Practical AI in Restoration: From Skeptic to First Steps
AI is nearly everywhere in restoration, but a lot of shops aren’t using it. The same four concerns come up again and again: “I don’t trust it,” “not another login,” “it just doesn’t sit right with me,” and “it costs too much.”
This workshop takes these concerns very seriously and covers where AI earns that skepticism, what it’s actually good for, and where human expertise is now even more important. Attendees will leave this session with the first steps to try AI safely, starting with the job documentation your team already creates.
Alrighty. Let’s get into it. So first of all, welcome to Practical AI in Restoration. I’m Kate, from Encircle here. So I’m going to be handling the boring yet important stuff before we get into it. So if you haven’t already, pop into the chat, say hi, tell us where you’re tuning in from. Okay. So why are we here today? And that… It really comes down to AI being a part of a lot of conversations right now, and that includes in the restoration space. And at at Encircle, we talk to a lot of restorers every single day. And from what we’re hearing, a lot of shops aren’t using it in a very restoration specific way. So many shops have a tech or a PM that use ChatGPT on the side, or maybe you’re using Gemini to draft emails, but most aren’t really taking it much deeper than that. And we’re hearing a lot of really valid reasons why, techs aren’t taking the plunge into AI. I am going to launch a poll here. Alright. So I… What I really wanna know is what is the biggest thing holding back your shop from using AI more? Some pretty interesting answers so far. A lot of people are already using it. Yeah. Okay. Alright. Let’s check on this… On the results here. Okay. So pretty interesting. So not trusting the output and not really knowing where to start. That’s a that’s a pretty… Well, we have a lot to to talk to you about today then. Okay. And without further ado, let me introduce you to today’s experts. So we have Jeff McDowell. So he is Encircle’s Chief Product Officer. Jeff has about 30 years experience in the tech industry, including as COO of Primal AI. For the last five years, he’s been putting that experience to work on some of restoration’s toughest problems here at Encircle. And we also have Joe Ledbetter, our Principal AI Innovator. Joe has more than 25 years of experience in residential and large restoration. So today, he is the restoration brain guiding and building Encircle’s AI tools. So without further ado, Jeff, kick us off. Awesome. Thanks, Kate. Really appreciate everybody being here today. This is actually a a webinar that Joe and I have… Oh, hey, Joe. I should say hi. Hello. This is a a webinar that Joe and I have have put on previously in person at a conference. It was four hours long. We have certainly cut it down. We we learned a lot by putting on that webinar of what people were interested in hearing about and what they weren’t. And, but we’re happy to to be here today talking to you about AI. I think as Kate mentioned, this isn’t typically… You know, we do a lot of… Encircle does a lot of webinars. Some are product focused, some are, product launch focused, new things that we’ve built. This is really more just helping people understand AI and get into it and use it to the best of its ability, what to do, what not to do, what to avoid, what to think about. And I do believe that there’s a lot to cover here, and, I’m excited about it. So, you know, I guess we can we can get into it. Joe, did you have any sort of opening comments, or should we just dive into the to the agenda? I say we dive right into it. I wanna get to the question at the end. So well, yeah. Oh, that’s right. Both Joe and I have spent a ton of time with customers, people in the industry. And, the way we actually structured today’s, chat was around the questions that we get asked most often, and they happen to be the same questions that were… That Kate just asked in the poll. And I thought that would be a great way to structure the presentation. So the thing that we hear all the time is I don’t trust it, and I do believe that was the second highest objection that people have. Not another login. That’s… That was that was actually less down on the list in our poll there, but that’s definitely something that we hear. The tech stack’s already gotten too complicated. I’ve already got too many apps. I’ve already got too many logins. You know, not looking for another one. I don’t know what I don’t know, or I don’t know where to start. I think everybody probably understands that AI is extremely topical. I think they… To a degree, they understand that there’s a lot to be gained and a lot of benefit to to adopt AI, but I don’t even know where to start. And I certainly don’t know what I don’t know about AI, so that’s another thing that we hear quite frequently. And then the third thing, it costs too much. And I think that’s a fair… It can be expensive, and I think there’s a there’s a fair comment here. But what I hope we are able to provide you today is is just ways to think about those costs, how to contain those costs, how to make sure that they don’t… You don’t get into a cost overrun. But most of all, understand how the cost of AI drive a a return on investment. And then… And like any software stack, we always try to to, you know, help our customers understand that the return that they should be looking for should be about 10 x of what you’re of what you’re paying. So, hopefully, we can provide some tips today on, you know, what is… How is it that you can use AI in a way that you should see about 10 times your return on investment in AI? I think you can go to the next slide there, Kate. So as I said, like, this is probably, you know, one of the top questions that comes up over and over is about trust. And I don’t I don’t trust that I’ve I’ve heard horror stories, or I’ve heard stories. I’ve I’ve got a story we’ll get to in a little bit. But I thought this was a this was a really good place to start. And the first pattern we see… You go to the next slide. The first pattern we see… And, you know, sometimes I wish the… I I love doing webinars, but the one thing I miss is being able to see everybody’s head nod because I have a feeling that this might be one of those head nodding moments here that I know that I experienced when I first started using large language models years and years ago is this wow wait pattern. What is the wow wait pattern? I think we’ve… I think the the poll said about 42% of the folks here have already started using AI, and I am certain you have come across this. You asked it to do something, whether it’s writing a plan or writing a scope or writing a brief or rewriting an email, and you went ahead. You went into a chatbot. You went into a typical language model chatbot, and you asked it to do something, and it provided you an output, a document, a plan. And when you looked at it, it was written really, really well. And you’re like, wow. This is amazing. I can’t believe this. This is gonna change everything. The grammar was perfect. The spelling was perfect. The way it flowed was, you know, immaculate. But then you read it again, and you start to really analyze what it’s talking about. And something seems off to you. You’re like, that doesn’t sound right. Or maybe you read an entire paragraph, and you’re like, I’m not sure that entire paragraph had a point. So then you read it a third time, and then you get into the details. And now that you’re familiar with what it’s written, you’re actually finding things that are just completely wrong. It’s misquoted you, or it’s it’s misunderstood what you were asking it, or it just completely hallucinated and and and just provided you a wrong answer. But the amazing thing is, and I still go back to the fact that the first time you read it, it just seemed perfect. So, you know, the the point here is as you’re dipping your toe under the water, as you’re getting into these things, as you’re understanding AI, you know, never judge AI output on the first read. Know, You in in in in the case of restoration, check it against the photos, check check it against your notes, the readings, you know, and make sure that everything that the output says is accurate and is what you’re looking for. So you can go to the next slide. So I’m gonna tell the story. Joe, I didn’t tell you I was doing this. Kate, I didn’t tell you I was doing this. In fact, this slide… Please excuse this slide. This wasn’t built by any of our marketers. I just stuck this in here at the last minute because I felt it was important. I told this story despite how bad it makes me look given I’ve been an AI for a minimum of 10 years now. So here’s the story. I… And this really demonstrates what I was talking about previously. Just about three weeks ago, my wife and I were out on a walk, walking along in a forest as we do, and I kept seeing this thing. You’ll see it in my hand there, this pretty ball looking thing, and I would often find it amongst a number of acorns on the ground. And I thought, I’ve never seen this before. So I picked it up, and I stuck it in my pocket, and I and I went home with it. And then I pulled out my phone when I got back home. And as you can see on the left, I asked it a question. I said, what is this seed looking thing that I came across in the forest? And I was using Grok at the time, although I’m not here to promote or detract from any one model, but that’s just happened to what I use personally. And Grok told me it’s a wild plum or likely a Canada plum, which I thought, cool. Something I’ve just learned new today. I’ve been in Canada my entire life. I’ve never heard of a Canada plum. It then went on to tell me that Canada plums can be used in jams. They have a tart skin and a sweet inside, which based on how well it was written, for some reason, compelled me to bite into it. And it was in that moment that I realized the mistake that I had made when my lips were set on fire, my tongue was vibrating like I just put a thousand volts through it, and I couldn’t control the amount of saliva that was… And I’ll spare you the details falling out of my mouth Yep. That’s fine. And, the pain that I was enduring and my throat closing up. And after about half an hour of flushing out my mouth with water and milk and everything else that I could find, I finally got control of myself. I realized that I didn’t have to go to the hospital despite the fact that a couple times I thought I may. And I pulled out my phone again, and I pulled out Gemini this time, although I don’t think it was, again, relevant which model I used. I didn’t ask about what kind of seed it was. I think I realized what my mistake was is that I misled the model by being specific about it being a seed. And I said, what is this I found in a forest where then Gemini proceeded to tell me that that is an oak apple gall, which is the larva of a wasp that lays its eggs in the bud of an acorn, genetically chain… Mutating that acorn into what you see in my hand. It, builds out a protective layer, an astringent that then protects it from being eaten by forest animals like squirrels chipmunks and so on. Despite this being an extremely embarrassing story, I think it’s relevant. The first time I put this in, the the the the result that I got couldn’t have been more confident. This is a wild plum, a Canada plum. I can’t believe I allowed myself to be fooled by this. You know? My 50 years spent in Canada, I’ve never crop… Come across a Canada plum. However, it seems like I just found one on the forest floor and decided to take a bite out of it. So I don’t know. Crazy story, somewhat embarrassing. However, my folly is everyone else’s learning experience, and I do believe there’s actually something to be learned here. Will I do this again? No. Do I think about this in every time I interact with AI? Yes. And so I wanted to tell that story. Joe couldn’t stop laughing the first time I told him, and I’m glad this wasn’t the first time he heard it because I don’t think he’d be able to continue down this path with me on this webinar if if this was the first time he heard it. Kate, I believe you can go on to the next slide. Unless, Joe, you wanna comment on my folly? No. I just… I I think… I mean, we’re so… We should all be thankful for you to share that, humbly. That is an amazing story, but it really points at a deeper truth. Right? Like, it points at something that we should all be aware of, and that is the the wait moment. The… We got this output from an LLM or an AI, but, like, something’s not quite right. Right? And the red flag for Jeff, well, it should have been the the Canadian plum that’s never existed. But, again, like, we’re we’re we’re… It’s so easy to to read it and feel confident that it’s right. You know? And that’s that’s one of the things we’re gonna be talking about today. So I I… It couldn’t have been a better story at the end of the day. Yeah. So where do we start? So, you know, something you hear a lot is human in loop despite I was in the loop in the last, but it… But it’s… But I do want to… You know, this is something that you’ll hear a lot from a lot of folks. AI certainly does a lot of things well. But we do wanna point out that none of it’s possible without that human diligence. You know, something that you can be confident where it needs a human is in… You know, to to dial it into a specific restoration scenario. You know, psychrometrics, drying calls that require certified expertise, cause a loss, data collection, like getting on-site. There’s nothing AI is gonna do. I’m not gonna predict robots and all of that stuff that you see on TV. But for… As far as I can tell, you’re still gonna need an expert human and a technician to show up on-site to do an inspection and figure out what the cause of loss is. You know, anything sent to a carrier needs sign off and accountability internally by a human. Equipment placement needs to be checked, containment, safety decisions. All of these things can be inferred or suggested by AI, but a human really needs to apply their, their expertise against it. Yeah. Yeah. And their and their physical attribute. Right? Like, the AI is not gonna come in and, you know, determine the cause of loss. Right? Half the restores can’t determine the cause of loss on the initial inspection. Right? So, like, AI is not solving that for us. That’s something a human ultimately has to arrive at. But, you know, the things that the AI can do really well in our workflow of restores is it can clean up notes. It can take a a really messy technician note and clean it up and make it more readable, legible, and proper grammar. Right? It can organize photos. It can describe things from us for photos. Right? It it it can do translation. Right? What a great what a great tool to be able to translate between two different colleagues or a customer. Right? AI is really good at spotting patterns in the data, right, when the data is there. And and also, it’s really good at, like, just keeping consistency within those types of communications. But at the end of the day, it all starts with the human. And when we talk about this, right, we talk about human in the loop. The reality is it’s it’s a it’s a… It should be human circle because it starts with a human, and then AI gets involved, and then it comes back to the human for review. Well, without the human doing the input and the review, AI… There’s nothing for AI to do at that point. Right? So… You can move on, Kate. Alright. So, you know, one of the things I wanna talk about today, there’s there’s a number of different AI solutions out there. There’s your your chatbots, your general large language models, Claude and Gemini and OpenAI or… Sorry. ChatGPT. And… But then there’s there’s purpose built AI. Right? Stuff that companies like Encircle have actually built into their software. There’s… In fact, I would say all of the different solutions in our ecosystem have probably built it in at some point. And we just wanna, you know, provide some insight into the difference between a purpose built model and something that’s a little bit more off the shelf and a little more general and how you should think about these different things. Because I think there’s use cases for both for both sides. We’ll talk a little bit later about getting your toe in the water. And, you know, definitely, the the language models and the chatbots are a great place to start. But the purpose built models certainly, provide a lot more value and a lot more certainty and a lot more trustability. So, you know, we’ve we’ve we’ve actually tested it ourselves against some of the the large language models, taken a ton of photos, ton of readings, ton of notes, and we’ve put it into one of these off the shelf models. First of all, I think the first thing worth mentioning is just the cost of that. You know, you can do a lot of stuff with these models for a fairly low price, but when you get into computer vision or or photo detection, photo analysis, it really can drive the price up quickly when it when it has to do complex calculations of, you know, equipment calculations, things like that. It can really run out the cost. So we were seeing, you know, costs up to $50 for one file, which is not something that anybody who who who builds a a a purpose built scoping solution would ever charge. I don’t think it would be as high as $50. But that’s something that we saw. The output, you know, we we… A grade c at best. There’s so much that goes into the prompts. We have over two years of engineering into our entire language model and prompting methodology to make it restoration specific. You can get some output that looks pretty good, but if I’ve referred to my last slide around the wait wow… The wow wait moment, things that often look good at first fail the test as you go through it, and you realize it started missing some things or misrepresented some things. So from what we saw, it was about a grade c at best. And then the other thing is is that when you’re working within a a a chatbot, it’s not really repeatable. You know, you can you can sorta get locked in on on writing these prompts, and, you’ll realize after some time, you wrote the prompt in a very specific way for that specific job that may or may not be transferable to other jobs. So, you know, often, you can get to an output. And even if it does look pretty good or it is pretty good, it’s not necessarily repeatable. So we just wanted to make sure that people understood that the chatbots are a great place to to dip your toe in the water. However, at the end of the day, looking for purpose built solutions in the industry that have been tested, that have been built and engineered with restoration specific workflows in mind, we do believe is the… And and that… And you just can’t replace good domain knowledge. So we please… That’s the place to start. Yep. We do have a white paper on it, actually. Joe, you worked on that, didn’t you? Did you? Yep. Yeah. So we have a white paper that, I believe what we can provide everybody. And and I I just wanna add to that. You know? Like, that was us trying to do the thing. Right? And so, like, I I feel for a restorer who may not know the technology attribute of what it takes to build even something in a simple chat. And so I would speculate the cost would even be higher on average and the repeatability be even less. It’s it’s a very complicated thing to try to attempt and very costly. Yeah. I think you had some… I think, Joe, you wanted to take this one. Yeah? Yeah. Yeah. So this this one is… This one’s interesting. Right? And it’s because the the whole premise here is that if you have three estimators and you give them all the same project and you say, hey, guys. I want you to go estimate these on your own. The definitive result, if you haven’t already done this and tried this in your own business, is you’re gonna get three different responses. You’re gonna get three different estimates. Now are they gonna be completely different? Most likely not. Are they gonna be fairly aligned with, what with what that estimator has been writing prior? Absolutely. But going to be differences. It could be, you know, two, three line items here, four line items there, five line items there. And where does that where does that delta come from? It comes from something painfully that we have to accept of of normalized inefficiencies. And it’s not that any of the estimators per se are better or worse than the other. Right? Some may have more experience and knowledge, but, basically, they all are forced to do the same thing on every project. And what is that? Fill the gaps from the field data. So, oh, they don’t have a picture of containment, but I’m gonna write the line item. That’s what this estimator did. This estimator decided because he didn’t have the photo, he was just gonna increase his labor hours somewhere else. And then the third estimator did a completely third different thing. Ultimately, the variances that you see isn’t just because of the estimator’s experience. It’s ultimately normally just dealing with the inefficiencies that we’ve come to accept, and so they’re forced to ask the field. And so this matters because this is a great place where AI can actually help you be more consistent. Right? And and clearly, starts with collecting that data in the field, but this is where AI has the ability to help you give the right inputs, right, so that you have the information. So then now the estimators actually all know the same things. This is where it becomes more powerful because now you can kinda level the playing field and ensure consistency. So, again, if you… If you’re interested in trying it, have your estimators write the estimates, see the differences, and then ask why. Why is yours this way? Why is yours this way? And what you’ll uncover, again, is is often just the the inefficiencies that we’ve accepted in our industry. Again, this is where we believe AI now is is poised to help solve that problem and bring a level of consistency that you can trust and that is dependable. Yep. You know? Yeah. Yeah. AI is not going to infer anything that it’s not supposed to infer. It’s not gonna make anything up. And, again, not to say that estimators do, but they could have different experiences that lead to different conclusions. And, you know, if they’re using this AI model, it can it can do two things. It can grade that consistent output, but it can also identify what’s missing. It can… And and instead of having to, like, you know, infer something or or guess at something, you know, AI should be able to tell you, hey. This specific reading was missing from the file, and this needs to be completed before we we move onwards. So… But definitely consistency is is key. Yep. You know, trust is earned in stages. So this is another thing that’s that’s, I think, quite important. And another thing that I would say we had to learn, that this was something just through our own journey of implementing AI in our product that we’ve sort of learned where trust can come from. You know, start small. Start with smaller tasks. AI can draft human rewrites. You know, I used to… When language models first came around and you were just… They were almost a novelty, this is, two, three years ago or three, four years ago, I remember using them because I would get them to rewrite an email that I wanted to send, and I would get them to rewrite it in Shakespearean prose or doctor Zeus. And and while that was kind of a silly use case, it it it just sort of helped me realize that, you know, I can use it for productive reasons as well, and and it it helped me sort of get my head around how it could actually help my communication. Sometimes life can be busy. I can write a an email to my employees who they ultimately get back to say, I I have no idea what you’re talking about here. And then I reread it. I’m like, I was, like, walking out of the biz… The the building to my car while I wrote that, and I just realized it was completely incomprehensive, and I probably should have had something to read that first. And I started to do that, and and and things… You know, you you immediately see that value when you now write communications that are completely clear, and they’ve been nicely rewritten for you with with proper grammar and punctuation. And believe me, my my teams appreciate that as well. So that’s where I started. You know, I could build a to do to do list. I would often record meetings that I was in. We we use Google Meet internally, and you can record the meeting and have it write a to do list of the things that you talked about. So that was, that was a great place to start, and I really do see that as as an assistance. It’s it’s using AI to assist you, and that definitely helped me get my head around what it was good at and what it wasn’t. Augment would be the next step, and I would say that’s probably where a lot of the tools are today in in the industry. It’s certainly where I would say we are at Encircle. That that human in the loop or that human circle, as Joe had mentioned earlier, basically, you take a number of inputs that are collected by the field, photos, notes, moisture readings, dimensions. And then in in in in a particular case, maybe you wanted to generate a scope, but then a human needs to review it. And the human goes through it, makes sure… Validates it against the photos and make sure everything’s accurate, and then it can be approved. So that’s what I would call human augmentation. And and I would say the last step, and I don’t… I’m not gonna say that we’re there yet. It’s something we certainly think about. It’s something that when our customers get their head around what we’ve built in AI, they immediately say, hey. Can’t this be automated completely? And we say, yes. It can. We will certainly get there. But automation is is the the sort of the next step for AI. You’ve probably also heard of the term agentic AI. That’s where this would play. Basically, AI can act. It can reason, and it can actually take action, and then humans only get involved when, there’s an exception to the rule. So, you know, Joe and I were bantering about, and and you can jump in anytime, Joe, but, like, what a practical application of this is. We talked about, you know, a job comes into your into your shop. You know, a statement of work needs to be generated in that customer’s name. It needs to get signed, and then a whole lot of things happen downstream from that after that statement of work gets unsigned. That job goes into production. You know, people gets… People people show up at the at the at the site, do their work. But you have this whole layer of communication to the adjuster and to the carrier that needs to happen. And right now, we call it dumpster diving. A whole bunch of dumpster diving has to happen. You go into the claims, you go into your CRM or your job management, and you’re pulling out milestones, and you’re trying to collect them all so that you can put a nice report together, well, there will be a time where all that’s just gonna happen automatically. You know, statement of work gets signed, the job goes into production, the customer, is contacted. All of the milestones, all of the data that gets collected along the lines, longer people don’t… Won’t be needing to dumpster dive. It’ll organize itself. It’ll create a report, and and and once it’s quickly reviewed, it can send itself off to where it needs to go, whether that’s an adjuster or, you know, work order is going to subcontractors, so on and so forth. So I would say that’s something to look to the future. I think there’s gonna be a ton of productivity gains there, but I really do think that augmentation is where most of the industry is at. Yeah. Can go ahead, Joe. I know this is a favorite one of yours. Yeah. So hallucinations. Right? So since the dawn of AI, we’ve had, this conversation happens a lot. Lot. And what’s interesting about it is there’s actually a lot of good red flags, if you will, that are really obvious when you’re reading an output from an LLM or an AI chatbot of some nature that can kinda tell you if it’s if it’s a hallucination, a lie, right, a misleading thing, or if it’s actually factual. And so here’s some really good examples. Right? So the the output is standard drying time is two days for this loss. That’s… There’s a lot of red flags about that. Let’s start with the fact that there’s no validation of an input. Right? So what what the output should state is something to the fact, like, this is a category two loss that’s been drying for three days. Based on the atmospheric readings, there’s probably one or two days left of drying. Right? That’s an accurate output. Why? Because there was substance at the beginning of it. Something where there’s nothing nothing specific at the beginning is clearly a red flag that it’s just an answer. There’s no there’s no proof behind… There’s no… Nothing driving its response. Right? Another real common one is when it uses words like typically or usually. Those are very different terms. Like, is standard really relevant in this case? There’s no standard drying time. No. And, again, to those who’ve dry… But to a new technician, you know, standard sounds official. Right? Sounds like Canada apple. Yeah. Like… Exactly. So, you know… And that becomes, so they… Those words are vague. Right? And then more importantly, they often will just contradict the actual conditions. Like, it’ll say something that’s not even at, like, a real thing. It’ll mention, you know, the the kitchen sink, but you’re in a you’re in a bathroom. Right? And so, yeah, there’s a sink, but it’s not the kitchen sink. You know? And so you you… The human in the loop is important, and it’s not because all the tools hallucinate, but it is a compute. So you should be responsible to read the thing. Now the difference, will tell you, is hallucinations often really only exist when you don’t have guardrails and you don’t have systems and you don’t have processes, when it’s not purpose built. The more purpose built it is, the more guardrails you have. Hallucinations really stops becoming a thing, that is an issue. But these hallucinations are things that you’re gonna find just using Claude or Chad GPT or Gemini or any of the open, you know, frontier models because they’re just generally not purpose built, and they have a deterministic want to satisfy you, right, at the end of the day. So hallucinations can be real, but they’re a lot less relevant when you’re in a purpose built system. That’s that’s for sure. But, you know, I think I think when we look at the the the attribute of trust, right, Jeff, like, you have to start small. Right? And when we look at the poll, yeah, there’s 40% of you that have already basically put your toe in the water. Right? And we understand that. But for the rest of you, this is the time. Right? If you wanna be able to trust AI, you have to start with baby steps and work your way into it. You have to know what you’re reading. You have to know what the outputs are. You have to really own that outcome because… Okay. Go to the next slide. It’s Sorry. It’s it’s really important, you know, that you’re able to, at this point, just start to do some basic things. And and the best way to to start is just doing simple things with, like, text, cleaning up your job notes. Right? Like, you’re making job notes, use an LLM to make them grammatically and and read correctly, right, without altercating things. Right? Use it to translate. I I love this one. You know, we’ve all been faced with a a knock on a customer’s door at 2AM. We’re reading off of our phones, and we can’t even pronounce the customer’s name, and they answer the door in a different language. Like, okay. Like, that’s a great use of an LLM. I’m gonna I’m gonna use that to interact and and have the conversation. I got a colleague who who speaks bilingual, but they prefer to speak Spanish. Let them speak Spanish and then use that l m to translate. Right? That’s a great way to maintain communications. More importantly, you gotta already communicate with your customer and your adjusters and everyone else. This is a great place to use an LLM to take the message, the email, the information you wanted to communicate, and have an LLM process it and have it more professionally organized, right, so that your presentation is always crisp and clean. And and these simple tasks, they don’t require you to be some massive prompt engineer or know how to use AI. Right? You can simply just talk into your to your chat and and say, I’m looking to make an email. This is the information. And then it’s gonna give you the email context. Read it. Understand it. Copy paste information into it. Give it more more context. But these are great ways to, like, do this. Another one that I’ve that I’ve really come to enjoy is the to do list. Right? So, like, if you’re really just getting started with this, again, open up your your LLM of choice and give it a bunch of tasks and say organize it into a a to do list. And then you’ll be really happy with what it looks like and and how it can be a functional thing that you can share to other people. These are great ways to start. And why are they great? Because the output is simple. It’s not complicated. It’s an output that you, for sure, can review it, understand it, comprehend it, master it, own it, and then go on to the next bigger task. Right? And then find ways to incorporate that into your daily routine, and then now you’re gaining that efficiency. Right? Key, though, is that you you own that output. Right? And so… So just some ground rules here, though, that, you know, we’ll talk a bit about private information a little bit later. But just as we’re talking about dipping your toe in the water, just be mindful of keeping customer names, addresses, claim numbers, anything that’s personally identifiable. Make sure you keep them out of those chatbots. You know, that’s just not information you want out there getting into the general, well, you don’t want it getting out into the general Internet, you don’t want them helping to train their models based off your private information. There… That is a main difference around purposeful tools in the chatbots as well is that anything that’s sort of purpose built and and with with any level of domain expertise, they should be able to provide you with their privacy policy on how the data is being used. So just something to be mindful of as you’re as you’re experimenting with some of these tools. You can move on, Kate. So not another login. As I said, we we we won’t… We don’t spend a ton of time on this, but, it’s definitely something that comes up all the time. You know, customers… And I I think this might be just sort of more, in general. Like, there’s just a lot of new apps, hitting the market. Text acts are getting more complicated and, you know, more and more logins to remember. So, just to address… You know, just to quickly address this, you can move on, Kate. You know, there’s there’s certainly a fatigue in the market with tools. Everybody… You know, a lot of new tools coming on board. AI has made it easier than ever to actually build these tools. But as I… You know, Joe and I both talked about using general ChatGPT and other things to dip your toe in the water. We do think a lot of the platforms are doing a a a really nice job in adding AI into their applications in a in a smart way where you don’t necessarily have to look to other applications for AI. Some of them ourselves, we do have a number of free AI little tools in our app that are just part of our basic subscription. Others are… Certainly require a different subscription level. But you you can use some AI features in in a lot of apps just as part of the the basic package. So, you know, use what you have available to you. I would hope that most applications out there have a trial period, you know, the ability to use that feature for some period of time for free using your own data against it. You wanna… You know, you don’t wanna get into copying, pasting, and things like that. And, again, that’s just more more more places for errors. So I would… You know, advice here is is look and almost expect tools that you are currently using to do a much better job at building AI internally to them. You can have an assumption that they’re they’re purpose built, and they’re they’re they’re built within a lot of domain expertise. So, you know, again, it’s inevitable that you’re gonna have a ton of different things out there in the market to look at. There’s certainly a lot of interesting stuff out there right now, but we we do we do hear this a lot that people are are getting tired of of applications and separate logins. And and, really, all you can do is make sure the vendors that you’re aligned with are building the the AI tools that you expect into the platform. So don’t know what I don’t know. This is probably when we went into building this and and when Joe and I were doing this, this webinar or… Sorry. Actually, did an actual seminar in person. This was probably the area that mow… Where most people were at was I know I I wanna jump into AI. I wanna take advantage of it. I heard it can help my business. I heard it can help profitability, but I don’t even know where to start. And there’s so much about it that I just can’t even get my head around. You know, where do I fit when AI shows up? Is this coming after my job? What happens to my customers’ data? What’s the future looking like? Like, everything’s moving so fast that it’s hard to keep track of. So let’s let’s tackle some of the… Some of these areas here. So the first one that, to me, is one of the most important things with AI is… And if you saw… I think we did a little teaser video on LinkedIn, and and I was talking about garbage in, garbage out because this is where I see just a lot of the a lot of the problems that people run into with AI and whether they think it’s valuable or not. Small stuff compounds, clean documentation is what makes AI work. So whether you’re documenting a loss or or, trying to figure out whether you should eat something you found on a forest floor, the inputs matter. I prompted Grok wrong, and it gave me a wrong answer. A lot of cases, in in the real world in restoration, AI can provide wrong answers when it doesn’t have the right input, when you don’t have a complete view of photos, you don’t have proper notes that were taken, all the moisture readings weren’t in, the dimensions were not in or incorrect. And, you know, we do get a lot of people telling us, well, that that creates a lot of work in the field. The thing is is that that work has to be done anyway. Joe talked about normalized inefficiency. That’s something we talk a lot about. Because when our customers talk to us about, hey. This is requiring us to put a lot more diligence in upfront and get our text to take the right photos and the right measurements and the right notes. We we we feel it’s like… That’s more work, not less, and and that’s actually not true. At some point, all of that data gets in. It’s just… It takes days where… Whether it’s the estimator or the claim administrator or somebody in the office calling the field saying, what did you do with this photo? Where are these readings? That that data does come in at some point. But if it doesn’t, it it it will create an incorrect output. And, but it would with an estimator as well. So the funny thing is is that this isn’t this isn’t really an AI problem. I think I think AI certainly catalyzes it a lot more because at this point in time, it doesn’t have the ability to pick up the phone and call a technician and ask for that missing photo. I think when we get into more agentic workflows, it will. But for now, it does require a little bit more diligence upfront, but the returns are amazing. Like, the the returns, once you’ve done the proper inputs and have collected all that data, you know, so much… AI can then take over and build your scope, your estimate, your work orders, your your your budgets, your your, your your cost accounting on a on a job. And so we believe, you know, you really wanna focus on the inputs and do know that you you can’t just take an incomplete file or an incomplete set of of of photos and job documentation and expect a good result. Ai is not magic by any means. So how do I engage AI? I don’t know, Joe. You tell the story better than I do. Do you want… Do you wanna tackle this one? Yeah. I mean, the best way to do it, guys, is you’re the experts. AI is not coming to take your job. AI is not showing up to tear out the drywall tomorrow. Maybe robots in 20 years, I don’t know. But tomorrow, it’s you. Right? You’re you’re the one who’s gonna be in the field. It’s it’s the human who’s gonna collect the data. And and capturing the information, like Jeff mentioned, delivers 10 x, you know, as you go through the project with AI. But at the end of the day, it starts with your expertise, your domain knowledge. You know, you’re the one who knows what’s wet in the building. Right? What really can be done is AI can do the messy part in the middle. Right? And and the thing that you don’t wanna do. And a great example of this is contents. I mean, an amazing example of this is our contents tool. And I know we’re we’re not supposed to talk about our products, but at the end of the day, you take a picture and it describes it. It does all the heavy lifting for you, and you just move on to the next item. And, and and historically, that’s hours of minutiae of work that nobody really wants to do. And, you know, AI just did it all in real time for you, and now you can sit back with your expertise and read it in about five minutes and know that you’re completed with the task that normally took you six hours. Now what’s interesting about that is you’d be like, well, then that means you get to work less. Funny enough, no. What it actually means is you can actually now apply the balance of your time towards the things that actually really matter. Right? Like, communicating with your customer, cleaning the actual contents, you know, communicating with the adjuster, making collections, all the other things that we really should be doing, to where now this is a great solution where AI can actually, like, solve a problem. So it’s really about… Everything needs to start with you, and then there’s places where AI can function and operate as a layer within that role. And then, again, it comes back to you as the expert where you have to do that review. More importantly, you have to own what it says. Right? Because that’s that’s your work product. You had a great tool help you do it, but it’s still your responsibility. You know? I I I just… It’s a great way to just start this process, you know, knowing with the right mindset that you are the owner of it. Yeah. K. So PII, there’s a couple boring slides here. I’m gonna be upfront with that. We have to talk about it. It’s certainly something everybody should know about. I think it is probably something everybody has heard about at the very least, but PII, basically, pry… Private identical… I… Identifying information… Privately identifying information… Personally identifying information. I don’t even know what PII stands for. I’m kidding. This is things in a job file that can contain homeowner names, addresses, phone numbers, claim and policy information, any information that can tie back to a… To I… The identification of of an individual. And this is an area that regular… Regulatory in North America is is sort of increasingly getting more increasingly stringent, where in Europe, it’s it’s been actually quite locked down for quite some time, and rules have been in place for… Around that data for some time as well. So this is where you don’t wanna get into pasting full job files with addresses and names into a into a language model. You know, they’ll all say that they don’t use the information, especially if… Well, a free version, they will tell you that they use the information. If it’s paid, it’s a little bit cloudy. You again, know, go one further to a a a… An actual built AI product by a software company, and they’ll they’ll certainly have a lot more rules against how that data is being used. But when you get using the chatbots, it can be kinda fuzzy on how they’re using that data. So you just wanna avoid it. Keep customer data out of the public chat tools. Any other tools that you are using, just ask the question, and we’ll go through a checklist later of different questions you should ask as you’re evaluating vendors. But ask the question about how long that data is stored for. Ask whether it trains models. That’s something we absolutely do not do. We we never use data for for training models and only use it approved… Unapproved, only use approved tools and real jobs. We’ll get into, just sort of the governance around using AI tools as well. Not the most exciting of topics, but nonetheless, as you’re enjoying the benefits of using AI, you have to be thinking about these things. You can move on, Kate. I I talked a little bit about regulation. As I said, in North America, it’s still sort of evolving and coming along. It’s probably worth getting organized before we actually… Where it actually starts coming down. There are some states where people are required to disclose when and how AI was used in a legal proceeding, you know, if if notes were summarized, so on and so forth. So there has to be record that AI was being used. I that’s think a fairly low effort regulation that’s in place, and that’s only gonna sort of get more more more structured from here. Not not a ton aimed at restores today. It’s more AI regulation in a general sense. I did wanna relate it to our specific industry, though. You know, there were times where there were no rules around or policies around lead and asbestos. I think the the good shops and and the operators out there probably started applying their own rules in handling asbestos before rules really came down. And once the rules did come down, there was a lot of shops out there that, you know, weren’t prepared and didn’t have the time to put things in place, and that’s generally where the fines were showing up. It’s those companies that just left it a little too late. AI will be the same. So, you know, again, you wanna you wanna think about these things and and get ahead of them. And some of the things that you can do, just write down how your shop uses AI. Just have a record of it. Again, it’s not something you need to spend a ton of time on. But the degree that you know where AI is being used, you should have a record of that. Know where AI is touching each job file. Is it just summarizing notes? Is it translating data, or is it writing your entire scope for you? Record, record who reviewed and approved each output. This actually should be built into whatever software you invest in. This is the kind of audit that you should expect, but definitely a record of who was, on that job file and who reviewed it. You wanna be able to explain… Like, again, if you used it for scoping, you wanna be able to explain that scope, not just… And this is something I’ve seen a lot of time. You probably have heard the term AI slop. It’s something that we’ve had to even on our own company get our heads around where people will generate a document and then send it off. Again, wasn’t reviewed, and and and and and the information contained wasn’t really digested. And I think that’s almost a bigger sin because you’ll stop learning if you’re not really reading the… Your output. So you definitely want to, not just generate a scope, but go through it and make sure that it’s well understood. And then revisit your policy as the rules change. Like, that’s just… As as things evolve, you always have to revisit things. You can move on. I’m boring myself as I talk about this stuff. Yeah. Anyone can build an app these days. It’s it’s… Every day, we see more and more apps coming coming online. I think there’s a lot of people out there feeling particularly empowered because they’ve lived in an area of of complexity and where there’s big problems to be solved that just haven’t been solved yet. And now AI empowers you to then give it a shot and tackle it on your own without being a coder. I’ll be perfectly honest. That’s how we met Joe. About two and a half years ago when we started down this path, I believe I did a webinar on AI, much less evolved than the one that we’re in today. And Joe happened to be on it. And he said, hey. I’ve been working on some AI solutions to solve problems in restoration. It looks like you guys are well aligned. Maybe we should chat. And it turned out that we were well aligned. We were solving a lot of the same problems, and I said, know, I think, Joe, we would both benefit if you were helping us and not competing with us. And here we are today, and we’ve launched, I don’t know how many products now, five or six, both with a well aligned well aligned objective. But there’s a lot coming online these days, and there’s just some things that you should need to, look into as you’re working with these vendors. You wanna know who built it. Are they… You know, where do they… Is is it somebody who just happens to be technical in nature and is looking for an entry point into a into a into a vertical, or is it somebody who comes with some expertise? Somebody who saw the problem firsthand because they were in the industry and they’ve lived it, and they really know it inside and out from a restore’s perspective, not just the technologists. You wanna know where does the data go? Where are they storing it? Are they cloud based hosted? They probably are, but, you know, where where is that data sitting? You know, can you show me where it’s wrong? This is something that… You know, even though it’s my job to build well tested, products, I’m very aware of where things can go wrong, and we’ve been actually very transparent with our customers in terms of where they need to be particularly careful or they should keep an eye out, when things can go wrong. A lot of times it comes with, you know, the gaps in the in the inputs. We talked earlier about the diligence around the inputs. So we are very aware of where things can go wrong, and we’ll be transparent about them. Anybody who tells you that they’ll… This is a 100% accurate, they’re they’re probably not quite telling the truth. There’s always gonna be places that you need to be aware where where the models can be, developed, especially around the edge cases. Can I try it on my own jobs? Look for companies to do free trials. The… None of this should be, you know, purchased or moved forward with until you’ve tried it against your own data and seen for yourself firsthand the results that you would get from it. So look for vendors who provide you a a free trial, not just a demo. Demos can be… Demos are nice because it gets you to the point quickly, but, you know, they can be put in the best light, so to speak. And then just make sure there’s sustainability there in terms of will they be around? Is this somebody who’s just seeing this as a hobby? You know, but they’re… Maybe they built it for themselves and are checking to see if other people wanna use it. Again, that’s interesting. But if end up building your business around it and they’re not here a year from now, that could cause problems for you. So just just look into the sustainability of that company and and and what their mission is and what they’re looking to do going forward. Yeah. Write your AI usage policy again. I’m not suggesting spend a ton of time on this, but that doesn’t mean it’s not important. You know, write a policy that talks about in your company what AI is okay for, cleaning up summarizing notes. Joe talked about this earlier, translations, drafting scopes, what AI is not okay for, final scope approval, cause of loss determination. These are all human led things, review, and and then, of course, the rules of data. So this could just be a one pager that’s that’s written down. But, you know, with with… Again, as I said, with regulations coming onboard, this is stuff you want under your control. We’re running out of time here, Joe. I’ve been, like, talking way too much. Yeah. It’s all good. We’re we’re in the… We’re in that last section. There’s some final… So let’s talk a little bit about costs. Yeah. It’s… It it can cost too much. That’s for sure in an ever expensive world. The problem is there’s no real free versions of anything out there. It’s gonna cost you something, whether it’s money or time. And when you start adding up the the the actual tooling that you need, the different subscriptions, you know, it’s… It becomes expensive. And, again, we’ve already demonstrated it’s not repeatable, and that’s that’s a that’s a problem. Right? And so we we feel the pain that it costs too much because if you don’t know what you’re doing, it’s going to cost you even more. And the problem is it’s it’s more of a hidden cost. Right? And go to the the next slide. And the hidden cost is this. Here’s a great example. So for the for the 40% of you that are already engaged with a, you know, AI in your organization, let’s just say that you have an OpenAI ChatGPT. Everyone has their own login. Right? And you got three estimators. And you don’t realize it, but you just paid three times for the same question. Because all three estimators had this question in their own workspace. They asked the question and they got this answer. And and here’s even the worst part. They all got different answers. Why? Because they all gave it different inputs. As we saw with the Canadian plum, one word drove a difference. Right? And so that’s a real big hidden cost. You just paid three times the tokens for the same question, and you are less defined at the end of it by all three answers. So, like, you’re you’re not only missing money, you’re missing context, and that’s a bigger hidden problem. And so, you know, when you look at one job and then multiply that times every job you do this year, I mean, you got 2,000 different answers and you paid for the same 4,000 questions three times. Right? Like, it it starts to compound itself and and it’s hidden. You’re not gonna see it. It’s not gonna jump out on your expenses and be like, oh, we got this problem. You know? So there’s a lot of hidden cost to to trying to do these things when they’re not, again, a purpose built system. Purpose built systems eliminate all of that, almost all of those issues, right, from, being where all these hidden costs that are gonna live. Go to the next slide as well. Well, people people probably… Just one thing on that. Just like, there has been some horror stories published recently on cost runaway. Uber confidently told all of their employee base. They gave everybody access to all of the tools. I can’t remember if it was Claude. I think it was Claude. And they had set a budget aside for AI use that the entire company hit in four months. So, basically, they went through the entire company’s AI budget because it was on… You know, what they were doing was unbounded. It was… There was no guardrails. There was no real mission onto what AI was to be used for, and the employees just went nuts. And and they blew their their yearly budget in four months. There was another company undisclosed what their name was, but apparently, this just happened, I think, last week. There was a company identified to spend 500,000,000 in clad tokens in a month. And I think it’s probably good for their investors that it was not made public who this company was. Now I’m not saying that anybody here is gonna do that. It’s just an example of what can go wrong if you’re not if you’re not careful. Yeah. Yeah. And that that that just be… The problem is they’re hidden expenses. They don’t just show up. You know what I mean? Like, they’re they’re trickled throughout everyday constant use. You know? So what good looks like, right, and predictable cost, measurable value. You should be able to ask your vendors this. You should be able to ask them, you know, about flat pricing, per usage pricing, you know, what is an actual job cost, like a job to me, the restorer. Like, what does it cost to run my, you know, my job? You should be able to ask these at any vendor, and you should be able to get answers. And if you don’t, you know, again, that’s that’s a potential red flag for you, you know, of of that you’re gonna spend more than you probably would want to. You know? Another thing is if somebody’s offering you the opportunity or a solution, you know, run a few jobs through it. Right? Like, challenge it. Don’t just, like, look at a demo. Encourage and ask that you can try it. You know? And and something else that, Jeff, you you brought up earlier, and I just wanna bring it back is the, like, where does the tool fail? When you’re when you’re evaluating an AI tool to use in your business, ask them where does the tool fail. Because if they’re not willing and capable and able to simply tell you right then and there, then they themselves actually don’t know. And with the best of intent, they can have the best intent to make the best thing, but if they themselves don’t know where it fails, then that’s a that’s a problem because you’re gonna find out for them. And that’s gonna cost you… Well, who only knows what that’s gonna cost you? And so, you know, AI definitely can be a cost problem for your business and a scary one because it’s not visible. However, a proper built resource tool that has clear terms, it’s an instant way to use all the benefits and efficiencies without the expense and and the overexposure to cost. K. You know? Next slide, please, Kate. Okay. So we’re we’re winding down. I think we have a a half hour or or some period of time. We have about 25 minutes left of of… We’re gonna answer some questions that we’ve been provided previously through the registration process, but then also, hopefully, we can take care of some of the stuff that’s been asked in the q and a. I actually just answered a question somebody asked about the backbone of Encircle. We’re hosted in Microsoft Azure. We use primarily OpenAI models, but I actually think we make use of all of them, don’t we, Joe? Like… Because we mix and match. And this is another thing that when you’re working within an LLM, you’re only using one model. We we process, you know, photos. We we obviously process language. There’s all kinds of different things, and we will make use of the right model for the right job. And Correct. That’s another reason why you wanna work with a vendor is because you’re not locked into one particular model that may have particular biases and particular outputs. So we’ve done cost optimization and and model optimization through… By using all the different models that are available to us, but primarily OpenAI. So I just wanted to… You know, I have this… There there was something else that I had mentioned, I think, when we when we when we did some marketing teasing on this webinar is, I think I certainly saw it this way. I’ve seen another people that I work with see it this way, but they saw they saw AI as an easy button. And we talked a lot today about different things. And if there’s anything that I can distill down to understanding about AI is that I I I I think it’s a mistake to see it as an easy button that provides the same outputs that you’ve always been getting but for less effort. Right? This isn’t… AI is not to be used to… So that you can put your feet up on the table and and just let some computer do the work for you and achieve the same objective that you’ve already been… Or the or the same outcome that you’ve already been getting. I would really encourage people not to see it as an easy button, but see it as an accelerator of your business. By putting in the same amount of effort that you’re putting in today, the same amount of expertise, the same amount of real life experiences, when you augment that with AI, you should achieve, like, massive improvements in productivity, efficiency, and and even profitability. You know, when you get into scoping and estimating, there… There’s areas where you could see line items showing up that are completely justifiable in in their billing, whether it’s furniture blocking or containment. Those are ones that we often see people miss. But the AI model looks at all the data and says, no. No. No. You can bill for these. So whether you’re finding new revenue streams or, you know, you’re you’re you’re you’re able to focus on higher value tasks like like working with your customers directly and making sure that they’re they’re taken care of in their loss and not having to sit there. And what was the example you used, Joe, around, contents? Right? Like, we’re we’re… We have a we have a contents identification application. So instead of spending all of that time describing contents, you could be spending more time making sure that your customers have identified all the contents that have been affected in a flood or a fire loss instead of just sitting there with pen and paper describing them all. You know, we want people to focus on higher value tasks, and I think that’s that’s what we wanted to leave you with today was was don’t don’t treat it as an easy button. Do the same amount of work, but just expect way better results, way higher returns, and and much more profitability for your business. So, yeah, I was gonna say, I think, Kate, can you jump back in here and and Yeah. So, we did provide a link to a worksheet for you all to fill out. If you weren’t following along during, we did kind of reference it a little bit in some of the exercises. But there’s also some templates in there that can help you kind of plan a a structured approach to introducing AI and and even trying it out, seeing if it… If it’s a good fit for your restoration business. So that will be in the, in the chat as well as emailed out to you as well. But I think we should get into some questions here. I’m going to kind of hit you with a curveball because I saw one that came in that I wanted to know. So Michael in the chat, he said, from what he’s hearing, AI really helps you sound smart. But what is it doing to help run a a restoration restoration business from day to day? Like, where where is that application? I don’t know. Joe or or Jeff? Yeah. I mean, I can… The way we’ve thought about it, we… I mean, there’s a number of places. Again, I I had mentioned scoping and estimating is the newest product that we we launched. We’re… It’s got a ways to go. It’s got a long road map ahead of it, but we’re pretty happy with where it’s at now in terms of taking the documentation that you do anyway. You have to you have to take that documentation to prove out the loss, to provide, all of the evidence to why you did what you did and justify the bill that you made. Once you collect all of that data, it should do all kinds of downstream stuff for you, create the scope, create the estimate, create the work orders, create your budget. Right? So so I think that’s an area you should expect AI to help. Other areas that we’ve looked at, again, we’ve we’ve we’ve we’ve tried to make AI pervasive without our app. So providing you that to do lists, note summarization, data… Language translation is another area that can help. And then you’re seeing a lot of contents workflows being supercharged with AI. So massive efficiencies there because I do believe… And, Joe, correct me if I’m wrong, but, you know, that that process of documenting contents and describing them all as you need to do to come up with a schedule of loss is is just an… Almost an insurmountable amount of time, and that can be massively collapsed. Serge, I’ll let you comment on that. Yeah. A massive amount of time, right, that it that it reduces. No question about it. The to add to your point, though, the… There’s… The simple answer, how can you best use AI to to help run your restoration business, is it’s using the AI that’s in the purpose built tools that you’re already subscribing to instead of trying the next new shiny AI solution. You’re gonna find more repeatable benefits to those tools than you will from something that maybe just showed up yesterday or last year. Right? So you… You’re gonna wanna kinda lean into that. But there’s not a single one AI solution that does everything that you need to do for restoration yet today. Right? That’s definitely coming on the horizon, though, as as we continue to build with AI. I just saw a a question in the chat there being involved with TPAs and carrier So this is a well understood… This seems to be a well understood concept in our industry, and I’m seeing a lot of tools, especially in the scoping and estimating place. Take that into consideration. Again, we’re we’re trying to stay away from being specific to Encircle here. We wanted to provide a lot of in… Generic information. But given the question was asked, we are building, our scope tool will have carrier rules built into it for sure. That is absolutely… Probably the number one request when you get into scoping and estimating is making sure that it’s built, in compliance with carrier and TPA rules. So it’s absolutely something you should look for and something you should expect. Yeah. That… That’s something you’ve definitely heard a lot of. Going back to being purpose built. Right? You gotta build a scoping tool that takes the guidelines. Oh, I just saw another… I know I know, Kate, this wasn’t what you wanna do, but I just saw somebody mention good time to mention going from system of record to system of action. Somebody’s been reading our internal product documentation there. That is, so just to explain what that is. So system of record is what a lot of folks would know as Encircle or Dash or or Albi or Xcelerate or PSA, they’re a system of record. They they record what happened on a job, what happens in your business, your financials, and they’re a place to put information. The the the world is changing and evolving with AI to becoming a system of action. And this is where you should expect the data that sits in your system of record should no longer just sit there and wait for you to go in and look at it or wait for you to consume a dashboard or wait for you to go into the inbox and go into a job and then go into the photos to find a very specific photo. A system of action is is kinda what I was describing in that three stages where that last stage was automate. The middle stage was Augment, and the last stage was automate. System of action is really talking about automation. That’s where the data, while… When synthesized by an AI agent can make a decision and, and then move… You know, based on that decision, move a job along. So whether you’re just moving through job statuses or sending, or communicating with the TPA to let them know that a milestone is reached, that’s what a system of action will do. And, and I would say it’s it’s to be looked forward to as sort of the next evolution of of CRM and job management and field documentation. Yeah. And and I just, I just opened up a few questions here that are in the the q and a. I’m gonna I’m gonna go ahead and ask them. I’ll read them out louder than answer them. But, Andrew says, if our staff is IICRC certified and if our staff provides an accurate narration during the walk through video, then what output can the AI provide that would be informative we are not already aware of? Well, I don’t… I’m not sure what you’re already aware of. And if we’re speaking about our products, well, you can develop a instant scope from that and then turn that into an Xactimate estimate within a few clicks. So if your staff is IICRC trained and knowledgeable and can do a great narrative of the loss, then you can produce an estimate within a few minutes. That’s what you can do, Andrew, for sure. Next one is Mark. Been a while. Hey, Mark. What’s up? I I I do know I owe you an an email of some nature. Been busy building. And so you ask… You made a key point on guardrails. And while it’s great to possibly use AI independently, do you believe that owners should find reliable resources to help guide you in the use to help streamline margins? Yeah. So, Mark, I I think what you’re asking is, you know, AI can have a lot of benefits, but should a company seek somebody maybe with more knowledge about AI that they don’t have to help implement us align it? Absolutely. Absolutely. So the message to all the owners watching. Right? Like, if you wanna try this, but again, you don’t you don’t know what you don’t know. And and you’re… And the worst thing you can do is not put a step forward, right, and try. So I would encourage, like, find somebody who is able to guide you down that path and show you the things and have them be your own expert IT guy you don’t have to have that knowledge. Right? That that is definitely a good approach. The next one here is by Mike. It says… Let’s see. That’s one… Yeah. That’s how I I like the… Kate, but I wanna… Actually wanna revisit that because I was just reading that, and it made me think of some things. So, you know, AI tools talked about essentially making fancy professional sounding scope. So here’s here’s the thing. You know, we are generating… Or a lot of folks are generating fancy professional sounding scopes. But the reason we are is because in our customer research, when we started down this path two years ago, we have a lot of customers telling us, hey. Here’s an area that we need help. Generally, we collapse the scoping and estimating process into one thing, and we we really wanna get to the estimate so quickly that we think in line items. We go through a house, and we’re like, okay. We’re tearing out two foot flood cut. So you’re you’re you’re basically compiling ex… Xactimate line items in your head as you’re doing an initial inspection. And what that essentially does is it skips the scoping process. So at some point, when you get into a discussion with an adjuster who’s asking… Who’s not arguing about the price of something. He’s not arguing that that’s what drywall costs or that’s what flooring costs. They’re asking, why did you do that task? Well, if you haven’t built a scope or if you don’t have a scope to fall back on and your memory is a little bit fuzzy like mine would be, you don’t have a good answer. So while scopes are certainly fancy and fancy sounding, they actually do provide this artifact, this this this asset that was traditionally skipped in in in in the in the the pursuit of getting to an estimate as fast as possible that then did leave a little bit of gap in your process and your procedures when you were being questioned as to why you took an action that you did. So I I I I get the point that there’s a lot of people producing fancy scopes, but the problem that it’s solving is that scopes weren’t really being built at all. A lot of people, based on the information that… The people that we were talking to, we’re saying that’s where we need help. We can get… You know, we can have people… We can think in terms of line items, but we really need help using the field documentation that we collect, turn that into a scope that we can depend upon, rely upon, and becomes sort of the ground truth to, why we’re we’re… We’ve done the job that we did. Right. Joe, anything you wanna add on to that? No. No, Jeff. Good. Yeah. Alright. So I do have kind of two versions of the same question. So Jeff and Mike, they submitted these through the, registration form. So we’ve got techs who don’t even fill up notes properly. How does AI help if I can’t get them to document? And then, how do I get my crew to take better photos and notes without making it about AI? They do not care. So Yeah. I mean, those are really… Basically, the the the same question. Right? How do I get my guys to, you know, to deal with this and adopt AI and do better documentation? Right? And the reality is is AI, first and foremost, is not gonna help you do that any way, shape, or form. And like Jeff pointed out earlier. Right? And so if we look at a standard water loss and the technicians go in and people go in and they inspect the loss, they capture the information, ultimately, something gets missed every job, on every day, all the time. I challenge you to show me a job that you didn’t miss a photo of something that you ended up needing later. Right? And so the goal is to close that window. And the reality is is… And again, Jeff pointed this out is is you may not collect it today, but you are gonna be forced to collect it on day three or day four or on day seven after you’ve already left. Because the estimator needs to know, hey, man. Was there was there this? Was there that? They’re gonna be asking these questions, and and sometimes it’s just a question. Sometimes they need something. And so the the… Instead of approaching it, like, you know, AI can help you do this or we need this because of AI, Let’s establish good processes that just ensure that we’re collecting the information minute one upfront. Right? And and here’s the deal. That’s not us as Encircle telling you as an… As a restorer, go do a better job. Right? That is all of us accumulatively as an industry saying, we need a better solution to collecting information easier, faster, and simpler to… So that we can document and become the ground truth. Right? And and and Encircle is clearly working towards that exact solution. But… Yeah. I I do think there’s an imperative that all of these tools need to think about that data collection process and make it easier to follow whether it’s some sort of, you know, a scope sheet, you know, sort of a more traditional way of doing it. Obviously, that was more of an analog. There… There’s certainly much more can be done in a in a digital sense that, you know, checklist can be much more… Can can be smarter depending on the context of the loss. The technicians can be asked specifically what they should be capturing, how to capture it. So I I do think that is a, as a lot of these vendors have realized that, you know, the success of their product depends on good inputs and good documentation, it does fall on, software companies to think hard about how do they help their customers put the best processes in place to collect all the data upfront and don’t, normalize inefficiency of that back and forth over two or three days where the estimator’s calling the technician or the PM. Where the hell is that photo? Where is this reading? You know, we do as an industry need to figure out how how do we get the technicians. I mean, I think the the the way that problem or that question was phrased was I can’t get my techs to do it. That’s a real problem. And if we don’t solve that problem, then we’re not really helping. So, you know, that… That’s definitely a problem to be solved, making it efficient to collect that data. Yep. So we… We’re running up against the clock, but there are two more questions that I would like to ask. First, Joe, you already kind of touched on this a little bit. But, Jeff, I think in a in a previous conversation we had, I think you have a a really nice perspective to share. So we don’t have an IT person who should be responsible for this. I’m the owner. Should I be the one testing everything? Yeah. Yeah. So that’s a that’s a great question. So I would say, while I was heads down and early on thinking about ways to apply I… AI to make our product better and more efficient and all of those things and provide our customers more value, our CEO, Paul, was really the the one that established AI as, a a tool on everybody’s tool belt in the entire company. He realized early that if if we wanted the entire company and and and, you know, he he would be at the very top of the authority chain of who wants the best for the entire company. So he did take it upon himself to be the the evangelist, the cheerleader, the… Ensuring that all of the, employees were using it in their job efficiently in a non wasteful manner… In a non lazy manner, that it wasn’t an easy button for everybody that we were using it properly. So so given that I would say the CEO has the most to gain from a a ubiquitous use of AI in your organization, it probably does… I know you’re busy, but it probably does behoove you to to be that cheerleader and to ensure people are using it the right way to make sure that the policies are in place. And, you know, the… You know, I’ve said a number of times that the objective of using AI is to make you more profitable, is to help you grow in a in a scalable manner. And, you know, hopefully, when that happens, you can hire an IT manager and make all of the policy stuff their problem. But I I do believe that that the CEO does have a role in in using, establishing a ubiquitous tool that that that helps horizontally across the entire company. So this is a question that’s come up a a few times in the registration form and in conversations we’ve had with clients. Is AI actually ready for restoration? Like, AI… Or sorry. Restoration is is such a variable industry. Like, it… So many factors go into it. And I know, Joe, you probably have a a really, really good perspective on this. I’m sorry. I didn’t hear that last part. Is AI ready for restoration since it is such a variable industry? I mean, yeah, it it needs to be. And I’ll tell you why. Just this month alone, AI labs have shattered records across intellectual domains in most major categories, specifically math. And so we’re seeing the adoption and acceleration of AI in every single industry. So restoration better be ready for AI. The question becomes is is the industry building enough tools or the right tools with integrity? And I think that’s where the users have to be selective. You know? And the the AI can handle the variability of restoration jobs. I know you mentioned, like, the structure No question. Typically No question about it. Is the same. Yeah. Okay. It it it it can handle that understanding of the dynamics and the nuances. But, again, I say that confidently based on a purpose built AI, something that we do at Encircle. So I’m not saying that for every AI tool in the space, I I can’t speak to anybody else’s tooling or or solutions. But purpose built AI can, should, and and in our case, can, handle the variances. Again, thank you guys for joining me. Thank you, Jeff and Joe. You’ve been fantastic, and we’ll see everybody next time. Bye.
Speakers
Jeff McDowell
Chief Product Officer
Encircle
Joe Ledbetter
Principal AI Innovator
Encircle
Learning outcomes
Attendees will learn:
You’ll leave with the first steps to try AI safely, starting with the job documentation your team already creates.
Barriers and costs
- The most common barriers to AI in restoration and how to work through each one
- How to evaluate the true costs of AI tools
Trust and accuracy
- Where AI performs well, where it needs human review, and why expertise matters more with AI, not less
- How to recognize bad AI output and handle errors with a consistent review and escalation process
Getting started safely
- Low-risk ways to start using AI in field documentation, and how better documentation leads to better downstream outcomes
- How to set internal usage policies, protect customer information, and ask AI vendors the right questions