Transcript for "The Architect's Edge: Smarter Specification Workflows With AI": we don't have as many specialists any longer. The specialist of a specifier, it's kind of a it's kind of a bonus if you have it. It's it's a incredible tool, but most of what we're seeing within the industry is you have a lot of what I call hybrids. You know, me being one of them in the past is you have hybrids in the industry responsible for the drawings or the models, and they're also responsible for the specifications. So how do we bring along those generations, or how do we bring along those folks that were more, focused on the graphics? How do we bring them along in on specifications? This is, after all, a more visually driven industry, and that's how Ask Della started. You know, we wanted to make sure that we had some of that data that we have within the master spec library. We wanted to make it more consumable, easy to capture, easy to find, easy to compare things. At the end of the day, we don't want specs to get commoditized. You know, that's that's the the opposite of what we wanna do, but we wanted to leverage AI to to make the process more efficient. So if we look at ask Della, so this is kind of a a brief, a little bit of a brief prompt that we have here. So Della is embedded directly in the specification workflow. So you'll find her in the workspace immediately on the right side of the panel, and it's built you know, she's built on master spec content. So before I show you what she does, I want a core principle to land because it governs everything that Della is assistive. It's not autonomous. There's definitely things that, Della can automate. It is proposing, but mainly it's the Della proposes, the professional decides is kinda how we frame it. So we're not trying to take away your stamp. We're trying to take away your busy work at the end of the day. Now if you look at the conversation over on the right hand side, this is just a really brief I wouldn't recommend this kind of prompt. We'll get into some of those prompts in a minute here. But you can see here that the basic ask is to compare acoustical ceiling panels for a specific building type, health care corridor in this case. We have a couple of requirements. We want the NRC, at or above point nine. We want the class a flame spread, and we want a current EPD as well. So those are some of the the criteria that enhance what our output is gonna look like. And notice how this is exactly what a lot of us are thinking when it comes to different design, options or generating options for products. Generally, you're doing a comparison matrix or some type of comparison to figure out, you know, based on my criteria, what are my best options? And you can see what comes back here. It's not a huge paragraph of fluff. I asked it for a specific output. I gave it my criteria, and that's kinda what it's delivering for me at a at a surface level. So it's giving me the product side by side, the criteria I asked for. It's all in plain language. You're also getting the citations as well at the end of it too. It's not just about, you know, Dela guessing and hallucinating and looking to come up with the right answer. These are all grounded in data. So it's pulling it from a specific knowledge source. It's putting together the resources. So, you know, if if if Della has shown you her work, you can actually go through and verify it. And that's the difference between, you know, a specific digital assistant that you can trust and then one that, you know, one that may not do the trick, one that's more commercialized and may not look at the same type of data. So if we think about the prompting, you know, talk about prompting quite a bit, and I know it's a little bit difficult. We're all starting to get into kind of this prompt engineering type of thing with all these different, you know, with Cloud, with ChatGPT, with Copilot, all those types of tools. The way that you can get the best out of Della is really the way that you can get the best out of any of any of those tools. But, specifically, when it applies to specifications, when you think about prompting, the the strongest step, the if you were to do any step, it would be step one, which is generating the task. So lead with the verb, something, generate, compare, analyze, summarize, any of those types of words. Starting your prompt point blank for what you're trying to get is gonna be the most effective way for the AI, for Della specifically, Ask Della, to give you the best kind of output. In terms of the context, I think of those as the sweeteners. Those just make your outputs stronger at the end of the day. So you have a couple of things as I mentioned earlier. I wanted a specific criteria. So anything like performance requirements, characteristics. You might also wanna get into sustainability. Anything that you're looking for in particular within that product, you know, it could even just be like a color, a thickness, anything like that. The context will help you, help provide a better comparison. Third of all, you have your examples and the examples referenced in your past work. So examples might be something that, you know, a specific style that you wanna see this in. You know, if you have a specific, any specific context that you wanna apply to Ask Della or any other tool, you're just giving it the how you like to see it type of, sweetener. And then the fourth is the chain of thought prompting. And this is mostly the idea of, you know, you you think of, you think of the AI and it's almost like, initially, we didn't wanna ask too much of it. You know, we we did these prompts individually, and we didn't want to bloat it because the AI generally maybe was going off in a tangent or a different direction. And that's actually typically not the case with Ask Della and many of these more, any of these smarter models. Actually, the more context that you give it, the more examples or the more more, what I call, chain of thought prompts come to come to life. The more that you push through it, the better it's gonna do. So if you're asking it to create a comparison table, if you're asking it to analyze a specific part of a section, or you're doing something else related, you don't have to necessarily prompt it three separate times. You can kinda give it You can give that prompt, all at once. So first, second, and third type of priority within that same context, and the AI will know what to do. It'll do a better job with that context as well. But the most important thing at the end of the day is the task. As I said, if you're gonna do any one of those, stick with the task, and you're already gonna be far ahead of where you need to be. Now in terms of the data sources, we get asked this quite a bit in terms of ask Della, what does it use, what doesn't it use? So what Della uses currently, it uses the AI master spec library. So all of our content library, those are, you know, essentially the sections, if you think about it that way. It also does use the supporting documents. So all the evaluation criteria, the, you know, the actual, coordination checklists for the drawings and for the specs, all that is built into it. It also does have the spec point product library. So the spec point product library is built up of you know, you have all the manufacturer lists within master spec historically. We also have products and materials for those specific lists, and we've also built the criteria or the, properties that are built within those products. All that is is nestled into a library that's updated in real time, and Della feeds off of that to do things like the product comparison that I talked about earlier. And then you have the help documentation, which is great from, you know, from the perspective of taking on a new tool. You're trying to figure out a new tool, what it does, best approaches, you know, to use certain features. If you're trying to find certain features, those kind of things, Dell is great great at. Now what Dell doesn't use just has to be said. These are the basics, but it's not, you know, using it's not using, web data, so it's not ungrounding itself in guesswork. So it does not use web data. It's isolated to a specific knowledge source that's based on Deltek and AI resources, if you wanna think of it at the end of the day. It does not use other firms' data, does not use open web scraping, and it does not use your prompts for training. So all the kind of basics that you might be thinking about, you might have already entered into the q and a in terms of security and legalities. We're making sure that we're keeping Dela clean. Currently, it's using our intellectual property and the AI's intellectual property, and it's also not using web sources to contaminate those sources. So that's something also to consider. Even if your firm is building your own AI or if you built your own AI, you have to consider things like, you know, is it connected to live data, or is that data based on projects that maybe have had a lot of RFIs against them or have a lot of inconsistencies against them or old standards. Those are the kind of things that you have to think about because that's gonna contaminate the types of answers your end users are gonna get. Now in terms of AskDella in action, a couple of main use cases, we've already gone through these in some way, shape, or form. But what the industry has been seeing or what we've been seeing from the industry in everyday use, the first and foremost has been the compare products, use case. So primarily, you're you're having, you know, you're providing your requirements, which become your constraints, and you're asking for some type of comparison matrix or table to compare the products, either in a category that you don't have much background in or in a category where you do have a standard or you have an owner standard in mind, and you're trying to compare that against alternatives within the industry. And then also, you might have specific, you know, for example, comparing product types. That kinda gets into the developing approach for number two. But if you're wanting to compare, unique identifiers even for similar products, so say, for example, you can create a comparison table and product a and product b match up with one another, what are the key differences? Ask Della can take you that layer deeper and identify, okay, what makes this one unique versus the other? Now developing an approach is pretty common as well, so that it's especially common for those that are getting into, pretty tricky sections. So I think of, off the top of my head, probably like glazing, anything insulation, anything sealants. I'd have a lot of questions for, you know, for example, like, if you're specifying one product type or another product type, what which one should you consider and why? And a lot of that information is already embedded in the master spec evaluations. You're just getting it out quickly and efficiently using Ask Della. Because each section think about it. Across the master spec library, you have 900 plus sections. You have supporting documents for virtually each, and then you have those supporting documents that are an average 25 pages a pop. So that gets to be quite a bit, a lot of manual review, and that's what Astellas streamlines. But then you might be you know, maybe you're a reviewer in Specpoint and you're getting into the specifications world and you're trying to understand what you're seeing. You know, how do you get started with MasterSpec? What does this specific product type mean? Or what does the specific, section intent? Or why do we use this particular language? There's a lot of that getting started with MasterSpec type of, inquiry that we also have in the evaluations that is critical when you're you're getting into specifications. And then finally, the learn the platform, there are certain things unique to spec point that gets into, like, automation, for example, best practices implementation. You think about bringing in your custom content, importing it, or even cloning master spec. You know, what are best approaches to implement the automation? What is you know, when should I consider import versus cloning master spec? Or maybe even just basic getting started with projects. Ask Della is very powerful for those types of u workflows as well. And if we get into the use cases, I won't go through each one of these, but you will get the you will get the slides after today's presentation. These are very in-depth, very much check all the boxes of what kind of prompts are successful prompts with AskDella and other tools, I definitely will encourage you to, just copy paste these. Literally try these this week. Take these prompts. Compare them with, AskDella's outputs with some of your own AIs that you like to use and see for yourself, you know, some of the some of the outputs and some of the answers that you get and, some of the, some of the resources that you get also, some of the links, the references that come from those. It it'll be pretty interesting doing that comparison exercise if you're comparing, multiple tools or even just getting into AskDella and trying to find out what, what what she does well with. These are these are great ways to get started. And I'm sure you can think of your own as you go along that, kinda go along these lines. So with that, I'll transition it back over to Jeff, and he's gonna take us through spec intelligence. Yeah. Thank you, Chris. So spec intelligence is something that I'm, actually, we are very excited about, something that we are looking to hand out to our our customers very, very soon. So let me just go in and get started in it. So what is specs intelligence? So, just a kind of a quick note, in this presentation we call it spec intelligence when we release the feature that will be called insights. So what I'm talking about today is going to be technically called insights. So what is it? It is a AI powered QAQC analysis tool directly embedded within the spec point application. So it offers many things when reviewing specifications. It provides clash detection. So one of the the great things about this is like missing requirements and complete articles and gaps in execution. So if you have a product type or product specified in part two but you do not have the installation requirements in part three it will flag that which is really handy. In terms of like multiple section analysis, let's say you're looking at the cast in place concrete section and for whatever reason you put the curing time or curing yeah, curing time at like twenty eight days and you go to the polished concrete section and it's at thirty one days, it can flag that for you as well. That would be the contradiction. It is very impressive with with what it catches. It will even eventually provide AI recommendations. So it won't necessarily tell you, like, explicitly what to do, but suggestions on how to potentially resolve the finding that the AI found. And then obviously, the easy one is cross section coordination. So if you're referencing a section down in part three that says, hey. Just make sure the contractor coordinates or install this piece of the of the assembly per section whatever it is and you don't have that section added to your project it will also flag that. So it is it is very very impressive. I was actually taking a look at it earlier today. One of the the great things about this is that how many hours are we spending doing QA QC analysis of specifications. Right? We spend many many hours trying to do this and nobody's perfect. We always we always miss things. I miss things, easy things. And this tool, again, it's a tool not meant to really replace it but help us improve that QA QC process and limit the amount of time that we actually spend on it so we can go spend time somewhere else on the project. And again here's just some real findings on on what it catches. So if you looked at the first one here cross section coordination, right, this is just telling you, hey. The thermal insulation section is not added to the project. You can either add it or just remove that that paragraph or sentence that's in that section to resolve it. It does find Arizona emissions here. So, again, we're defining Plazadeck products, but the part three doesn't have any installation requirements for the a the AI recognition. Easy. Add it to part three or just remove it. And then the contradictory language, you're looking at hot fluid applied rubberized asphalt. It's a long section name. Thank you for that. 071354. Right, we we're specifying allows installation above 40 degrees Fahrenheit, but then when we actually look at it, it actually is saying something a little bit different so we can actually identify those within the section themselves or across multiple sections within the entire project which is extremely extremely helpful as a specifier architect or or engineer. We can even provide a little bit more. Here's some more examples here of what it provides. I won't go through every single single one of these And again just the last slide here of examples of what we can find. Again it is very impressive what it can do and it's all because we can take that structured data of mass or spec put it into a machine readable structured format and produce these types of results for end users in spec point. And if you look here, right, this is just kind of a simple workflow of I want to analyze these two sections o seven thirteen fifty four and 071413, and we're gonna check for for everything. It's gonna analyze those sections and it's gonna provide you feedback. So it found 30 issues, 29 of them are are AI that we found. Two of them are potentially high, and then you haven't resolved any of them yet. So we'll let you, accept or dismiss the finding. We do save that and report for you, so nothing is ever lost. There's a history of it as well. Again, it's extremely efficient, at what it does and and hopefully will provide a lot of lot of valuable time back to our end users with it. And so where where is this headed? Right? Where is, insights headed? So, again, we can take this in a lot of different directions but we have a really firm direction internally of where we think we can help our end users based off of the feedback that they provided us. The real world example potentially is something like this in 2027 where partner changes an exterior wall assembly and Revit. Revit automatically tells Specpoint to update the o seven forty two thirteen secondtion. The submittal law report then flags three affected products, spec intelligence catches a new clash, and then Della can draft the addendum narrative or whatever it needs to be, change order or whatever it ends up being. So we have this entire workflow built out all through the connection with with Revit and Specpoint and our internal tools with Della and and Insights and a few other features as as well. One of the cool things about this again, what makes this possible? Structured data. But one of the cool things is the bidirectional BIM. We can we can have data exchange to Revit and we can have data exchange from Revit. So historically every application out there on the market is just one directional. Now we can change that up with spec point. So if you change a thickness of a wall, in or I should say you don't if you change the thickness of a wall in the Revit model that changes, you know, say you say you change that half inch chipboard to five eighths chipboard. Right? Or maybe you added another layer. We can send that information back to spec point and inform the end user of this change. In some cases, we can probably actually make the change, not necessarily in all cases, though. But, again, we can move that data back and forth. And by doing this, we can build up what's called the project knowledge base, and I'll get into that a little bit here. But if you think about the project knowledge base of just the brain of your project, so it includes everything, design, specs, submittals, RFIs, change orders, addendum. Right? How many times at least well, let me back up a little bit. When I started in the industry, ten years ago, my firm had an off-site locker, and that's where every archive project went. Paper copies of the drawings, paper copy of the specs, and it went into the into the locker, into the storage unit. When I got there, we were still doing that, but we had a guy who was basically scanning everything and and digitally archiving the entire project onto, under the computer in the cloud or whatever. When you're looking at project history, if you're doing a lot of the same repeat projects and you wanna look at something, how are you finding that information? Right? How easily accessible is that information? It's pretty hard. Right? I remember I had to pull out a CD, find the project number, go to that CD, put it in my computer, and basically read through read through the the specs or the drawings. It was very, very tedious to do that even when everything was manually put over. Alright, the Windows search explorer not really great very difficult to find information on a project depending on I guess I should say no matter what you're looking for or especially if you didn't work on that project this changes everything this project knowledge base really changes that entire workflow so you can literally prompt Della and say hey Della I worked on a gym project five years ago What was the gym floor that we used? What were the details that were used? And did it have any RFIs or submittals? And within the reasonable response time of of AI, you could have all of the information right at hand to then make the decision that you need for your current project. That's how powerful this this project knowledge base really is. It can literally centralize your firm's entire knowledge about all of your decisions in the project, everything that that's happened, all the design decisions, all the product decisions, everything that happened with the change orders, the RFIs. They can help you automate decision support so we can eventually build agents that really help automate a lot of this workflow. You can learn from past projects and standardize and compliance. About the time I left, we were trying to really capture knowledge within the firm. So we had several high profile architects leave, and it was like, oh my god. So and so is leaving. That's thirty years worth of knowledge right out the door. How can we capture that within a month? And so we developed this this platform to be able to put up project history and be able to standardize everything. And it was a long term project. This was without AI. Now, Dela can do that. Insight can do that, and this project knowledge base is basically the core of all of that within the spec point application. And so in practice, right, taking all of this, we can, like the previous example, design conflict detection, sustainability assessments, that can all happen within spec point and all because AI makes it possible to be able to to do this, which is from my perspective, if I was still a specifier, would be amazing. I would want this as as soon as soon as possible, basically. So, look for this later on here potentially towards the end of quarter. And, again, it's gonna be called insights. Everybody will have have access to it and be able to run all your projects, current, new, past projects that are in spec point. That would be the key. They gotta be in spec point. But you can certainly import your word docs and and run it as well and get those those findings. It will work on on all sections that you have in the spec. So with that, I will pass it over to Chris for closing remarks. Yeah, Jeff. Speck Intelligence, the insights as we're as we're transitioning into calling them, very, very exciting. And, you know, I I think when you think about intuitiveness, ask Della is wonderful if you know exactly what you need to know, like, exactly how to prompt it. The nice thing about, you know, this the insights feature is that it takes a lot of the guesswork out of it. You it really the prompts are are are working. They're underneath the surface. Really, on your end, you're just kinda hitting the analyze button and reanalyze button. And like Jeff said, you know, this is a wonderful tool in surfacing a lot of those inconsistencies. But at the end of the day, like, if you're thinking about model coordination, for example, you know, this these types of tools can give you the delta. It'll tell you, well, this is what the spec says. This is what the model has. This is what the last time this was updated and who did it. But you ultimately have to decide what is the what is the source of truth. Are the specs the correct version? And then you have to go figure out, you know, how to update the model, or is it the other way around? So AI can get to the point where it can probably reason for us and and probably take a stab at automation. But, ultimately, at the end of the day, you know, human in the loop is critical. It's it's slashing our time, making it kinda like a fraction of the time to get some of this information. Like, I remember some of the pain with, with coordination or the pain in, you know, trying to even figure out, inconsistencies with reference standards or inconsistencies across the manual. You know, even the best of us will miss things here and there. And over time, you start to figure out, okay. Well, I know I know how this consultant works, you know, more or less, I know exactly where I might have to, give a second look. But that's all based on experience. And this is what I love about the spec intelligence side. The insights is that it's doing a lot of that for you. So it's a great foundation to build off of. And as Jeff said, I think the historical precedence, like, there's many a times where even I pulled details or families on the Revit side that were used on previous projects but were, were not the greatest references, but I didn't know any better at that time, until somebody had told me way down downstream that it was too late. So having the ability to to have that context on why something was used, how it did, you know, was it successful. Some of those prior learnings to to do better on future projects is gonna be critical, but it all comes you know, having the data in one place is gonna be critical. So adopting the tool, making sure that you have this common data environment, that's how this all works. Now in terms of AI too, before we land the plane, just a quick word, you know, how to be intentional about AI. It's it's a lot of the basics as with any new tool or new technology. It's putting together the process. So making sure that you pick the tool, that, you know, that basically matters far less whether your actual workflow changes around it. It's more about what the tool can do for you at the end of the day. So build a process that rewards the new behavior. Don't try to force your existing process on a new tool because most likely, that's not really gonna work. You don't wanna reinvent the wheel, of course, but there you know, with any sort of new adoption of a tool, a feature, and enhancement, there has to be some sort of workflow adjustment so you can get the most out of it. So just kind of being honest about that is number one. Number two is making sure that you have a pilot, running pilots on real projects, not just this kind of, like, endless demo environment. You know, some of us can get into these test projects and, you know, we drop the test projects. But more than likely, you know, more often than not, what we've seen on our end is that the more successful teams that implement new features like an AskDella, like the insights tool that Jeff was talking about is actually putting it through its paces on a project where it makes sense. So, measure, you know, expand or measure, learn from that, expand it, and then repeat on future projects. And then finally, the trusted environment. You know, when you're thinking about these tools, as we've talked about previously, using tools that ground themselves in connected data always provides better results. That's that's gonna be critical. Like, for example, in SpecPoint, when, you know, ask Della, It looks at master spec sections, correct, as I mentioned earlier. When anything changes or, you know, anything changes with a section is being updated, it's updated in real time. So once we've updated and pushed all those updates in spec point, Della is also feeding off of that as well. So these are live sources, you know, live releases that we're pushing out, and it's actually it's you know, Della is not using a previous version or there's no lag in between when Della you know, what data Della is seeing versus what's in Specpoint. It's all live as opposed to kind of this document centric approach. So, the environment is gonna be critical, not only getting the live sources, but as I talked about, getting a common environment where everybody is is using one tool as much as you can or centralized tools, that talk to one another. And as we said, AI is not gonna replace the judgment, but, of course, it's gonna accelerate the work that leads up to it. And that's kinda what we have in our key takeaways here. So key takeaways, the copilots will accelerate what you do. As we talked about, spec intelligence or insights is gonna catch errors, contradictions, omissions. It's gonna accelerate our work to get to the decision point, but the decision point ultimately sits with the professional. The structured specification data is a prerequisite. So whether you're using master spec, you know, if you are using master spec, I should say, you're gonna be in better shape because we're doing that day in and day out. Every update that goes out not only contains more updated and current information, but it's also, some of that information is normalized or structured in a way that, not only AI can take advantage, but all these other automation functionalities, you know, all these reports and things like that, you want more accurate reports and accurate outputs from AI. The best way to do that is to structure your specs in that manner. So if you're if you have your own custom specs, of course, we do have also resources where we kinda share how our team goes about our normalization process, and you can apply those same principles over time to your custom content as well. And then Grounded AI wins. Open web AI loses. So, again, with the web, you're gonna get what you're gonna get. Sometimes it'll pull great resources. Sometimes it won't. You know, as you might have found out already, a lot of our AI tools, these commercial AI tools, they they're not, exactly accurate all the time. They even if you feed it the exact same prompt sequentially, it's gonna you know, it's potentially gonna pull a different source, especially if it's looking at the web. It's just gonna find whatever it finds first, and its answer might change. Now if you do have what we call grounded AI, you know, we've talked about this AI where it's the live data. You're working off of knowledge source as Jeff talked about, the knowledge base. That's gonna be critical. That's what you wanna look for in a tool, and that's what's gonna get you more consistent outputs and make sure that what you're getting out is actually, you know, it's actually trustworthy. At the end of the day, pilot unreal projects. Again, can't say this enough, but the firms that start now, unreal work and quietly built that advantage, over time, the rest will be catching up for years. You know, a lot of we hear it very often. A lot of customers will will sort of wait until a feature is fully complete, and then start the work. And I think we all know that kinda making small progress, even getting in early on a specific feature, even if it's not in its full evolved form, getting in early, at least learning about it, if not applying it to some extent, will help you in the long run. And you don't have this big pile of work waiting for you if you're kinda waiting for it to be feature complete, which, you know, that that finish line may always change. So those are some key some key things to take away. This is our contact information. Definitely encourage every one of you. You know, Jeff and I always like to say we have an open door policy. We love having conversations with customers, training sessions, strategy sessions, even just talking about our backgrounds. So please feel free to reach out, email, phone, LinkedIn, whatever works for you. And let's jump into some of the q and a. Jeff, I'll ask you anything that you wanna uncover that you might have seen from the q and a that is worth bringing up to the masses. Well, we just got a question. here, that I think. I can answer. So, I how will some q and a, Adela, insights, just first off, integration, just, and dynamic know, apologies for the late start for some of you. We had a a couple sharing, technical difficulties that we, the model, or not yet using to solve. here. So that that's, interesting. So know, it is about seventeen minutes we're building, is is meant to be where all with the webinar, are in everybody. will get the the recording after this. the you'll be able to catch those first, is we have this capability to seventeen guest, users. So as long presentation. as if you're if you're an architect But just wanna go through Specpoint, some questions here that we had during the the presentation. have Specpoint a good well is will will Della, emails are in the system, or SpecPoint can go ahead and add them to your account be integrated with? up a seat is something that we are talking as a guest user. in the future. They don't see all of your projects. have anything depict on add them to the. specific project, We we're they can all work couple different exact same of if m if an MCP is the right approach. So you don't there's to worry about headers, else. You know, that that coordination be, is simplified a little bit more value for Della. our our customers. And then with our Revit capabilities, And then another question really provides well does does nice coordination for. for accuracy? can also do this with the BPM is actually well. really good. A lot of our, BPMs, it is it is AI, point accounts it does depend on, on the can. add them as guest users if you need help, on a spec it is. really good. Dela that's not hallucinate, how I would approach this is, that that is, part sure is really good. have we do have some point and then, quality built in checks as a a guest an AI prompt. about that. When. you ask questions also say contributors library, too. That's another popular product data, comes from the manufacturers. think of it as long as the product reviewer. data is know, we have a lot of, from know, the lot, of specifiers that we talk to, Della know, will be accurate. expressed that, you know, I don't wanna We do don't another to give full access or too much access, to, you know, the person these specs Revit. do these AI, even though they they have to interact with be used in Specpoint? with the specs, we don't want them to be used in Specpoint. up. Yes. Same thing with young them outside architects it. or young engineers. However, if you have, them, to learn about specs, Word docs necessarily that things up specs with that are outside of Specpoint, along those can import a lot of these different, Specpoint, and the AI tools. will work with imports is a great way to. get everybody can fully common environment spec permissions and a little bit more control. with, imported word docs. do consider that. And then the last question I have here. is, But for anybody that's more serious getting into own key. import, So, setting, up automation, can you know, administering projects, that type of thing, the the front end of Della wanna consider elevating of Della to your own author or? administrator, no, type level. you cannot. So it is, solely provided by Deltek, and you use everything through see. here. Another question came in. Does Della work for all disciplines, GPT or is it mini specific? model. But next week, So it's gonna be upgrading it to the, specific. GPT Maybe some of the examples that we went. over the responses a little bit more architectural based, but MEP, better. The, performance is going to be better, table when it comes to. what AskDella that capable will. be a nice welcome We do also have additional learning resources, Ask Della. videos, and also some, And we do documentation do gets into, look at performance and practices and quality prompts or that sort of thing. for some of the prompts. We did, if you're really deep dive something, evaluation you're looking at a section that's mechanical or electrical, available. I shouldn't say all models, getting into lighting, some of, the the popular ones. So we did do of the prompt, the the actual, GPT, the structure, of the prompt is, a lot of the a lot the same, but you'll get, several know, Claude, and Gemini something that relates well. to your discipline will off of land a bit better. quality of responses, have a a couple of, examples around that, too. you know, cost as well. We identified 5.1 was the best model for our users. Anything else that you found on your, end, I? think that looks like the only what? questions last question have. that that came in we'll we'll end good one. a couple minutes early. Is AI available or all should say, thank you again for attending today. or is this an element apologize? for the late, start, but you will be getting it is available recording everybody catch up. Specpoint. And then like you have, access to you have any questions, Della today. know, please insights reach out. comes out, We love have access to and. and talk and shop do you. have access to best very much. Have a, great rest of, your great. rest of your day, everybody. So, Bye. yeah, it's it's all there for you to to use. And then Della does only