April 9, 2026
How AI Agents Are Reshaping the Solar Industry Right Now? #343
A 10-person, three-month estimating process. Compressed into 12 hours by a single AI agent. That is what Jesse Anglen, co-founder of Ruh AI, is building for construction and solar companies right now. In this episode of The Clean Power Hour, host Tim Montague sits down with Anglen to break down what a digital workforce actually looks like in practice, how solar contractors and EPCs can start using agentic AI today, and what it means when AI agents take over knowledge work at scale.
Episode Highlights
- Anglen breaks down the three categories of AI agents and explains why most of what people call "agents" today are not actually agents at all.
- One construction firm with projects in the hundreds of millions of dollars had a core operational process that took 10 people three months to complete. Ruh AI turned that same process into an overnight task.
- Solar contractors are sitting on a lead generation opportunity that most have never considered. AI agents make it possible to act on it at scale, at almost no cost.
- The administrative burden of running a billion-dollar construction or solar company is staggering. Anglen explains exactly which back-office functions AI agents are already handling, and what that means for headcount.
- Anglen gives a clear breakdown of what it actually costs to build and run a custom AI agent, from the entry-level option any business owner can start today to the complex systems designed to replace entire departments.
- Anglen shares a number that reframes the entire AI conversation. It is not about chatbots or writing emails. It is about the total size of the knowledge economy and how much of it AI is already capable of doing without a human in the loop.
Solar and EPC companies are already operating under margin pressure, competing on thin spreads while administrative overhead continues to grow. The tools Jesse Anglen describes are available today, at a price point that is lower than hiring a single full-time employee. The window to adopt these systems before competitors do is narrowing fast.
Connect with Jesse Anglen
Linked: https://www.linkedin.com/in/jesseanglen/
Website: https://www.ruh.ai/
Connect with Tim
Clean Power Hour
Clean Power Hour on YouTube
Tim on Twitter
Tim on LinkedIn
Email tim@cleanpowerhour.com
Review Clean Power Hour on Apple Podcasts
The Clean Power Hour is produced by the Clean Power Consulting Group and created by Tim Montague. Contact us by email: CleanPowerHour@gmail.com
Corporate sponsors who share our mission to speed the energy transition are invited to check out https://www.cleanpowerhour.com/support/
The Clean Power Hour is brought to you by CPS America, maker of North America’s number one 3-phase string inverter, with over 6GW shipped in the US. With a focus on commercial and utility-scale solar and energy storage, the company partners with customers to provide unparalleled performance and service. The CPS America product lineup includes 3-phase string inverters from 25kW to 275kW, exceptional data communication and controls, and energy storage solutions designed for seamless integration with CPS America systems. Learn more at www.chintpowersystems.com
The Clean Power Hour is produced by the Clean Power Consulting Group and created by Tim Montague. Please subscribe on your favorite audio platform and on Youtube: bit.ly/cph-sub | www.CleanPowerHour.com | contact us by email: CleanPowerHour@gmail.com | Speeding the energy transition!
WEBVTT
00:00:50.780 --> 00:01:19.900
The things that I've done so far that are the most exciting to me would be like, we had a we had a guy that owns a large company, they only do like, you know, billion plus projects, and maybe, like, 800 million plus projects. And so when they go to do an estimate, it takes roughly 10 people three months to put an estimate together for them, give or take, and we built a system for them that can do that entire estimate in one night.
00:01:21.340 --> 00:01:53.960
The clean energy industry is moving fast. The deals are getting bigger, the technology is evolving, and the stakes have never been higher. Welcome to the Clean Power Hour, the podcast for solar, storage and micro grid professionals who want to stay ahead of it all each week, your host, Tim Montague, industry advisor and president of clean power Consulting Group, brings you unfiltered conversations with the leaders actually Building the energy transition. Now here's your host, Tim Montague,
00:01:55.460 --> 00:02:25.632
today on the Clean Power Hour, a digital workforce is coming for your job, but it's going to help you grow your company, go faster, cheaper, better. You know, I am a catalyst for the clean energy transition. That's why I created the Clean Power Hour to help others go further faster into the energy transition. Well, my guest today is a gentleman who is building a digital workforce and helping companies automate aspects of their operations.
00:02:21.980 --> 00:02:29.860
That is really going to blow your mind. So welcome Jesse Anglen, the co founder of Ruh AI Tim,
00:02:29.860 --> 00:02:33.599
nice to nice to hang out, man, I'm looking looking
00:02:33.599 --> 00:02:36.900
forward to this.
00:02:33.599 --> 00:02:37.740
I'm so happy that your bot found me on LinkedIn.
00:02:39.000 --> 00:03:17.680
Yes, I am. I am as well. You know, I try to put my money where my mouth is and actually use, use AI agents to do work. For me, it's funny how I meet a lot of people who are like, I love AI. I'm like, How are you using it? And they're like, Well, I use chat GPT to help me write emails. Like, okay, yeah, yeah, I love AI, too, like that, but like, you do know that you haven't even begun to scratch the surface, if that's all you're using AI for is, is you go use a large language model to help you write emails. It's fun when you start getting them actually doing the kind of work that you have to get done, and it just does it on its own without thinking about it.
00:03:18.580 --> 00:03:55.099
It's a strange time, because we're very early in the AI revolution. Only a fraction of knowledge workers are still using llms, Chachi, BT, Claude, perplexity, Gemini, et cetera. But at the same time, the agentic aspects of the technology are getting good enough where you can set them loose and they will do real things, and that's what we're going to talk about today. But Jesse, for my listeners, paint us a picture of how you got interested in AI. You're also a web 3.0 entrepreneur, so just give us a picture of why you're here now.
00:03:56.180 --> 00:04:11.500
Yeah, I was just trying to solve problems. So I had a I had a large software development company that was building decentralized applications on blockchains. And when I say large, it wasn't large, it was like 370 developers that worked for me.
00:04:08.900 --> 00:05:00.439
And one of the, one of the issues that I had was that we only worked with startups and first time entrepreneurs, or at least people who were building novel technology for the first time, which meant that as a part of our process, I had to sit down with a guy like you, and you had this crazy idea in your head, right, that you now want to get built and bring into Reality, and I had to somehow pull that completely out of your head, and I had to get it into a concrete format so that I could give it to a to a developer, to a software engineer, and they could build it, and if a mistake was made during that process, the engineers would build the wrong thing, and I would deliver You a product that wasn't what wasn't what you wanted, and then we would have to go back and make changes and fix things.
00:04:56.180 --> 00:05:20.500
And, you know, I don't know if you've worked, you know, solely with startup founders, but they're often crazy, ethereal vision type people. They're not good, oftentimes, at articulating the details of what it is that they want. And in order to do a really good job of building a piece of software for someone, you actually have to get all the details. And so I had this massive bottleneck.
00:05:21.100 --> 00:06:21.940
Well, you think we'll go hire people for it. I tried, but there were very few people that I found that were able to sit down and have those conversations. And so the process for me to sit down with someone took about six weeks. I found two other guys that could do the same thing, which meant that we could each onboard. We could onboard like three clients at a time, and we weren't going to be able to grow and in early 2022 we a guy that worked for me, Hansel, one of the smartest guys I've ever met, and I sat down and we thought. Okay, we have to solve this problem. We found this tiny little company called Open AI that had created something they called a language model called Da Vinci 01 and they actually, right after we started messing around with it, released another model called Da Vinci 02 and these models were really weird, because you could talk to them, and they almost sounded smart, like if you had a simple conversation, they could fool you that they were, that they were completely human. If you had a complex conversation, everything would fall apart.
00:06:19.179 --> 00:06:56.179
They forgot everything after about three paragraphs that you had said prior. I think their context went on. The one of them was 2000 tokens, or about 1700 words, and the other one, I think, might have been 5000 tokens, or about, you know, 4000 words and and we thought, well, you know, let's, let's try to make a system where you give all the transcripts from the conversations that you're having with the clients or the we brought on, and it produces questions, and then we can record all those questions and slowly feed those transcripts into these models and see if it can make all the documentation.
00:06:56.779 --> 00:06:59.840
And so we over engineered the system. Took us several months.
00:07:00.259 --> 00:07:13.480
But the thing that was nuts about is it worked. It took our six week process and turned it into a four day process that not only myself and those other two guys could do, but that probably 10 people on my team could do.
00:07:10.340 --> 00:07:32.700
And so all of a sudden, bottleneck gone, and I and I thought, Man, this is crazy, like it's like having access to almost like having access to intelligence on, like at an API call. And at the time, I was talking to a friend of mine, and he's really the guy who inspired me and got me really excited. We were talking about photographs.
00:07:32.820 --> 00:08:49.860
And what happens when you take a take a biological problem, or a chemistry problem, or any problem, really, and you turn it into a math problem, you're able to scale that solution infinitely. Chemistry is a really good example when you look at photography. Photography was a chemistry problem for years and years. And as soon as they took the chemistry problem of photography and they turned it into a math problem, and we got digital photography, all of a sudden, pictures are disposable. I was talking to my son this weekend. We were fixing his truck, and I took a picture of something before we took it apart. And I just had this random thought when I was a kid, when you went to do that, you couldn't have done it. Like, I wouldn't have got out the the camera stuck in a roll of film, took a picture, brought it down to Walgreens, developed it, and then eight days later, you know, picked it up and then had that as a reference so I could put something back together, like it was just unreasonable. Well, now we take pictures like, like they're, they're worth nothing to us, right? Like they're, they have no value. Well, in my mind, I made the association between the chemistry problem in photography and the biological problem in human labor. And I thought, I'm going to, I'm going to pursue this as far as I can.
00:08:46.379 --> 00:09:23.439
And I want to figure out how to get computers to solve work problems like, typically, the only only you know that humans have to do, that biology has to do. And so we started working on stuff that year, 2020, in 2022, in November, open. AI released chat GPT 3.5 which is the one that everyone kind of started getting on board with and what was interesting about what we were building is that when they released that and we plugged those models in every all of the work that we had done in 2022 became 100x smarter overnight.
00:09:23.980 --> 00:09:39.539
Wow. And it blew my mind, because the systems were already useful, and they really became, overnight, something that that like, fundamentally, was going to change my entire business.
00:09:39.539 --> 00:09:48.839
So give us some context, though, on what, when you say, making it 100 times more valuable. What exactly were you creating?
00:09:49.860 --> 00:10:10.159
So when you, when you when you take somebody and you pull an idea out of their head, and you have to go build a piece of software. It requires a product specifications document, which basically says, this is the product and here's all the specifications. It requires a Technical Specifications document that says, here's the tech stack we're going to use, the technology we're going to use, why we're going to use it?
00:10:10.159 --> 00:10:29.559
Here's the architecture that we're going to use. These are like, here's the database. These are the schemas. Is that we're going to organize things, because you want to put it together in a way that you can scale it. It's kind of like if you go to build a house, no one just dumps a pile of lumber on a piece of property and starts building stuff, right? Like, generally, they have a plan.
00:10:27.339 --> 00:11:06.980
Because if you just start hacking things together, and you just start building like you're gonna you, you'll get something like that is true. You'll have a house, it'll have a roof, it'll have windows like, you know, but you might, you might not have a bathroom, or you might have to stuff it into the garage when you're done, or, who knows what'll happen, right? And so at the beginning of the software project, you really want to take time to understand what put a blueprint. Other figure out what it is you're building. Well, that ends up being, for most products, around 250 pages of documentation that allows a developer to always have answers to every single question that they could ask. Now, it doesn't work out like that in real life.
00:11:04.519 --> 00:11:14.980
You still have to meet with the client. There's still other stuff that would come up. You know, there's another 250 pages of documentation you probably come up with, but that's what those models were doing for us.
00:11:14.980 --> 00:13:48.649
Was they were We were feeding them the transcripts from calls that we would have with them, they would then come back to us with questions we needed to ask in order to gather the requirements that we needed to build the project, and and then once it had all of those questions asked, we would take all the transcripts from those calls, give it back to the models, and the models would put together the documentation. And when chatgpt 3.5 came out, it had a 16,000 token context window like that was like 10 pages of text which was which was incredible, compared to the, you know, 10 paragraphs of text we had access to before, and so we would feed that stuff in. And we had an interesting system for making it work, because we were coming up with hundreds and hundreds of pages of requirements, requirement documentation. And so we did some clever engineering, but we got this thing to actually output everything for us. No manual writing, you know, no anything. I look at it today and it's like, well, yeah, of course you can do that. It's like a super simple task today. But back then, in 2022 that was that was like magic. And when chat GPT 3.5 came out, and I watched what it did, I decided that in my company, anytime anyone left, or anytime anybody had to be let go, we were going to try to replace them with an AI across everything. And so we did. And over the course of the next, I guess, two years, we went from 370 people down to about 85 people. And we were doing the same amount of work, if not, if not, we had probably grown 10 or 15% in the amount of work that we're
00:13:48.829 --> 00:13:57.710
doing about rapid innovation, your software company, yes, and you're making software for what audience
00:13:59.569 --> 00:14:17.049
it was, it was back then, it was really all of the entrepreneurs In the blockchain space, was the niche that we were in. And so we, you know, we were building out exchanges, and we were building out, I mean, everything that you can imagine inside the inside the blockchain space. I mean,
00:14:17.589 --> 00:14:23.549
let's put our finger on a few things, banking contracts, smart contracts. It was almost
00:14:23.549 --> 00:14:45.889
all, almost all smart contract development. So it was, it was decentralized exchanges and so like swap contracts and things along those lines. It was, it was, we did several things in the supply chain in order to, like, track products and goods that were manufactured all the way, you know, from kind of the farm from farm to table tracking for different supply chain things.
00:14:45.950 --> 00:14:54.413
Okay, well, let's fast forward. Let's fast forward and get into the roux AI story.
00:14:49.769 --> 00:15:35.309
Fast forward to 2025, in April of last year. So a year ago, you, you founded Ruh AI, r, u, H, A, I, check it out. And what was the seed for that. And obviously the agents, or the AI tools, have gotten much, much better now, and now we're in the takeoff. I mean, technically, we've always been in the takeoff, but we're starting to see elevation, right? We're starting to feel the speed at which things are getting better, and the recursivity the AI is improving itself. The companies open AI, etc, right? Will tell you that 90% of the work is being done by the AI itself. Yep.
00:15:35.309 --> 00:16:04.250
Well, what's funny is, when I used to tell people that AI was going to change the world back in 2022 I would get laughed at because they're like, I wouldn't use it. It's like, I don't understand what you're talking about, like, you don't see it. And even in 2023 where after chat GPT 3.5 was released, they're like, I tried to use it, like it's useless. And it's funny, because I was literally making, you know, like, gobs and Gods gobs of money using it, and everyone else was telling me, oh, it's useless. It can't do anything.
00:16:04.250 --> 00:16:07.450
You had a very specific use case that it was very good at early
00:16:07.450 --> 00:16:34.300
on, yeah, which I think, and what we've seen over the last several years is it's broadening its ability to help people. And so for me, what happened, and the reason Ruh exists is because we had a lot of clients, right? Well, they're seeing us do tons of work using AI. And the question they always ask came to me with was, hey, do you see any applications for AI in my business? And these are people that have, you know, very different businesses than me.
00:16:31.230 --> 00:17:54.353
You know, like they're in banking, or they're in the supply chain business, or, you know, they're in they're in, like, over land transportation and. On, you know, they're in manufacturing. And so I started looking at other people's businesses, and we started building digital workforces for other people and and they were all custom builds, custom employees, where we'd go into their business, find levers that we could pull, and then build them digital employees that would help them, that would help them leverage AI in their business. And I did that enough on top of the foundation that I had spent building for the past several years that it made sense to separate the companies. And I had become obsessed with this idea of digital labor or or AI agents, or human emulators, or whatever people want to call them today, and and so I had, I had basically fallen completely out of love with Blockchain. You know it was, I mean, you could say I was having a an illicit affair with AI, and what had completely ignored the blockchain space for so many years that I couldn't even really have an intelligent conversation with people about the current events. So I thought, You know what, I should just step down as CEO from rapid innovation. We should divide these companies into two, I'll sit on the board over there, and we'll start another company dedicated to this new thing that I think is going to fundamentally change the world.
00:17:51.059 --> 00:18:00.997
And we have already tons of clients, because people are coming to us, and when I show them what it is that we built for ourselves, they want them.
00:18:01.052 --> 00:18:41.302
They want the same systems built for them. And so we, we, I, we split the company in half, started Ruh AI, and started taking this architecture that we had built mostly for ourselves and for our clients, and thought, how do we commercialize this and turn it into something that solves real problems for real people and and so we started building and organizing, really, and getting the foundation to a place where it was commercially viable, all this, you know, all the security and, and, you know, the the compliance stuff that you don't, you don't think about when you're building something for for your, for yourself, and and then when we launched, it was like, Okay, where are we going to go? What are we going to do?
00:18:41.358 --> 00:19:22.054
Oh, I have a background in construction. I spent the I spent the early part of my life in the first couple of companies that I built were in the construction industry. And one of the things that I knew about construction is that it's very, very it's very complex, like you have all of the work that I would have to do, as far as organization and administration and knowledge work, and then you also have to go put people on the ground to do things in the real world, right? It's almost like having, it's like having five or six different companies that you're trying to run at the same time. It's a lot of moving parts and and the software as a service world, or the SaaS world, had basically left all these people behind, right?
00:19:22.110 --> 00:19:35.172
Like, if you look at the software that makes construction easier today, you've got Procore, you've got blue color by Oracle, like, and that might be it like, there's not really much more than that. And so there's software that exists.
00:19:35.228 --> 00:20:33.677
Everyone in the industry hates the software. It's a necessary evil. It's slightly better than using spreadsheets and whiteboards. And so I thought, you know, why don't we go into the construction industry? And so I had a conversation with a guy, told him what we were doing, showed him a little a couple of things. He's like, Yeah, let's meet next week. I'll bring a couple more people. And this was earlier this year. And and on that call, when I showed back up, those 50 people on the call, because they're like, oh my gosh, maybe this guy has a solution to all the pain that we feel on a daily basis, and AI can actually start helping us, at least with some of our administrative, you know, the administrative over overwhelming administrative burden that sits on us at the level that they were doing projects and and then it just has turned into a niche, which is funny, because I got, I got into software development to get out of construction, and now I'm come full circle. It's bringing me back. They say, once you get in, you never get out.
00:20:33.731 --> 00:20:35.910
Maybe I it's like the Hotel California.
00:20:36.029 --> 00:20:46.069
But I love your I love your analysis of construction. I agree. It is several different companies in one, and it's way more complicated than meets the eye.
00:20:46.430 --> 00:21:07.450
I think society tends to downplay A, the value or B, the complexity of running an effective, efficient, quality construction company. There's lots of companies. There are a few very good, well run, profitable and efficient construction companies. Well, my listeners,
00:21:07.450 --> 00:21:25.170
none of them are profitable. Like, and when I say none of them are profitable, it's an it's insane to me to go find someone who runs a really successful, you know, like, let's say two, $3 billion a year in revenue. Yeah, like they're making they're happy when they walk away with 3% profit at the end of the year.
00:21:26.910 --> 00:21:32.910
Well, things are a little different in solar. You know, I cut my teeth working in the solar construction industry.
00:21:32.910 --> 00:21:37.350
I am a consultant to solar contractors. We call them EPCs.
00:21:37.650 --> 00:21:47.930
They're installers and EPCs, and some of them are residential and some. Of them are resi and commercial, and some of them are commercial only or utility only.
00:21:48.529 --> 00:21:53.569
Is there more profit margin in in the solar construction world?
00:21:53.809 --> 00:22:09.370
Well, certainly most installers are striving to achieve a 20% profit, okay? Are they actually achieving that is a whole other thing. Meanwhile, there's tremendous downward price pressure on the industry as a whole, because
00:22:09.370 --> 00:22:19.809
it's because of competition, right? Like, I'm assuming anyone like, the barrier to entry is low enough that you can get a lot of people in, and then there's a lot of people competing, and it drives the price down.
00:22:19.870 --> 00:22:28.410
Yeah. It's become very competitive. Yeah. So anyway, and it's a good thing that the price is coming down.
00:22:28.410 --> 00:22:34.289
Two things are happening in our industry. The cost of electricity is going up, which increases the value of solar.
00:22:34.470 --> 00:22:41.190
Any solar array produces electricity, so solar electricity becomes more value.
00:22:37.650 --> 00:23:56.930
As electricity becomes expensive and the price of technology is coming down. We're automating the fabrication of solar panels, racking, inverters, etc. We're in the process of on shoring a lot of that as well, which is counter a counter balance, because American manufacturing is not cheaper than manufacturing in Southeast Asia or China and but anyway, the price of technology coming down, we call it the cost adoption curve, and the price of electricity going up means that there's anti gravity for the industry. And the industry is growing despite headwinds from the federal government, which is hating on us right now. We're growing and we're growing in select markets. It is very geographic, geographically specific so, but Jesse, when you think about construction software, we have project management, we have estimating, there's all kinds of communication tools, ways to capture information on job sites, you know, pictures, progress, etc, bringing it all in. Give us some examples, though, of how you've leveraged this digital workforce and are building custom employees, digital employees, or digital I mean, once you build one, you can multiply it by a million for nothing,
00:23:57.529 --> 00:24:24.990
yeah, yeah. Once you build it, once, that's the, that's the, that is the chemistry to math problem that we're talking about, right? It's infinitely scalable. Once you take a biological problem with meaning, I have to hire people to do this, you turn it into a math problem, all of a sudden you can scale it infinitely. And that's the leverage that you have access to that I think very few people really understand, because we don't get exponential growth as humans. I think we just struggle to understand it.
00:24:21.850 --> 00:25:09.670
But the main things that we're doing is mostly administrative, obviously, right? Like the AI is not to a place where we're going to go out and being installing solar panels anytime, super soon, it'll get there. Like I've seen that. You know, people laugh at me when they say that, because they go look at the robots that do stuff, but it will get there. Like, I can see that we're making making progress quickly. The AI has to get better, and the actual then Robotics has to get better, and the AI's understanding of the physical world has to get better, because it has no model of the physical world and what it means to actually walk around somewhere and move something from point A to point B, or any of those kinds of things. But So the stuff we're doing right now, I would say, is, is all very either top of funnel, meaning it's lead generation and sales or administrative back office.
00:25:09.789 --> 00:25:18.970
The things that I've done so far that are the most exciting to me would be like, we had a we had a guy that owns a large company.
00:25:19.210 --> 00:27:20.049
They only do like, you know, billion plus projects and maybe, like 800 that million plus projects. And so when they go to do an estimate, it takes roughly 10 people three months to put an estimate together for them, give or take. And we built a system for them that can do that entire estimate in one night. And so you feed it all the information, and it does the estimate. And not only does, does it do the estimate, but it does a gap analysis and says, Hey, this is everything that I'm missing in order to put together a good estimate for you. And it sends those people that it has that would be normally doing the estimate out to go gather that information and to bring it back to the AI so that it can put together something really, really tight that they know is accurate, that isn't missing anything, to take a 10 person, three month process and turn it into a 12 hour process is like that was, that's that was super exciting in the manufacturing space, like we another fun project that we did was, this was very top of funnel. It was for incoming sales. We took a company that had 300 in incoming sales people, and we basically built a system that replaced all of those people. It took us about three weeks to build the agent, and they replaced all 300 of those people and saved themselves, like three and a half million dollars in in payroll and then the other stuff. Is like a lot of its lead generation, like, you know, imagine as a, as a, as a, as a contractor, if you had the ability to check, you know, every single permit pulled in, every single county that you work in, or every single request for, you know, every single request for new construction, every single like, someone goes and does a request for a big solar farm, it's like, well, you could know about it, and they may not have a contractor yet, because oftentimes those guys are talking to the county and trying to get stuff approved before it is the, you know, before they actually start contract contacting installers and all the and all those other people that they need, right?
00:27:20.049 --> 00:28:21.009
Well, installers could go and monitor every single county website, looking for every single, you know, permit request and all that stuff. But it would be so much work, like the buy it's there's a biological inefficiency problem that just says you can't hire 30 people to spend all day reading data on people's website. So you can go get an AI agent to do this stuff really, really easily, and then that gives you insight into who you should be calling and who's doing what. And so you can build relationships with them. And so it's a lot of those kinds of activities. And then on the financial side of things is, you know, for the larger companies, like, if you've got a company doing a billion dollars a year in revenue, you're dealing with around 10,000 invoices a month, and you have to make sure it will that, and let's say, 500 subcontractors. And so there's a lot of just administrative workload that comes along with doing that, with compliance, financial administration. You know, just juggling W nines is, you know, would be a nightmare.
00:28:21.309 --> 00:28:39.509
Well, you can plug an agentic system into that finance, that administrative back office that just does those things for you so you can focus on the work that you should be doing. You know, go out and build stuff, install solar panels, etc, etc, and for my vast majority of what we've been building
00:28:39.569 --> 00:29:23.370
for my regular listeners, check out episode 322, with Spark, AI, Anu, Segal and Julia Wu, they are doing what Jesse was just talking about with Spark. Ai, they've created a platform that goes out there and sources what's going on in all these different AHJs across the country. So no matter what jurisdiction you're working in, you have a central repository of signals about, are there friendly hJS? Are there good, you know, policies in place, or is there a ban on solar? Some counties do this periodically. They ban solar and wind. They get freaked out. So anyway,
00:29:23.430 --> 00:29:27.450
it's, is that real counties will ban solar and wind. This is new. I did not
00:29:27.450 --> 00:29:43.630
know that. Yeah, it's a real thing. It's a real pain in the butt. You know, nimbyism is, is a powerful force in the energy transition. People look at a solar farm and go, Wait a minute. We've got corn and beans today and tomorrow.
00:29:43.700 --> 00:29:52.212
You want to put this, you know, steel and glass structure on my farm. BTF, bro. And I'm like, You know what the enemy is?
00:29:52.283 --> 00:30:01.288
Actually urban sprawl. It's not renewable energy, because we're only going to convert one to 2% of the landscape to solar farms.
00:30:01.358 --> 00:32:05.950
That's all we need to completely green the grid. So it's really a it's a non issue, but they don't understand this. And it's not the landowner, it's the neighbors of the landowner that are objecting. The landowner can triple their income by leasing their property to a solar developer and but the neighbors go, Hey, I don't want to look at a solar farm. Sorry, I'm going to object to this. The Clean Power Hour is brought to you by CPS America, maker of North America's number one three phase string inverter with over 10 gigawatts shipped in the US. The CPS product lineup includes string inverters ranging from 25 kW to 350 kW, their flagship inverter, the CPS 350 KW is designed to work with solar plants ranging from two megawatts to two gigawatts. CPS is the world's most bankable inverter brand, and is America's number one choice for solar plants now offering solutions for commercial utility ESS and balance of system requirements go to Chintpowersystems.com or call 855-584-7168, to find out more. Back to, how do we automate construction operations? Okay, I get this question a lot. Tim, how are people using AI, I'm like, well, they're using it in all different ways, and you're the first person, really, that I've, I've come across that is building tools for construction, I guess you're the second person. And I think that, I think that bidding is estimating is extremely time consuming to do it accurately with with humans, and it is a job that is crazy. Clearly, you know, attackable with these, with these AI tools. What are the AI tools, though, that my listeners, who are company owners, okay, want to be aware of, is it? Is it off the shelf, and what is it? What is it? What is the lift financially, in time and money to take a process and automate it.
00:32:07.870 --> 00:32:10.809
So that is, that's a really, really good question.
00:32:10.809 --> 00:33:04.870
It's the question I think everyone should be asking, actually, and it's got a very simple answer. But in order to understand the answer, you first have to understand what is an AI agent. And right now, we have a problem in the nomenclature, where people say, I've built an agent, and then someone else says I built an agent, and they're two vastly different things. And so I put AI agents into three categories. Category one, I don't personally call an AI agent, but everyone else does, so I have to call them AI agents, and that would be automation 2.0 right? Like, we've been automating knowledge work for 20 years, 15 and all of that automation that you've been able to do. So this would be things like Nan workflows, even like, if, if you go and use MailChimp and send out, you know, a campaign to an email list. Like, that's an automation that was built times past.
00:33:01.370 --> 00:34:42.829
People used to sit down and send all those emails up. And so automation 2.0 comes along, and they've injected large language models into the workflow, and it makes the workflows more intelligent. And so you might call it Intelligent Automation, or automation 2.0 or something along those lines. But these workflows don't have any idea as to whether or not they have failed at a task or not. And so they start at point A they end at, you know, they're into the workflow. They have an input, they have an output. They don't care what they output, they don't care what they get on for their input, and they just work, right? And so that's, that's category one. People call those agents. I don't think they are agents, because they don't have agency, they don't know if they've failed. Bucket number two is what I would call human orchestrated AI. The vast majority, the vast majority of the agentic tools that are on the market today are human, orchestrated AI. My favorite is probably Claude co work and Claude code but these are things like Codex 5.4 like Manus. And what they are is their systems that are intelligent enough to really almost do anything that you need them to do. You know, like, for instance, my son today wanted he lost the buttstock for his AR, his airsoft rifle at an airsoft party. And so I asked an AI to go find me one for his gun that can be 3d printed, and to download it, open it in elegue slicer, set it, set up the print with, you know, the the filament that I've got on my 3d printer right now, and get it printing.
00:34:38.730 --> 00:35:00.170
And I it's actually done that while you and I were on this call. So it went out, found it, downloaded it, opened it up in yellow, go size layer on my computer, started printing it on my 3d printer like that's work I no longer have to do, but I had to. I had to orchestrate it right. I had to give it a task.
00:34:57.349 --> 00:35:53.692
I had to watch over it. I'd make sure it was doing the right thing. And you can basically do almost anything you can leverage AI in your business right now today, and it's very, very inexpensive. Like, if, if you go get caught code, for instance, or get a cloud subscription, you can pay $100 a month and basically run it 24 hours a day, seven days a week. And like, that's one of the things that I use to go find podcasts to be on. I use it to do i right now while I'm on this call, I have 123456789, agents that are doing work for me while I'm doing this podcast with you, and they're all just doing random stuff that I need to get done today. So one of them's doing some banking things for me and reconciling some accounts. And I've got another one over here that's doing some experimental lead generation.
00:35:53.749 --> 00:37:19.792
I've, you know, I've got a bunch of different agents that are just doing things for me, and those are not my non human orchestrated employees. Those are just my human orchestrated employees. I'm micromanaging them and telling them what to do. So quad code is probably the easiest place to start. There or cloud co work, I should say, if you're not a developer and they're both, they do the same thing, and then the next bucket that you have would be fully autonomous, AI agents. Now what I mean by this is, not only do they have agency, meaning I can tell them what to do and they can go do it, but they have autonomy, meaning that they can operate independent of me, because they have a memory and they understand who they are, where they work, why they work there, what their main objectives and goals are, and what self improvement looks like. They can modify their own code and. To become better at their job, and they can self reflect on the work that they're doing to understand whether or not what they're doing needs to change, and whether or not they're failing at things. And so like, for instance, I've got a if you go to ruh you'll see one of the first things we're selling, just as an off the shelf product, is a lead generation agent that you can hire, and you explain who it is, and you onboard them, just like you do a human you now work for me. This is what we sell. This is who we sell it to, and it goes out, and it does all the work of just selling and booking appointments on your calendar.
00:37:19.849 --> 00:38:23.970
And if somebody emails them back, it they it'll have a conversation back and forth until they set an appointment on your calendar, and so as long as you do a good job of explaining to them where they work, why they work there, what you sell and who you sell it to, then it's going to go out into the world, and it's going to do its best to go and sell those people and having a meeting with your Sales Team, and you never have to intervene. You never have to tell it to get better. You never have to tell it to start up again. You don't have to tell it to run this morning or in the afternoon, or any of that, because it works for you, like a fully like, like an actual employee. And so, like, I'm thoroughly interested in the third category right now. You really need developers to build those things and get there. But everyone should be playing with the human, orchestrated, agentic systems right now, because the the barrier to entry is almost nothing. For $17 a month, you can get started with Cloud co work, and it'll do $17 worth of work in the first hour that that you know that you actually figured out how to use it.
00:38:24.630 --> 00:38:36.510
So we only have a few minutes left. But to get from using co work, which is, you have to download Claude onto your desktop. To use CO work.
00:38:32.070 --> 00:38:52.970
First of all, you're not interacting through the browser anymore. It's not you're you're not chatting through most people know just go to claude.ai. This is a step beyond that and then, but to get to this fully agentic entity is a digital worker.
00:38:47.930 --> 00:39:05.770
Jesse, what does it take in time and material for a company to say, Okay, I want to build a digital sales force or a controller, a digital controller for my finance department.
00:39:07.330 --> 00:39:27.330
It really depends on what it is that you're building, like, I've got a complex agent that we're working on right now for somebody that is doing mergers and acquisition so it's going out and finding companies for him to buy, doing all the due diligence on those companies and starting the negotiation. And that's all this digital employee does is just for acquisition opportunities.
00:39:24.090 --> 00:39:42.150
That employee that like that process is is complex and legal and other things. So to build a digital employee like that is going to run you $75,000 probably, and then to run it on a monthly basis is gonna it'll cost you another $2,000 a month.
00:39:37.890 --> 00:40:15.310
So his first year investment will be 100 grand. And then after that, it'll fall to 24,000 but if you build something relatively straightforward and simple, like you want a controller that looks into your finances and does all that stuff, you're going to be spending between eight and$12,000 to build it, and between 507 $100 to run it on a monthly basis or less. Yeah. And so you know, if you've if you figure, I go high, if you go hire a controller for your finance team, it's going to be far more expensive to build
00:40:15.309 --> 00:40:24.509
something so in our last minute. Jesse, what happens to society when 90% of knowledge work is disrupted by the AI,
00:40:27.270 --> 00:40:30.390
you know what?
00:40:27.270 --> 00:41:05.290
It's a cheating answer. But I tell people I'm not here to solve that problem. I'm just making it so no one has to work, but somebody had better start solving the problem of figuring out what us humans do when there's no work left for us to do. I was having a conversation with a technologist who who's got her doctorate in in psychology and sociology, and one of the things that that she brought up is that, like human beings, have to have purpose, or we like that we commit suicide, is what ends up happening. We just don't do well when we have no purpose. And one of the one of the ways that we get purpose in our lives is having something to do, IE or
00:41:05.290 --> 00:41:23.370
hobbies or creative outlets or sports. I mean, there's a lot that you think of humans that retire from working right they're doing something. Some of them are sitting on the couch, but some of them are volunteering and giving back to society and creating value that way.
00:41:23.910 --> 00:41:45.410
I think that's what happens. I mean, I think at the end of the day, AI provides value, the value gets distributed to the people, the people live the lives they want to live. I think that, I think that work will become optional in the future. If you want to work, you can, and you'll work alongside, alongside AIs. If you don't want to work, you won't have to, and you'll. And you'll, you'll make money.
00:41:45.410 --> 00:42:01.490
Or, like, my question is, where does that paycheck come from? For some people, like, I could see it in my own business, fine, if I have an army of digital consultants that are doing the work that I used to do, that there's an income stream there. But, yeah, I don't you know if the answer
00:42:01.490 --> 00:42:56.990
to that question, it's terrifying to me, actually, you know what? I think, like, people are worried about it, the AI, you know, doing something and taking over the world. And, you know, like, we've got Skynet and robots and we're all enslaved. Like, that's not the future that scares me. The future that scares me is when, you know, knowledge work makes up about $70 trillion a year globally is paid out to people who do knowledge work, roughly between 50 and 70 trillion right now, the estimates say that AI can do at least what can do 100% of 70% of that meaning no human in the loop, no human intervention, no Human needed at all. If you look at the unemployment rate that would happen if you, if you get rid of, you know, 3040, $50, trillion worth of work globally on an annual basis, you have an unemployment rate that's insane.
00:42:56.990 --> 00:43:06.250
So it's that transition that scares me between where we are today and getting to that place where, where the AI is doing stuff, and honestly, I don't have the answer to those things.
00:43:06.250 --> 00:43:31.230
What I tell people is, you ought to learn AI now. If you don't know how to use it, you need to learn to use it today, because the people who are going to get more screwed over in that process are the people who don't know how to use it. But all of us are going to get screwed over in some way over the course of the next decade, if the right people don't come in and solve the problem of, what do we do with all these humans who are no longer valuable to us because of what they can do?
00:43:32.250 --> 00:44:40.218
Yeah, check out mogadot. M, O is his first name, G, A, W, D, A, T. For my listeners, mogadod has a great YouTube channel, and he thinks very deeply about this. He's a technologist, he's an AI entrepreneur, and he thinks that we should lean into our humanity. And that's really my only answer, is we have to lean into our humanity. That's something that the AI can't can never be is fully human. It can do a lot of knowledge work, but it can never be fully human, and unfortunately, we have to leave it there. Jesse, I want to thank you for coming on the show today. Jesse Anglen, check him out@ruh.ai and please check out all of our content at cleanpowerhour.com. The best thing you can do to help others find this content is to just tell a friend or two about the show and do it today. There are so many people who do not know about this library of content, which is super valuable to any energy professional. Hey guys, are you a residential solar installer doing light commercial but wanting to scale into large C&I solar? I'm Tim Montague.
00:44:36.902 --> 00:45:19.175
I've developed over 150 megawatts of commercial solar, and I've solved the problem that you're having. You don't know what tools and technologies you need in order to successfully close 100 KW to megawatt scale projects. I've developed a commercial solar accelerator to help installers exactly like you just go to cleanpowerhour.com click on strategy and book a call today. It's totally free with no obligation. Thanks for being a listener. I really appreciate you listening to the pod, and I'm Tim Montague, let's grow solar and storage. Go to clean power hour and click strategy today. Thanks so much.
00:45:15.222 --> 00:45:22.810
And with that, Jesse, is there another way you would like to connect with our listeners
00:45:23.710 --> 00:45:44.210
now they can go to ruh.ai r, u, h.ai, if they got questions about what I do, or you can look me up on on LinkedIn or Twitter if you just want to reach out and have a conversation. Yeah, happy to talk with anyone who wants to know more about, more about how AI is going to affect their business and what they can do about it.
00:45:44.210 --> 00:45:46.550
I'm Tim Montague, let's grow solar and storage.
00:45:46.550 --> 00:45:49.190
Thank you so much. Jesse, thanks. Tim.
00:00:50.780 --> 00:01:19.900
00:01:21.340 --> 00:01:53.960
00:01:55.460 --> 00:02:25.632
00:02:21.980 --> 00:02:29.860
00:02:29.860 --> 00:02:33.599
00:02:33.599 --> 00:02:36.900
00:02:33.599 --> 00:02:37.740
00:02:39.000 --> 00:03:17.680
00:03:18.580 --> 00:03:55.099
00:03:56.180 --> 00:04:11.500
00:04:08.900 --> 00:05:00.439
00:04:56.180 --> 00:05:20.500
00:05:21.100 --> 00:06:21.940
00:06:19.179 --> 00:06:56.179
00:06:56.779 --> 00:06:59.840
00:07:00.259 --> 00:07:13.480
00:07:10.340 --> 00:07:32.700
00:07:32.820 --> 00:08:49.860
00:08:46.379 --> 00:09:23.439
00:09:23.980 --> 00:09:39.539
00:09:39.539 --> 00:09:48.839
00:09:49.860 --> 00:10:10.159
00:10:10.159 --> 00:10:29.559
00:10:27.339 --> 00:11:06.980
00:11:04.519 --> 00:11:14.980
00:11:14.980 --> 00:13:48.649
00:13:48.829 --> 00:13:57.710
00:13:59.569 --> 00:14:17.049
00:14:17.589 --> 00:14:23.549
00:14:23.549 --> 00:14:45.889
00:14:45.950 --> 00:14:54.413
00:14:49.769 --> 00:15:35.309
00:15:35.309 --> 00:16:04.250
00:16:04.250 --> 00:16:07.450
00:16:07.450 --> 00:16:34.300
00:16:31.230 --> 00:17:54.353
00:17:51.059 --> 00:18:00.997
00:18:01.052 --> 00:18:41.302
00:18:41.358 --> 00:19:22.054
00:19:22.110 --> 00:19:35.172
00:19:35.228 --> 00:20:33.677
00:20:33.731 --> 00:20:35.910
00:20:36.029 --> 00:20:46.069
00:20:46.430 --> 00:21:07.450
00:21:07.450 --> 00:21:25.170
00:21:26.910 --> 00:21:32.910
00:21:32.910 --> 00:21:37.350
00:21:37.650 --> 00:21:47.930
00:21:48.529 --> 00:21:53.569
00:21:53.809 --> 00:22:09.370
00:22:09.370 --> 00:22:19.809
00:22:19.870 --> 00:22:28.410
00:22:28.410 --> 00:22:34.289
00:22:34.470 --> 00:22:41.190
00:22:37.650 --> 00:23:56.930
00:23:57.529 --> 00:24:24.990
00:24:21.850 --> 00:25:09.670
00:25:09.789 --> 00:25:18.970
00:25:19.210 --> 00:27:20.049
00:27:20.049 --> 00:28:21.009
00:28:21.309 --> 00:28:39.509
00:28:39.569 --> 00:29:23.370
00:29:23.430 --> 00:29:27.450
00:29:27.450 --> 00:29:43.630
00:29:43.700 --> 00:29:52.212
00:29:52.283 --> 00:30:01.288
00:30:01.358 --> 00:32:05.950
00:32:07.870 --> 00:32:10.809
00:32:10.809 --> 00:33:04.870
00:33:01.370 --> 00:34:42.829
00:34:38.730 --> 00:35:00.170
00:34:57.349 --> 00:35:53.692
00:35:53.749 --> 00:37:19.792
00:37:19.849 --> 00:38:23.970
00:38:24.630 --> 00:38:36.510
00:38:32.070 --> 00:38:52.970
00:38:47.930 --> 00:39:05.770
00:39:07.330 --> 00:39:27.330
00:39:24.090 --> 00:39:42.150
00:39:37.890 --> 00:40:15.310
00:40:15.309 --> 00:40:24.509
00:40:27.270 --> 00:40:30.390
00:40:27.270 --> 00:41:05.290
00:41:05.290 --> 00:41:23.370
00:41:23.910 --> 00:41:45.410
00:41:45.410 --> 00:42:01.490
00:42:01.490 --> 00:42:56.990
00:42:56.990 --> 00:43:06.250
00:43:06.250 --> 00:43:31.230
00:43:32.250 --> 00:44:40.218
00:44:36.902 --> 00:45:19.175
00:45:15.222 --> 00:45:22.810
00:45:23.710 --> 00:45:44.210
00:45:44.210 --> 00:45:46.550
00:45:46.550 --> 00:45:49.190
