Subscribe
Share
Share
Embed
Summary
Explore how AI innovations are transforming energy management by 2026. Host Marcus Hazelwood discusses with Mohamed El-Sayed the revolutionary impact of AI on building performance and sustainability.
Key Takeaways
Resources
About the Guest
Mohamed El-Sayed — CEO, EA Energy Solutions
Mohamed El-Sayed, licensed Professional Engineer & Principal at EA Energy Solutions and has more than 20 years of experience in energy management, building performance, commissioning, energy modeling, and sustainable design.
Connect with Mohamed
🎧 Listen On: Spotify | Apple Podcasts
📲 Connect: X | LinkedIn | YouTube
Chapters:
Clear Haze Exchange is an interview-led podcast focused on simplifying sustainability while uncovering the hidden systems, decisions, and long-term impacts shaping our built environment and energy future.
Hosted by Marcus Hazelwood, an energy engineer with 15+ years of experience in sustainability and energy efficiency consulting, the podcast has evolved across two seasons:
Season 1 focused on breaking down the fundamentals—making energy efficiency, building systems, and sustainability concepts more accessible and practical for a wide range of listeners.
Season 2 builds on that foundation, diving deeper into the systems-level thinking behind sustainability—exploring how decisions in design, operations, policy, and business strategy create long-term outcomes.
Each episode features thoughtful conversations with industry leaders, engineers, architects, researchers, and decision-makers who are actively shaping the future of sustainability. From building performance and infrastructure to sustainable chemistry and organizational mindset, Clear Haze Exchange connects ideas across disciplines to provide real-world insight.
This podcast is for professionals and curious minds alike, those who want to better understand not just what sustainability is, but how it actually works in practice.
If you're interested in smarter investments, better systems, and meaningful progress, you're in the right place.
Subscribe to Clear Haze Exchange on your favorite platform and join the conversation.
For inquiries or guest opportunities: hello@clear-haze.exchange
Hey everybody, I'm Marcus Hazelwood and welcome to Clear Haze Exchange. I'm an energy engineer with over 15 years of experience in sustainability and energy efficiency consulting. Many find these topics a little complex, so I created Clear Haze Exchange to help provide clear, actionable insight, clearing the haze, so to speak. We'll cover plenty of topics under the umbrella of energy efficiency and sustainability. Let's get into today's conversation. Welcome to season 2 of the Clear Haze Exchange local series. This season is focused on uncovering the hidden systems and decisions shaping sustainable investments and the built environment. Uh, today's guest, we're going full circle, is Mohamed El-Sayed. He's a licensed professional engineer and principal at EA Energy Solutions and has more than 20 years of experience in energy management, building performance, commissioning, energy modeling, and sustainable design. Mohamed also has a special connection to the Clear Haze Exchange. He was the very first guest, but Mohamed El-Sayed. I'm coming season 2, you know. So how was it? How's it been? Like, so I'm glad to be here at the end of season 2, but you're the one who been in every episode. So how was the journey so far? It has been good. It has been a great journey. Well, like we were saying earlier, you brought your phone the first time and I recorded hard data onto that phone and had to transfer that file. So along with our topic of today, AI helping me figure out how to streamline different ways so I can take less time producing all this. So we love efficiency. We love efficiency from the beginning. We even in our work on a daily basis, we do process improvement and things like this. Yeah. And AI is just like our last or the most recent step of doing this. Um, adding efficiencies in the process and so forth. But yeah, it's, um, it's exciting and weird times, to be honest. So my relationship with AI started a little before the big, um, GPTs came out. Yeah. So we were doing regression analysis of like large datasets for building, you know, uh, just to correlate data with you know, performance behavior of what's happening in the building. This counts as AI. It doesn't count as machine learning. It doesn't count as neural network. It doesn't count as GPT, but still in the general AI umbrella. So maybe we'll start there. So AI is kind of the biggest umbrella you can fit machine intelligence under. Hmm. Then you have machine learning. That's subcategory. You have the statistical analysis we discussed, and then you have other models that are not very familiar to the public in general, but you know, it gets used for things like picture sorting for Facebook. There are different algorithms than the GPTs that we use to do this. Um, so this is kind of a general idea of my relationship with AI and what AI really is. So in 20, I think '22, when ChatGPT came out, this was like a miracle child. Everybody was like, oh my God, my computer talks to me. Oh my God, my computer can do this. Oh my God, my computer can do that. Uh, the research on this started in the '80s. So it was the DARPA project, which is the research arm for the Pentagon, funded what came out later on as Siri. So that was the very first GPT generative AI. So let's talk about GPT. So GPT is a generative transformer. So the transformer speaks to the architecture and Uh, the generative means it generates. So we see it generating text, images, audio, video. This is all coming from the generative aspect of it. And this was like a wow moment for everybody. And people start, you know, playing around with doing more productive things. Yeah. Let's have a to-do list. Summarize this for me. Yeah. Or write my email, make me sound a little bit more Clear Haze Exchange. Oh, absolutely. Especially when you come from like a place where English is your second language, like me, and you're trying to impress people. Yeah, absolutely. And also working international. So our office in Egypt, you know, we use this to just to make communication easy. Yeah. Um, so this was great. This was our first kind of productive things we do with it. Now. And then we start using code to help it generate code. And then, you know, people start using it to do all sorts of things, good and bad. But yeah, it, um, it definitely have came a long way since 2022, the last 4 years. And that's scary. When would you say was the moment you realized this was really going to change the way we were. So you kind of gave us a history of it, of how AI was already there. And I, I remember talking to a lady I ran into, but she was probably in her 70s and her major was artificial intelligence and she went to school in the '80s. So like she was majoring in this long before we kind of understood, I guess when I say we, um, as popular as it is now. Everyone knows about AI, but that term didn't resonate, um, back in the '90s and even early 2000s. So yeah, it, it had kind of its debut with the general public in 2022. Uh, before that, I started picking up on analytics that Facebook is doing, generally speaking, social media sites, how accurately it can identify us in pictures. Yeah. How accurately they can, you know, like find our locations, our likes. All of this was AI that was behind the scene. People weren't really interacting with. Yeah. Not in the form of a GPT. So you are the product in the Facebook case. Yes. But, but you're the user in the GPT case. So that's around the time that I closed my Facebook account and deleted all my social media accounts. I only have a LinkedIn account and that's it. So I went back, I'm like, right. So that was the moment, if you're asking about the moment in AI in general, but in the generative side, the like GPT, when I start using it to generate code and the code was better than I was writing. So just to give your listeners or viewers an idea, I'm about 45. I start writing code in high school or before. So I don't sound very nerdy, I'm going to say high school. Uh, so like I wrote projects in college, you know, nothing commercial, but I never stopped. And before we had to go online and search for answers and stuff like this, but now AI is answering code questions better than I am. So just kind of to start from a very basic way, and the name of this podcast is Clear Haze Exchange. So. The hope is that we can bring clarity to this complex system, artificial intelligence. There's so much data out there and so much to discuss. You can get really into the weeds, especially when it comes to engineering. So diving into kind of the practice of using AI for our listeners, could you break down for us today's tools and what they actually allow us to do? ChatGPT, you can use that for basic research. Emails, like I mentioned, trying to figure out the best way to understand something or pull from data, but understanding that dataset, you can use it for analysis, communication. Then you move into Codex, which is another level of that intelligence and AI agents, and that can do work with code data flows. So could you break down what those different levels of datasets and how they can be utilized. Okay. I, I can. So I'm not going to be too specific. Um, and I'm not gonna mention my references so we can save time, but everything that I'm going to say now, I read it somewhere. Now, this is important for the people to know that references are available if they want to know where this came from. Yeah. I don't like misinformation and don't wanna be one of the ones spreading misinformation. That's one thing. And the other thing is whatever that I'm going to talk about today, I'm going to keep it tool agnostic. So if you are, let's first talk about what tools are available. So for the general public, you have 3 providers, 4 providers of tools in general. That's OpenAI, Google, and then Groq and Claude. Anthropic Claude. So these are available to the general public. Amazon have one that is running Alexa off of, and that has been putting labels on your boxes for the last 10 years. But they make it available for businesses only for, we'll get to this at the end with the workflows and stuff like this. Yeah. 'Cause Amazon is fantastic workflows if you are making a cent a box. You need to optimize the heck out of the process. So Amazon is very good. So let's talk about the levels of the tools. So chat, you have Gemini Chat, you have Bing Chat, you have OpenAI ChatGPT Chat and so forth. These are usually good for quick answers. Something that's either available on the internet, just don't feel like searching for it, or, uh, something that's not in depth. It's hard to gauge what not in depth is, and this is changing by the second. So as of this recording, the most recent model is the GPT-5.6. So if somebody's listening later on, we could be on GPT-20 and you just don't know. So this is, uh, 5.6 GPT. And so the chat is good for the topics we discussed. Now in the chat, you have a mode where it says deep thinking. Mm-hmm. So deep thinking, if you don't want it to find the first answer that it find, we want it to spend time looking for the most relevant one. Mm-hmm. So fast will give you the first right, correct answer for your question, not necessarily 100% relevant. Then think deep, give it a little bit more thought, a little bit more thought, but it's still chat also. So next level will be kind of an organized chat that can be a notebook and Copilot and Codex or a project in Anthropic Claude or whatever the case is. Names change, but the concept is the same. So it's kind of a notebook where you can put files in it, organize documents, and you can ask it questions about these documents. You can have it do research for you. So this is more of, let's say, an end of college year. Project, not graduation project kind of stuff. So I'm trying to make it relevant to the audience. Like, yeah, you can put your notes from the classes and how to prepare for the exam. Yeah. So this is very good for that notebook or equivalent. Then you have, um, every one of the provider give you pre-built agents. So you have the researcher agent, the analyzer agent, the coding agent, and so forth. So these guys, their names are very helpful and they are very strong in their job. So if it's a researcher agent, it can take a day working. So this is not, I'm, you're gonna stay in front of the computer until you get the answer. This is important enough for you that you're willing to wait a day for a lot more information and a lot more accuracy. Yeah. So there's the prebuilt agent. And then you can custom build your agents. Mostly research agents, kind of a static, you give it documents to review, uh, documents to summarize. It, you can hook it up to your email and can give you like email digest. Um, and it also can be customized for code. So when, you know, when we talk about customization, that's later, we can talk about this. Um, but the customization really is advanced though. Like you, we have software engineers in the business and they laugh at the instruction that they give to my agents. And I am, I think I'm very detailed. I think my instruction is good, man. Like, it's like, they give me a hard time. No, it's not good enough. Okay. So what is good enough? And when I read the instruction, I know that it's not good enough. Like they give 4 pages of instruction to an agent. I give it like 4 paragraphs. Yeah. So that's the difference. I'm just gonna mention it briefly. There is inherent lack of predictability in a statistical model. These are all statistical models. Yeah. They are not 1+ 1= 2. They're what, what's the most likely answer to 1+ 1 equal is. Uh, so. 1.9, 2.1 are likely because it is a statistical model. So you add them up, it guesses an answer. Yeah. So for you to prevent it as much as possible from trying to guess if 1+ 1= 1.9 or 2 or 2.1. Yeah. You have to give it specific instruction. Use a calculator. Yeah. For example, that's an instruction. Don't guess at this number. Use a calculator. So you can go as far, as deep as this, because when it's doing 1+ 1, there is no calculator in it. It's guessing at the answer. It's gonna give you a statistical answer. Yes. Well, there's a few studies that show that 1+ 1 is 1.9756. Speaking of studies, there is actually a statistical study that proves that a coin flip is not 50/50. Uh, okay. It's $49.51. So, so one thing that comes to mind is when you mentioned that is that makes sense. Cause AI is looking at research available. I also, I heard this and I'm curious the validity of it. AI can go only go back to a certain year to reference information and pull it from. Is that correct? Okay. So you're paying $20 a month to get the service and it takes $20 million a month to run a server that the AI is on. So I'm sure there is a limitation that even cost-benefit analysis, like how far back do I really think that this will change your answer? So I think answering your question straight out, I don't know, but I can see why would they do something like this just to, from a cost standpoint. These are massive data centers that are using very expensive hardware and like a small city kind of electric infrastructure. Mm-hmm. So makes sense to me. Understanding the limitations. There is a limitation on what research you can do, and that is critical, right? To know. So there is a couple different limitations here. So the first limitation is training limitation. what data was it trained on. So part of this, the most recent trainings have utilized hardcover books because they, the models on the internet already have consumed everything on the internet. So now they're scanning hardcover books and feeding it to the agents. Amazon is doing it, Anthropic is doing it. It actually was revealed, this fact was revealed. In a lawsuit against Anthropic that they were buying old books, they were scanning them, uh, feeding them to the model and then throwing them away. Uh, so this is a fact that your viewer can validate from the lawsuit information. Um, so this, so this is the limitation number one. I know this was a digression, but it's all helpful to the people listening. Um, limitations on training. There is limitation on what the actual models can do. And then there is limitation on the tool itself. Are you using the right tool for the task? So let's start with, so we talked about that it's a statistical model. So this, this is a limitation because when you want 100% consistency, you may not get that. Different models have different training. So ChatGPT-2 or GPT-2.0. And 5.0 are completely different in their capability, structure, size, and, you know, the first one will eat the second one alive. So, or the other way around. Um, so that's the model itself abilities. And then there is the tools. So some of the tools like ChatGPT, if you're trying to run like school homework, maybe, um, a quick project, probably, but really if you need memory. If you need reference different documents, if you need to have more deliverable, more than one deliverable, then, uh, a notebook is good. An agent is overkill. So you need to know what's like too little and what's too much. Yeah. You know, kind of, so you don't drown your plant, you just need to give it the right size water or the right amount of water. So agents is when you want to implement workflows. So they can connect to APIs and can connect to your organization and they can connect to your data. Now, whether you want it to or not, that's a completely different story, but it can. Um, so I think agents are more for if you want the custom behavior that's as consistent as it can be, it's an agent. Okay. Uh, but where are people underestimating AI because they think it's basically a chat box and where are people expecting too much? So kind of the variance between what you just explained? I think this, the problem with AI, it's self-taught. Everybody is having their own experience. And it rely on 3 things. What your technical knowledge before you went into it, what tool you started with, and what task did you need to accomplish. So if you ask 10 people, You may get 10 answers on how useful AI is because everybody have went into it with completely different experience, expectations, and stuff like this. I think once you have gotten past the chat, you should advance to something like Notebook or, uh, I'm using, well, uh, specific terminology that I'm used to because of this specific tool that I use. So think about like your, in a, in a computer, your processor do the processing and your RAM do the storage. In AI, it's doing both. So it, whatever the abilities is, you can have too much intelligence and not enough data or too much storage and not enough intelligence or the right size. So you're always, we're always looking for what's the right size task for an agent. Okay. That makes sense. And it's not static because they are not stopping developing them. Yeah. So, yeah. So thinking about what we've learned using AI in our business, so we work together in developing professional services at EA Energy Solutions. We work a lot with architects, engineers, building owners in the development stages of the design through construction. We also work through the construction process. with teams, um, for green building certifications, so on and so forth. There's a lot of data, um, especially in energy modeling. Um, and we're advocates for utilizing the ability to gather that data as you're making critical decisions, especially when it comes to building infrastructure design decisions. So from that explanation on what we are doing on the business side, but now back to how we're using AI to help build our process as a smaller business, um, to streamline in developing business? What's one task where AI has genuinely changed how we work? But what did we do before and what does the process look like now is kind of the question that I want to talk about. So two things comes to mind when you ask me this question. The first thing is we learned that they are lies. Yeah. Yes. Intentionally. It's not, it's not confused. Yeah. No, it's not. It's a reaction you would attribute to somebody who's lazy. But I know that's not the case. I just don't understand what they're saying. So we're just calling it lying. Uh, so AI lies. So we have to be very careful with where we insert it. Yeah. Like if it's a responsible task that will lead to consequences. Yeah. That has to be reviewed by human. We don't skip steps into AI. So this was the lesson number one. The repetitive tasks that required intelligence before, we were automating the boring stuff, right? Yeah. Now we're still automating the boring stuff, but it requires minute intelligence. Yeah. Intelligence, I should say. So what we do is, let's say we process documents through AI. So drawings, project drawings, say I want to get out of this drawings, maybe the square feet of the building, or I want to create some sort of a schedule with Let's say the finishes of the room or something like this. So we have this, we tried this 7 different ways and we failed in the 6th, 6th measurement. So this is a very sophisticated, what appears easy on the surface. Mm-hmm. Turn out to be holy moly. Like really, does it need all of this? Yeah. And the re— and this is when we get into the technical core. We really don't understand how computers work. Like, we have outsourced this to the process a long time ago. It's like, we don't care, we just want, like, you know, our Facebook 7. I know I'm picking on Facebook a lot, but it's not personal. It's just, I'm old and this is the name that I still remember. But, you know, this becomes crucial when you're dealing with AI. This is a funny story, to be honest. So we created a notebook, we upload the project documents in it, and we're starting to quiz the notebook about the drawings that we uploaded there. In the setup of the, of the notebook, we did not close the company resources, so it could still access the company docs, right? So I'm asking it, um, what's the building total square foot to come up with By 10,000 square feet or so less than what I expected. Okay. This is not your couple digits wrong. That's a completely wrong number. What happened? What happened is it have read a document in the past that had the square feet number in it that was on our shared drive and just spit out that answer instead of looking into the document. So if you don't restrict it from something like this, you don't guarantee the quality. If you don't give it enough instruction, open this document, get the information from it, put information in Excel file, close Excel. It's like sometimes you have to be this literal for you to get the instruction that you, or the information that you want. So what have we automated? I would say for the, we automated a lot of the entry point information. That's taking a 5-year experienced engineer having to do. So now we have a 5-year experienced engineer and an AI agent doing this. So it's not this or that. I think that's the future. Like the idea is AI is going to replace us. We're not here yet. And that's personal opinion. Yeah. Um, yeah, you can chain them as many of them you want, but you need like a, half a million server for, for you to replace me. It's not cost effective. Yeah. So I think the future is more of cyborgs, human enabled with AI. Yeah. Yeah. That makes sense. For, for those, um, feeling nervous about the replacement of jobs in some ways that, that, you know, it, it does make sense that there are jobs that are getting, um, eliminated because of AI, but the reality is that there still has to be, like you said, the, so it has to be some level of intelligence that can think in a way outside of what AI thinks. I think you gave me an example and I'm trying to remember what you said. I remember the conversation with all the information freshly squeezed, so I can't remember what I said at the time, but this sounds very true. Um, I think Geoffrey Hinton, the godfather of AI. So this guy is, I think, the only software engineer that received the Nobel Prize for inventing the neural network. He said that AI intelligence is jagged intelligence. So it's not the same in every task. Where us, you know, it kind of, if you're a smart person, I don't wanna categorize this, but You know, like our intelligence is like up or down together. So that jaggedness of AI intelligence make it shocking when you ask it to do something and it's doing it great. And you ask it to do something about the same intelligence, a flabby drool. That's, I think, because of how it's trained and how people focus on certain aspects of it, knowing the users. So they focus on training, improving the skills, and they let other one go. So let's go back to the, what I just said about cyborgs. So the replacement of employees is not going to be the way that You hear about it everywhere, but my personal opinion is based on what we're hearing, the majority of the people who are not going to get hired are entry-level positions. Mm-hmm. So now I can have, let's say I can have a marketing analyst or an accounting analyst or a paralegal, but all of this, like regardless of the field, I'm mentioning plenty because it's not field specific, regardless of the field. Now with AI, I can replace 5 paralegals with 2. So the 3 that just graduated, what do they— I think this is the question that we need to recognize. It's not that it's going to— you as a senior person, I know people are like, uh, AI, like even some CEOs, like, I'm worried, see, AI is going to replace me. Um, I think you haven't been putting effort, man. This AI is gonna replace you. I think you didn't put much effort. Yeah. So personal opinion, I viewed massive amount of videos and read articles about this just because I'm trying to educate myself about my own future. Mm-hmm. So not as an employee will get fired, but as a business owner, how can I keep educating my staff and so forth. Mm-hmm. Uh, and yes, as a business owner, you will sit and try to think about it 2, 3 different ways. Should I hire an AI company, come build an AI system? Should I hire people? I think this decision was relative to so forth. Okay, so we did all this over the AI and we hired 4 people this year, right? I don't know. So. It's not that we are leaning one way or the other. I think it's a combination of both. This is basically find a successful recipe for what you're doing. Mm-hmm. Number of people, AI, and so forth. But for anybody out there watching and thinking that AI is going to replace me, I think either your job should have been automated to begin with if you're being replaced now. Yeah. Or, um, the company is just mindset is we're going to automate everything. Yeah. And we're reading and hearing about companies rehiring people because they thought that they could eliminate— they cut too much people. So thinking about on the same track of what we've learned in our own experience in the business using AI, can AI give us capabilities that previously required substantially more people time or resources? And you kind of answered this already. Your example was 5 hires down to 2, but how does that scale across other areas of intelligence, like coding? Because I know coding, you still need a person. So is that scale that goes across even technical areas of expertise, or do you think it's just that entry level? No, this is where the operator comes into play. Yeah. So It's, it's basically level playing field out there, or almost. Um, yeah, bigger companies have more resources, but that doesn't necessarily mean more success, by the way. We have, we have, I don't know if this was mentioned or not, but we have a software development team mostly writing code for building optimization and process optimization for us and building optimization for our clients. And the team lead, uh, have master's degree in AI and he's helping us with this. And this is an information coming from the research that he was working. Um, the tools out there, there is no special things behind closed door that we're not seeing, except maybe for different governments. What you need right now is idea, creativity. Mm-hmm. You don't need muscle mass. Yeah. So hard the smart, not the hard work way. The lazy, actually somebody said this, but I don't know if it's, you know, very good to say, but out loud, they said hard the lazy, they will figure it out because they just don't wanna do the job. So they will figure out the best way to do it. It makes sense. So thinking about what happens to the energy and building industry next, could AI finally help owners move from here's another dashboard to here's what's wrong and why it matters and what should they consider doing next. You're talking more about hiring people with intelligence, creative ideas. So where would you say AI is going to help more building owners in this space? The AI tools that existed to analyze building data have always been around. What's beautiful about the tools that we have now It's, it's animating me. It make it talk to you. Yeah. So you can have a conversation with the building instead of reading a report. Yeah. So I always dreamt of that day and I'm now trying to make this dream come true. Yeah. So we're working on building that chatbot that will make you basically mind-blowingly easy to know exactly anything that you wanna know about your building, whether we're pulling this information from your utility bills, from your BACnet server, from surveys we're doing with your employees, whatever resources that's needed, sensors on the roof. And then you can have all this data augmented on your phone. And you know, this like, uh, chat agent summarize my meetings tomorrow or help me prep for my meetings tomorrow. Yeah. So it's like, help me plan this or help me do that. It's. It's constantly bringing you updated information and stuff like this. So this can be a way that AI can help building auditors. Mm-hmm. Um, this also another way that can help building auditors is, um, or building consultants, I should say. Uh, a lot of the tools that we discussed in analysis, in audits, in design, um, All of these are tools that can be customized. So it, it is leveling the playing field somehow, but you have to have the right ideas. Yeah. And I think it goes, the other thing you mentioned is working on, you know, solutions, cuz this is a space that we're in. The critical challenge is there's so much information out there and so much competition with the softwares that can pull information about your building. So you have this kind of information war. Around what is useful and should I hire this company? Cause they specialize in getting me this information. Because what it comes down to is then after you have all this information, what is actually useful to then go and make decisions to build a business case that's actually profitable. So I think as well that there's, there's a lot of information and dashboards out there, but back to your point about creativity, there has to be a way for a human consultant to understand, here's what you have, let's look at automating this, animating that, and this is how it will help you. You should kind of need that, that kind of guide, that in-between guide in all of this, because there's so much information out there. So, great point. Uh, Noah Hawari. I think he was a, uh, a researcher. He said that we live in a time where we have so much information that it's very hard to find relevant information. Yeah. And it's even harder to find relevant and useful information. Yeah. So I think You start with the problem. What is the problem you're trying to solve? Is it automation problem? And really get into the ball, spend time. Like, let me bring people a little bit on this journey. We spend about, now that AI can be our muscle, so to speak, we spend criticizing our own ideas and our own question the majority of the time. Are we formulating the question wrong? Yeah. Are— is this the real answer to our question? Is this the real problem? So we're spending most of our time now not figuring out solutions, but really defining problems. So if you spend enough time defining the problem, then it will crystallize to you what information you need. Yeah. And this has been the most successful way I found to drive solutions, not research. If you wanna go in circles and do research, start with the solution and then try to find the problem that will fit to. Yeah. But if you have a real problem that you're trying to solve, spend time defining it. The muscle mass is there to help push you across the finish line. That's AI's job. Yeah. So criticize your ideas. I think this is what, where it's the most beneficial. Yeah, that's good. I mean, this is crash course in understanding AI. Like, there's so much to know, but these are really important nuggets, uh, to consider and to utilize as, as you navigate forward. The time has, has flown by, and we talked about doing maybe a, um, a part 2 because there's so much more to talk about here. Yeah. Happy to talk about this. But what knowledge comes from 20+ years of actually working with buildings? In clients that you don't believe an AI tool can replicate? AI cannot solve a brand. Mm-hmm. It can make recommendation to a lot of things, but at the end of the day, a human have to evaluate this recommendation. AI will not have the full picture unless you plug into it a lot of data. Yeah. So based on the tools we have today, AI by itself may not be able to troubleshoot. Let's go back to your, uh, electrician example, troubleshoot that a wire is loose in an HVAC unit. Yeah. But it'll tell you that I can't get any signal from your fan. So it's not a problem solver tool, it's a pattern recognition. It's so it can dive and say, oh, this is the most likely 3 answers for what you're facing. I think at the end of the day, uh, the role of human is to formulate all this information in an actionable solution. Yeah. So that, that's what I think where AI is. It will get there if we see it flying drone over our heads. We know that it can propose solutions. Yes. Yes. So for someone who's listening to this and they must be living under a rock, but if they barely used AI— oh no, I don't wanna say that 'cause a lot of people just don't even touch it because they don't like removing human thinking. But I see different train of thought when asking people about AI depending on the generation. The younger generation is checking everything out, getting, soaking in all the information, streamlining as much as they can. Whereas the older generations are, well, we used to, you know, they're used to pocket bank calculations and they'll get irritated by doing things different. We were just talking about a planner that I started to, to organize things. The reality is if I don't write things down and keep track in a centralized book, it's gonna be harder for me to go back and find where did I put that file or what is this? So, and it becomes more of a balance. I know you have plenty to say on that one. Well, I don't think we have time to answer this question as it deserves because you touch on so many things. I did. Um, like my dad is trying to teach the chatbot, don't do that. The chatbot doesn't learn from you. So my dad corrects its mistakes. My dad is a university professor and he always work on books in his retirement. So it have it solve problems for him as a TA, basically. Yeah. Uh, and then it does mistakes and then my dad try to talk to it, to correct it, and it just doesn't learn from us. Yeah. It get— it learn on the server where it was generated. What you tell it is just information it keep in its short-term memory. So every agent we have have like, or every tool we have have a 6-month short-term memory. So every day, the day, I don't know, 183 gets dropped from the backend and it adds today to the memory and so forth. So you may feel like it's remembering and it's learning, but that's just an illusion, like, um, bought and paid for by you. Yeah. Um, so this is one thing. I think there is so many good, um, that can come from this, but the reliance itself have a little bit of the bad aspects of it. You talked about the library. That's the skill. There is, there is a step that we are skipping, JR, which is the search step. It is doing search, not it. Yeah. Sorting out what relevant information, your ability to discern signal from noise, knowing what's important and what's not important. We're conceding this to AI outright. Mm-hmm. So I personally use it after all what they said. I personally use it as little as possible. Yeah. So I use it mainly in research. It's very good at collecting information, format it, and if I want to learn about something, I'll just cut it loose. Yeah. Uh, I think some of the personal experiences were, you know, some unfortunate situations happen with people who committed suicide and stuff like this. Yeah. Think of social media on steroids. So you go to social media so you can find validation. And we can debate this for hours, but it's not worth debating like that. Yeah. There is nothing that will give you more validation than AI engagement. Yes. It's not, I ask the questions like, great question, Mohamed. But tell me, it's just like, what 5+ 5 is equal? Oh, fantastic. So if you're looking for validation, nothing will give you better validation than AI. Yeah. It's, it's perfected that literally. Yes. And so you know this, but your audience may not know that my wife is a patent examiner. So she sees what's coming into AI and some of the stuff that, you know, I happen to glance on is some stuff related to, uh, emotional intelligence and scoring the agent emotional intelligence and its ability to, to recognize our needs on an emotional side. That's completely the Pandora's box that we may have to crack in our lifetime. Wow. So I, again, you mentioned so many things in the question, so if I missed something, just circle back. I didn't really finish the question, uh, cause I got carried away. Cool. Like AI, there's so many things, uh, to discuss, but the, the point was for someone listening that's barely used AI, what's one useful thing you would tell them? To try this week? Learn something that you always wanted to learn. Use a research agent to summarize something for you or create an audio podcast, like 5 minutes of that. If you always wanted to take it to the next level. So if you have the chat experience, go to a notebook or XAML or so forth. If you have a notebook experience, Try, try and fail with agent or call me, you know, just like Dwayne. So we have coworker, you know, he's building multiple agents right now and I'm getting all calls of all hours of the day and night to talk about, hey, why isn't it doing this? So if you run into why it's not doing this, just call me. Um, but try it with like, I think it's the best experience and still try. Yeah. So I think these are two good things to do. Yeah. And don't forget to learn, 'cause AI is making not learning very easy. Yes. Yes. Yes. And I think before you type in a question, it's like, how might I phrase the question better to give more detail, like you had mentioned, about where I'm trying to get at. Don't stop learning is very, very important because I had to understand, and you don't learn unless you have errors. And yeah, when you have errors or you, you end up using something of like, oh, I didn't check that. Um, AI gave me this information, but I didn't check the resource that it, it put on there. Speaking of, uh, speaking, just to add a little bit on what you said, because we're professionals, we're professional engineers and professional architects and, you know, landscape architect and so forth. So we are professionals and we stamp and seal and sign and we have You know, a duty to the general public. Mm-hmm. So how do you fit AI with that duty? Well, you better review what you're putting your stamp on, cuz as far as I know, AI is not going to stand in court instead of you. Mm-hmm. So we are as professionals still responsible for everything that comes out, whether, whether that's done by a junior staff member or whether that's done by AI, honestly, cuz that always existed. You have. Junior team member working on projects and you review any STEM. AI is just another one of these. Yeah. So, yeah, that's good. Well, thanks again for coming back on the show. I think the biggest takeaway here is be careful how you use AI and, but at the same time, learn it and understand how it generates the answers that it gives you. Don't be swept off your feet by the first answer you get. Yeah. Scrutinize. And, um, I wanna say something just as we close. Sure. Um, if you feel a strong hype, it's for somebody's benefit. Mm-hmm. AI right now is, uh, very good and very wrong in some instance. So it's not the doom and gloom you hear. Yeah. And it's not our Lord and Savior either. Yeah. So. If somebody is selling a strong statement, there is a financial benefit there. Yeah. So, and I'm giving a strong statement now, but I don't have a financial. So most likely since I caught myself in that, most likely if there's somebody making a strong statement, there's a financial benefit to go with. Mm-hmm. But right now AI is just like any other tool that existed. A human can never more equivalent, just like the ATM machine and all the other AI that you were interacting with before. Yeah. It was out in the open. Yeah. All the algorithms that controlled our life, the security algorithm that make you check the picture, hey, are you human? Yeah. That have always existed. That's AI. You interacted through it from behind the scene, but it was interacting with you without you know. Yeah. So it's not doom and gloom and it's not saving us from anything. Yeah, just to do that, maximize it. Well, thanks again for tuning in to this episode of Clear Haze Exchange, and we will look forward to the next episode. Thanks for listening to Clear Haze Exchange. New episodes drop every 2 weeks. You can catch Clear Haze Exchange on Apple Podcasts, Spotify, Pandora, iHeartRadio, and more. Let's clear the haze about sustainability and energy efficiency together. Subscribe now and join the conversation.