to engage with the 3rd party around. And that's the bit that starts to get like, really interesting about how we can just iterate to your point, kylie, like, does the product move too far from the customer? I think. AI, if it's used properly, gives us the opportunity to actually reduce that iteration with the customer. The product will always be wrong when it goes to the customer.

That's just like an assumption that has to be made. And it's like, okay, how can we? How can we push the product more sort of at a higher cadence to the customer to get that feedback rather than this massive release schedule. That's always wrong. And then you got this massive feedback schedule like we're pushing code every day into production, and a customer will always just see this like a slight change in the system.

And I think that for us like that works for us. But it may not work, for, you know, in the sense of sort of Leanne's business. I think you know, Banks might shit themselves if those broke out Banks might, you know, get a bit wary of all, like all the time. There's this sort of like change, even though it delivers the best outcome for them. And that's like, okay, how can you? How can we, Molly? Kind of the customer, basically, so that they understand and are comfortable with the outcomes that are being delivered to them, because that's really all that matters. The end of that. That's what they pay for.

Leann:

Absolutely.

David:

Yeah.

Leann:

And it's an interesting conversation, because, in fact, we've been using forms of AI for years. They just don't always call it that right. So we've been having a facial, you know, OCR labs and racial recognition, etc. I have been using it for years. But now it's AI. So part of our job is to demystify for our banks and lenders. And I mean frankly, sometimes I don't care what it's called. As long as it's doing the job it needs to do for our customer and user. But there is a distinction between how we demystify for our banks and lenders, because I think you elegantly put it. They do shoot themselves at times.

David:

I think the thing to remember here is when you talk about agentic AI and autonomous type AI, everyone thinks well, they think of it as a person, and that they can do anything and everything within the business. But I think you know that the early agentic AI is very, very isolated, and, you know, has a very small context. To try and make it, you know, easier to to handle the memory and all those types of things. So you'll have roles within the business which could be replaced by Agentica, but very, very narrow and very small job descriptions if you want to treat them like people. And that's again where the tools coming in will help the Agentic AI make those right decisions. Otherwise you need to, you know.

Think of the system prompting as your onboarding and training manual for an employee, you know, and it can only be X long before you run out of context, so you need to keep it as short and concise as possible, so that they don't go off and do the wrong thing. So yeah, and I just noticed a question from Darren on the chat, is there anyone out there building Proptech specific AI code or MCP servers that can be applied industry wide. Well, we are. We're gonna build. Not only are we building MCP service for obvious forms and agreements, we'll be doing other tools like address, matching property address, lookups, things like that which AI needs and particularly Agent AI needs down the track. Being able to do those things. So yeah, it's. But as I said, it's still very early days from that perspective.

Kylie:

So what are some of the things that as an industry we should be advocating? For in order to achieve best practice.

Damian:

For our customers, or for.

Kylie:

Well, yeah, for our custom. Well, so for ourselves. And then also because if we have standards and best practices and we communicate them, then our customers know what to look for when they're seeking best practice.

Leann:

I mean, I guess I would just reiterate that it's about looking under the hood and knowing who you're partnering with and what's happening to the data. Where is it happening? And what are the vulnerabilities there? And I think the comment earlier on was there was a comment earlier about testing it's a feedback loop. And my 3rd comment would be. I don't think AI lives with any one person anymore. We don't have an AI team or an AI expert in an organization. I think it's upon us all to be articulate and to be well versed in AI and application of it. So we can understand, and we can interpret it. We can scrutinize it and I, you know, when we look at it I heard a quote the other day that it is estimated, we spend 1 billion dollars a day on AI research and a quarter of the world's money. Wealth is reinvested into R&D on AI like it's not going anywhere, so it's incumbent on all of us to do and to be well versed in it.

Damian:

So there's incentive to sell the AI hype with that level of investment.

Kylie:

There we go!

Damian:

Yeah, from a, you know, customer's perspective. I think you know, we have a bit of a responsibility. We have a high responsibility to make sure we're not walking them into, you know, breaches of the privacy act or security breaches. And yeah, there's good best practice that we can do at a dev level like end to end encryption. And you know, making sure data is processed here in Australia and whatnot goes.

You know, there's some basic things you can do there. But from a customer's perspective, you know, I just, I reckon the context behind everything. Everyone's saying today, context is king right and garbage in garbage out. That's something that's been spoken about. We're all blue in the face for years and years and years, but it is for our customers.

The quality of outputs that they can expect to get from magentic platforms are going to be very much correlated to the quality of data that they're inputting. So I would encourage anyone in, say, the real estate world, to get really serious about data hygiene and best practices around that. And the more you can, and also, you know, when it comes to selecting your props, your tech stack, it's, I think, connection and connectivity is really highly important, you know, I think channel of choice is a good word.

And I think if you've used selecting systems that are closed and you can't even access your own data back. That's going to really limit your context window to actually provide the context needed to perform really high level problem solving agent functions. So that's something, I think is, you know, and I've seen that in the Proptech world that there is a movement towards channel of choice which is great, and I think anyone that sort of excludes themselves may do so at their own peril. Because I think customers, you know, I'm already seeing customers asking about that. So, yeah, I think that will.

Aonghus:

A totally different perspective here, if I can. So we run everything on our own infrastructure, like everything that sits in our platform that a customer has runs on our own. You know, it's cloud infrastructure, but it's on dedicated infrastructure. We probably run like 5 TB of computer, like its significant amount of compute would have like service running so like an enormous amount of, you know, of heavy sort of compute.

But if a customer wants to benefit from benchmarking, they must opt into a multi region model for that. And the reason that we do that is because it reduces the bias that you get within a region. So in Australia you know the way a customer manages, you know. They got it, you know, certainly pretty consistent among them. But if you are in Australia and you have a membrane roof. There are a tiny number of buildings that have membrane roofs in the US. Everyone's got membrane roofs. So actually for us to give a better product to our customer, we need that data to move across regions.

We don't have personally identifiable information. And so that is, that's a benefit to our setup. But certainly, you know, if a customer wants the benefit of the sort of the benchmark, and that's totally anonymized, like everything that is stored in a multi region. Sense, for a customer, is totally anonymous. The customer would know it's being stored in that way, etc. But if they said to us, Okay, we want it to be, you know, within a specific region. We say, Okay, but these are the restrictions that then happen on the product.

And unless there are regulatory requirements basically for a customer, very few would opt in for that. So example here in Australia is like defense, or, you know, critical infrastructure, they're obviously impacted. But there are ways that you know, I'm actually a big advocate of a multi region setup, in a sense, for a customer, because for a train customer being on an access in US Benchmark data sets open you up to like 50 times better inference. And so for a customer. I think it's a much better outcome.

And often they, you know, are terrified around where data moves. But I say, Okay, why are you terrified because it's moving in a place. It all runs on our own infrastructure. So I can tell you absolutely where any data that sits in the system, unless it unless the customer has a 3rd party integration. That data does not move off our system ever and so.

Damian:

That. So you'd be storing that here and then anonymizing it, though, if you push offshore right. So it makes it a lot easier like, that's the thing like, yeah.

Aonghus:

Yeah, absolutely, absolutely. Yeah. But I think, like, there's this skew. What needs to be done to be regulatory sort of compliant in a tech stack sense. But you know, in the sense of an Australian provider cloud here costs 30 or 40% more than it would in the U.S.

Region, and so you know, to like, all of us. Customers are served from us 1st instances. All Australian customers serve from Australian 1st instances, but where we run that sort of inference depends on the region that these customers are sitting in. But yeah, it costs a lot more here, and there's a lot more restrictions on what can be done.

And we sit across all crowd cloud providers. Basically. So like, you know, I think we have good exposure to it, in a sense, and I really appreciate when people have to all have it onshore, because that actually costs us a lot more money.

Damian:

Hmm.

Kylie:

So if you are, what should we be telling our clients they should be looking for? So we're telling them that they should have their data as clean as possible. They should be clear about where you guys are. You know where their data ends and where your data starts and the benefit, what the value exchange is by allowing us as Proptechs, to have access to their data sets. Is there anything else that we should be that they should be that they should be doing from their end to.

David:

I mean, and most of it is common sense, you know. No, no, but you know Sam Altman, what he came out with recently saying that every single conversation on ChatGPT is recorded and could be used against you. So if you're using a therapist. It's, you know, probably not the right use case. But yeah, you know. And that's the thing.

So AI, like Leanne said before AI is here, whether you like it or not, it is here, it's going to be used, it is being used. It's being used instead of Google. And all those types of things. It's going to infiltrate your platform, either whether it's going to simulate a person using your website, or if it's going to use your Api directly, so we are better off being as open and transparent box as possible to make sure that our customers can make the right decisions, because at the end of the day you can only control so much.

So, you know, we've got these MCP servers which will guide them in the right direction, but they can still hook it up to ChatGPT and you know, potentially send send that information to someone else, or you know that they right now they could get an agreement, and they could store it on a another server somewhere else, and it gets you know it. It gets infiltrated. So there's only so much we can do. But you know. Obviously, best practice is always the best way to go.

Aonghus:

That's and that's maybe where, as an association, the value to a customer like I don't think there's much value in AI standards, because I think that's a super reactive thing. But can we actually help customers buy AI smarter, you know, like, so when we do an IT assessment. 80% of them are based on on-prem software install. And it's like that doesn't mean a thing for 10 years like, why are we? Why are we getting these assessments? And I think that's like, how can we help our customers by AI like, what are the outcomes? It's almost like a better education, these events, you know, these sort of webinars, I think, actually really help, and maybe pull in some.

Leann:

Cameras.

Aonghus:

Customers going like this is how we saw value in a city. This is how we use it. We thought this was totally useless, and we didn't actually dive into this aspect of it. But we thought these other aspects were really great because it lets people understand the use case. What the outcome that delivered, you know, for them, and why that actually existed in an Roi sense.

Leann:

Can I add to that, Aonghus? I think one of the things that many organizations can do is to rethink how they do procurement and due diligence. If you think about the way that we used to do things, if you look at on prem servers, or every instance got its own server. So the way that businesses will, you know, a lot of the newer stacks and new tech companies are working, and I don't think procurement processes and due diligence processes caught up with that.

And, in fact, we cannot get a steer on whether we think there's going to be an adoption of AI or is something innovative, just by an organization's procurement processes and due diligence in terms of their risk, not their risk appetite, their change, appetite, their readiness for change and adoption of new tech.

Kylie:

That's really interesting. That could actually really cut short a whole lot of sales calls, couldn't it? You could basically say to people before they, you know, when they inquire about it. Couldn't we have a look at your procurement documentation and due diligence processes? And then basically say, yeah. Now, yeah, no, you can't buy us.

Leann:

I could pick the ones that we're if I look at their due diligence procurement, I can tell you exactly the ones that are ready to change and the ones that aren't. And you know, I think organizations need to be change ready. But that requires the education that requires everyone to have an understanding of AI, because it, you know, really isn't going anywhere, as we all know.

Kylie:

Yeah, Darren fries just shared that on-prem now, the equivalent of black and white TV. So I think that's very apt. Hey? I'm conscious of the time, we might just do a quick comment from everybody to wrap up. And so, Aonghus, do you want to go next? Like, what's your final thought?

Aonghus:

Oh, I think you know AI is exciting. But knowing why it's exciting is probably the thing that people need to need to understand, because I think it's exciting, for every time I find excitement is vastly different the way our customer does, but it's still much of a word, and we might be the same excited about it. So I think, knowing, knowing why it's exciting is probably the thing people should think about in the world. Maybe.

Kylie:

Yeah. The end.

Leann:

My final thoughts on. It's probably apropos, the elephant that's behind me. It's safe and secure, but a vision of the future. And for us. It's always staying ahead of the game. And I'm really curious about what's happening in the world of machine experiences. So I'll leave that for everyone to consider Mx alongside Ux and Cx. And all the other X's in the world. It'll be interesting to see what happens with AI with machine experience and machine to machine.

Kylie:

Damian.

Damian:

Yeah, it's exciting times. I would, I guess, encourage, you know, our customers to remain curious and and maybe test and and ask, familiarize yourself with with this space because it is coming, whether you like it or not, but you know, don't give in to the fear of missing out, or the fear of being replaced, and I think training getting that attitude embedded in your team so that they're, you know, welcoming or embracing change. But then, mapping those tools that you may be exploring to real world problems that you have. So that's what it's all about solving problems and doing things more efficiently and productively. You know. Yeah, that's what I would end with.

Kylie:

And yeah, David, last time.

David:

I look, I'm excited about what the Proptech community will come up with and that's I guess. Historically, you know, we form. It was the 1st ever open API forms platform in Australia, and we encouraged innovation from startups who, you know, didn't have to go through a Paywall, or those types of things. They're the things that I met when I 1st started my business. You know, I had to pay $10,000 to get access to an Api, all those types of things. And that's why we want to build these MCP servers so that we can encourage innovation.

So, you know, I might have one idea, but someone else might use this thing in a completely different way, which completely revolutionizes the way things happen. And we've never thought of it, you know. So, that's what I'm excited about, and I know it's early days, you know the toddlers. But some of it 's fun, it's fun as well. So, it’s very exciting.

Kylie:

Yeah, awesome. So look, guys, thank you so much for your time. Just a quick summary on some of the things that I've heard is that as Proptechs, we need to be putting our customer first.st Not AI first.st And we really need to be thinking about the use case through the view of what our customers is, we need to be creating clear boxes, not black boxes when it comes to AI, so that everyone is very clear on what's in there.

Our customers knowledge, and in fact, even their processes and systems, for the way that they do things at the moment is an important part of both the data that they're going to supply to us, and even whether they are going to be able to use us or not, or whether their practices are so out of date that they're going to struggle. We need to be really clear, I guess, with our customers on where their data starts and where our Proptech starts and where our Proptech ends. And so, therefore, what that kind of ecosystem looks like. We need to be really transparent around where data is processed, who we partner with and our clients themselves. I think what I've heard in all of this is what their policies are.

Oh, no, they need to be asking us questions. And we need to be really transparent around things like, what are our policies and protocols around data management around security, privacy, feedback loops where our data is hosted and why it's hosted there and be really kind of open around all of that stuff. What we've heard is that context is king. I love this idea of. Well, I don't love this idea, but I'm terrified of context. Rot! That's now my new favorite saying, and that the quality of the outputs is directly related to the quality of the data that we need to be a channel of choice.

And I think this idea around thinking about procurement processes and due diligence is, and whether our clients have got the right processes in place, and what we should be encouraging them to replace those with could be quite powerful. Here, too, I think one of the most important and exciting things that I hear about about AI, though, is that we've gone from having to think like the machine, to being able to have conversations with the machine, and often and what I see sort of generationally, and I find myself having to constantly think like this, too, is that it's no longer about trying to think about what should I?

Well, what question can I ask? And it's just an AI really opens us up to being curious, and so asking lots of questions, because now we get answers as opposed to having to think of the, you know, the right prompt to type in. And so I think this whole idea of like well, I don't know what I don't know, but I know I've got lots of questions, and not being afraid to treat the AI like it is a slightly stupid intern, because we do have to train them. And we have to have as Proptechs really strong training and helping our clients have, you know, strong training and processes for that.

So look, David, Damian, Leanne, Aonghus! Thank you so much for an awesome, awesome conversation. Thank you. Everybody on the call who joined us, and we will be recording this, or we have recorded this, and we will be sharing it out afterwards. It's been a great conversation. We will definitely be continuing it, and certainly following it up at the Proptech Forum in Melbourne in November. Stay tuned. We'll be releasing details on that soon. So thanks everybody, and we hope to see you at the end of next month for another cracking Proptech panel topic to come very shortly. Thanks, guys.

Leann:

Thanks. Everyone.

Aonghus:

You, too, guys. Bye.

Kylie:

Thanks.