Morning. Good morning. Welcome to LCP Delta's webinar on generative AI. You should currently be able to hear some music. Probably sounds like you're in an elevator. If you can't hear the music, you need to check your settings because you should be able to hear it. We'll start the webinar in a minute. I can see that some people are still joining us, and so I'd like to wait a minute. But I'm not going to wait too long, right? We've got quite a fascinating topic to cover. And not if people are late. They can always watch it. Well, they'll just have to miss out a little bit. But please just be patient with us while we let one or two stragglers join. While you're waiting, have a look at the various resources on the desktop. There's lots of buttons to click. Perhaps a key one to mention is you can ask questions at any time. Now, just to be clear, I'll handle questions. We won't respond straight away in the Q&A unless we, you know, unless it's really obvious, but I can't promise to do that. But I will take your questions and ask them of our experts who I'll come to in a moment. Worst case scenario is you ask a question, you don't, we don't respond in the session because we don't have time. We promise to come back to all questioners offline as a basic minimum with an e-mail response. So so you will get an answer to your question, it just might not be immediate. You'll either get an answer during this session or you'll get an answer offline by e-mail. There are subtitles if if you want to use them. It says here. Click CC in the media player to enable caption captions and blue means it's on. So captions alive. Accuracy may vary, which sounds a bit alarming but but basically the captioning is created in real time. So I guess every now and again it will make a small mistake. But please use that facility if you find it helpful. OK. I think that's all the housekeeping out of the way, so. Just. Get to our introductions while we're still waiting for the last few people. So there's three of us here. So I'm going to introduce myself and then I'll ask my friends Bram and Steve to introduce themselves. My name is Nigel Timperley, I work for LCP Delta and I run 2 Research Services. This webinar is produced by a service called Energy Insights Plus and it's purpose is to help you engage, learn how to engage with B to C customers better. Generative AI, we believe could be part of that that dialogue. So that's why we're here today to find out more about generative AI. The people are going to help us find out are Bram and Steve. So, Bram, would you like to say hello? Yeah, Hello. Good morning from me from the Netherlands. My name is Bram. As mentioned working for Net Grid, I'm a director of products mainly focused on product marketing and product strategy at Net Grid. And our main product we will introduce in a in a in a second. Nice to meet you all and looking forward to this session and. Steve. Good morning. Good afternoon everyone. My name is Steve Dawson. I'm the founder and CEO of Power Connect AI, started the company end of October 2023. We're based out of Tampa, FL and we offer AI powered customer service solution for the energy and utility industry. That's brilliant. Thanks guys. Bit of sympathy here for Steve because where he is, it's stupid o'clock. He's in the States at the moment. So thanks for joining us both. OK, so I think most people will have will have joined. Quick reminder, you can ask questions. If you just joined, you can ask questions. We'll either answer them in this session or offline. I want to begin with a poll if I may. And I want to ask you what you, the audience think. You know, we have an audience here of energy industry professionals and so be very interested in in testing the water, OK, to get a sense of where the industry feels we are with generative AI. So the question is, we will see generative AI transform retail energy customer services in the next three years. And I haven't allowed you a middle option. So please click ABC or D to to give me your sense of where we see generative AI transform, whether we will see generative AI transform that customer conversation that happens in retail, energy, customer service departments across Europe right now. Will it be a fringe? Will it be transformational? Where's your sense of where we're going in the next three years? No, we're not talking 10 years here. We're talking, you know, this side of the horizon, right? If you hover over the screen, your cursor, you should be able to select an answer. I don't quite know how that works from your end, but please, please click on it. I can see the answers coming in now. I'll just give you a moment more. There's some interesting responses coming. Still some people joining us. If you're just joining us, where do you think generative AI will be in retail energy, customer services the next three years? I've just got to give you a few more seconds. I'll count you down if you like. You've got 54321. OK, we're done here. We're done. So let's go look at the answers. And what we've actually got is a very positive response. I would say over half of you or let's say half of you think retail energy customer services is very likely to be transformed by generative AI, which is really interesting. And perhaps most interesting of all is very unlikely scores 0%. So out of this large audience, nobody has said it won't have, it's very unlikely to have no transmission impact. So, so thanks for that. We'll hold that thought. We've got another poll for you later on and I'll be interested to see your responses there. So stick around and you'll get a second chance to give us your expertise in the light of what you're going to hear today. OK, so let's dive right in. Thanks for that. So 3 bits of this, the role of generative AI in customer service. I want to paint A5 to 10 minute picture. That's all just some highlights. Why should you care, right? Why? Why does this matter? Will it matter? Is this all hype? Where are we? Is it mature? Is it not ready? All that stuff. Then I want to get into the meet with these guys. These guys are actually doing it. I'm just talking about it. They're the experts. They're the ones who are actually, they've rolled their sleeves up. They're, you know, at the cold face doing the hard yards. How many cliches do you want? And so I'm going to I'm going to talk to these guys and they're going to tell you all about it. And finally, we'll have a set Q&A session where you can ask your questions. But let's say if we don't get round to your questions, we will come back to you offline. OK, so I want to tell you a story. The story is a glimpse of the future picture. This Emma is an energy customer and she's she's this is 2030, right? This isn't this isn't 2025 S Emma has got a heat pump and an EV. She's pretty smart. She's she's she's going with the flow now. But what she wants to know is what does that mean for her energy use, right? We have seen an explosion in the development of Smart Terrace in the last two years. It's all very confusing. OK, so Emma's trying to understand what that means for her. So she contacts her energy company. She starts asking difficult questions, right? What's the best tariff for me, right? I've got this heat pump. I've got this EV. What, what what's best, right? You know, because if I, if I get an overnight tariff, does that screw around with my heat pump because it's good for my EV? But you know, or maybe if I get one of these time of use things, maybe it's good for the device. But actually overall I'm losing because my general consumption is now priced higher to compensate. So Jenny and I will engage with Emma and have a nuanced conversation based on her, her particular circumstances here. It's saying, OK, so you drive 250 miles a week, your heat pump usage spikes in winter, taking account of regional weather outlook heat, an EV saver tariff could save you €232.00. Do you want me to break that down? You want to show what that looks like? She goes, hang on a minute, hang on. I, I'm, you know, what if I changed my working practices so I only work two days a week, you know, does that change the profile, right? Just because I did this last year, I might do that next year. What's the story? And it goes, OK, well, then I take that on board. Even if your heating usage increased by 20%, you'll still save €140.00. Do you want to see a plan that breaks that down? What we're having here is a really rich, nuanced conversation with an advanced energy user. And we're having that with a glorified bot, right? Quite interesting situation. You cannot do this today, our thesis as you will do it tomorrow. OK, so the thesis becomes that Jenny and I becomes more invaluable tomorrow than it is today. And by tomorrow, I'm talking three years, five years, maybe, maybe one year, right? Who knows? But, but, but why is that? Well, customer needs are evolving. They're changing because they're, they're no longer just passive bill payers. These guys are starting to engage with the transition. They're starting to get solar. They're starting to get EVs, they're starting to get batteries, they're starting to get heat pumps. They're starting to say, how do I integrate all these things so I can be in control of my energy use? They're trying to do simple things. I get energy insights apps that that give them insight over there of their usage. They can deal with energy waste. So you start to go through this journey, what we call it execute, engage, empower and collaborate. Execute isn't even on this picture. That's the basic operational integrity. Get the bills right. You've got to do that first as an energy company before you do anything else. Then you start engaging around my consumption. You start to show me one where my what my usage looks like. Maybe you disintegrate my demand and help me get a handle on where the real hotspots are. And then you empower me with new energy assets that allow me to take back control of my usage. And finally, we in the like Emma is at the collaborate stage. She's she's really becoming a full node on the on the smart grid. And the key idea here is that as the customer advances on their transition journey, the need for that nuanced dialogue that Emma was having increases. And that's where Jen AI comes in, in this more, in this richer conversation. You need a more powerful engine. And we believe that it's you should at least be looking at the possibility that Jen AI could be that engine. OK. And now at this point, somebody usually says, hang on, what about hallucinations? Right. And and that's a very good question because the one thing that Jenny is notorious for is making stuff up. And and one of the reasons it does this is because it doesn't actually do well with the idea of truth or falsehood Is the pattern generator effectively, and if it doesn't have something to go on, it will make something up to fill in the gaps. However, we have a fix for this called retrieval augmented generation. And retrieval augmented generation basically combines AI with real data to deliver to ground it in your corporate data, your account data, your corporate guidelines and policies. So coming back to an Emma's question, what's the best tariff for my EV and my heat, heat heat pump? This thing in in with RAG can retrieve specific date data about Emma, She can get her account data, her tariff data. It can get your corporate guard rails that say, well, you can talk about this, but you can't talk about that. Here's our product set and from and that limits the risk of hallucinations by grounding it in reality. And so it's able to generate a meaningful response that's actually relevant and meaningful for Emma. And so the conversation goes and you get into this innovative iterative loop. So RAG is a very important idea to ensure we get tailored, accurate responses. Now there are a whole bunch of claimed uses for Jenny. I, I'm not going to get into more right now. I'll put them on the screen. I'll give you some examples. I guess what I'm trying to say here is it can do a lot and a lot of people are experimenting with it. But at this stage, you know, there's one here, 2 there and so on. We're starting to to have those early implementations and and we're starting to see where it can go. But for today, I want to talk about this one. OK, these guys are with me today. And so I introduced to you Bram from Net to Grid and Steve from Power Connect and they're helping call centers with energy insights and that they're providing scripted rag based support for agents. And they're going the extra mile beyond that, which is to move towards replacing legacy chat bots with Gen. AI power chat bots ratio completely different beast. So what is this solution? Well, this point I'm going to ask them, but I'll just I'll just throw out one slide to put a bit of a frame around this. Essentially what you have is net to grids, energy AI architecture in the background, which is is doing a lot of things that really advanced classical AI does around energy measurements, disaggregation, insights, all of those good things that we've started to see normalize in the industry. And then we're overlay, they're overlaying that with Jen AI from power connect AI to to to power that customer conversation with agents and chat bots. OK, So I'd like to come now to Bram and Steve. So Oh well, there we go. Let's start with you, Bram. Do you want to tell us a little bit, briefly introduce net to grid and explain what you've been doing in this space? Yeah, sure. Thanks Nigel. So maybe you could go to the to the next slide if you want. So net to grid is a energy insight service provider. We, we basically run analytics as a as a service for energy retailers and utilities and we do that across the globe. So you see some some logos here in Europe and the North American markets that we serve. So we are AB2B service provider. We don't deliver energy insights as a as an app or as a business to consumer products. And we do this already for quite some time and we work with partners mostly to deliver that service to end users. So in this constellation of the energy inset plus service, we will also focus on the on the domain of energy inside services. And what, what you can see also on the on the next slide is how that works. So what we basically do is analyze smart Metre data that's let's say the 15 or 30 minutes data or 60 minute data coming from smart meters. We have our algorithms being able to determine in the total energy consumption what households, appliances and products are being used every day. We we make a so-called breakdown of consumption for appliance or appliance categories or events in the home and report out on that so that the energy retailers can consume that in their own digital channels, be it a portal, a mobile app or even setting a marketing campaign. So it's more than just energy insights today that's basically used across different disciplines within the energy retailer. So we we call that energy disaggregation. That's the technology used and rerun that in the clouds as well as in the North American markets and that it's inside the next Gen. smart meters as real time disaggregation is done is done possible. And that that's about the introduction of the the service. And here you see some examples how that looks like in a customer engagement setting. So when you start from the left, the the service delivers not only the the breakdown, but also comparison data, benchmarking data and and also has some intelligence to really drive that conversation towards end users. So also a customer experience as a service is part of that service. And besides the historic energy insights, we're also giving insights on how much of your solar production was being self consumed and also do predictions of what households would benefit from certain products. For instance, if solar households have a certain amount of solar production and self consumption, what would be the ideal size of of a home battery for that households? And these kind of upsell or propensity models is also something we deliver to our clients. So they can make segmentations of users and do upsell and cross sell of new products and services. That's, I think the introduction as far as Metagrid goes. OK, thanks. Thanks, Bram. Steve, could you sing similar briefly introduce power connect so, so we understand how you're playing in the in this space? Yeah, absolutely. So, so I was really excited, Nigel, when you invited me to the to the webinar today, because that's kind of like how power connecting I started was really based off my own experience I had with my utility because I started my electrification journey in 2023. And when I contacted my utility, I was just like, hey, purchase the EV. What kind of EV rate plans do you have? And unfortunately, the customer service representative didn't know. Long story short, it took me an hour and 30 minutes. I went through eight different agents and nobody was able to help me. The chatbot was outdated on the website. And I was kind of like, I was a little bit frustrated. I was just like, wow, you know, And in the back of my mind I was just like, if we could like develop some sort of AI widget, right, that they can just kind of plug into their customer service Organisation. So customers like me when they do call asking what kind of solar programs, rebates, EV time and use rates, right agents, the first agent can be able to help the customer immediately. And that's really how Power Connect AI Star was really based off my own experience with my utility, that's sometimes how it starts with startup companies. So our mission is to really revolutionize customer experience using generative AI. And what we're really doing is we're bridging the gap between the service provider and the customer and helping them across multiple channels. And one thing we do very well is we have the industry expertise and the knowledge and, you know, just the use the, the best practices that we've already done that really represents the future. We're not a big company, we're just 10 employees, just myself and, and 9 developers, but let me tell you some of the smartest people. But within the year, we were able to already bring 10 utilities and even more onto our platform that they can really start using and testing. One thing about Power Connect AI, instead of claiming we can do everything, which we don't want to do, we're like, hey, we're going to just work with the best vendors out there. People who vendors like Meta Grid who can do energy AI sites for the disaggregation companies like SAP and Oracle who dominate the North American market with the CIS, the Metre to cash, we're going to just work with them. So we're this kind of like open API platform where we can basically plug, plug and play and bring all the information together to really support the customer service agents. If you can, go to the next slide. Sure. SO11 particular customer that I would like to talk about today in this webinar is Hillsborough County. They are the first government owned utility here in the state of Florida who's using Power AI, which is Power Connect AI. They're the third largest utility. They have around 800 employees, 40 customer service representatives. And exactly a year ago, we had the first kind of initial discussion with them where we kind of just showed them the possibilities, right? It's just kind of giving them some ideas. They like what they saw and we were actually moving very quick, which is somewhat unusual with the utility. But then cybersecurity came, right? It's AI, cybersecurity, they kind of put a stop to it. But what we kind of showed them was we're going to take this in a phase approach. We're going to start doing like the data readiness, looking your content, looking at the information that we can use to just give, just allowing you to kind of test AI maybe just for training purposes, enabling agents, right, without pulling various customer information into our platform. So that took about three to four months. And then we became a test phase last year around September till December and this year January because they, they, they spent a great amount of time during the test phase. January we kicked it off and now they're they're actually using it in production. So you know, Hillsborough, they're very well known utility here. And you know, I'll talk about a little bit about the the improvements and the success that we've already seen in this short amount of time. But if you want to go to the next slide, Nigel. I think that's my only slide actually on this. So we're. OK, that's all I want. All right. OK, that's fine. No, that's very useful. Thanks. SO as, as I understand it, you're collaborating with Net Two Grid in Hillsborough and so, so how do the two services work together? Great question, Nigel. So one thing that net two grid does very well, and I'll let Bram elaborate a little bit more, but one thing they do really well is they provide the insights, right? The consumption, how the how the customer is using energy, water, gas. And with the insights that they provide, we can take that data and feed it into our platform to really show like the visualization, how they're using it, how they're using among the peers. So a agent who is not really familiar like on the whole consumption, we can provide better insights on what's really going on with the customer. But not only that, we can actually also provide personal recommendations where the customer can eventually save energy, gas, water and provide really recommendations to the customer on the fly, either through chat, through, you know, speaking to the customer and so on. And Bram, if you want to elaborate and add anything more to it, you're more than welcome to. I think it's a good summary here, Steve, and maybe maybe one thing to add is that this this works indeed already out-of-the-box, right? So you need just smart Metre data and then the the engine starts crunching. So there's not even requirements to have end users all down a download a mobile app for instance, as as long as there's access to metering data and then this service is operational as such. OK, so. Would you say then the smart meta data is actually the central element? But that's the bit you got to get right. Exactly. So to generate those insights you need access to to the smart Metre data. Of course, in Europe you also need consent from end users to be allowed to process that smart Metre data with the the analytics that that we do. But besides this, the, the, the serving out of those insights can go to different channels. And in this case, our output is delivered in in the rack of, of, of power connects. So that's basically one of the sources they use like a knowledge base is also another source and that you feed into the large language models that power connect and really fuels that the agent cockpit with. OK. So, OK, that's really interesting. So coming back to the Gen. AI piece around, you know, there's a lot of concern around hallucinations and around kind of honesty and trying to control the quality of those responses, you know, give, I mean, not just at Hillsborough County, but generally, you know, you know, had one question here. What did energy companies generative AI be be honest enough to recommend another company's tariff if it would benefit the customer, right? So how do you you know, how do you know what this thing is going to do, right? Yeah. And that's where the the, but I mentioned the, the knowledge base come into play. So as a, as utility, if you're implementing such a feature or such a, such a product, you can set the guardrails. So you can set the sources that can be trained basically for the agent to respond. So an agent, I mean also the the rack. So that's the, the really programmable solution that then really also delivers the output to the different applications. And that's quite powerful because that really gives the, the energy retailer control of what are the kind of things that are allowed to be said. Like do I put my frequently asked questions list as a knowledge base item? Do I put my tariff plans as a resource for that, for that Gen. AI agent? So you can really set the domain of where it is allowed to look into to give responses back to to end users or call Centre agents. OK. So, so it is a constant theme. I get what, you know, in terms of our research, when we we've been researching this, we found a lot of fear about, you know, this kind of idea, these things, this kind of runaway train, right with a mind of its own. Now we've got the word guard rails on there. Is that essentially what you're talking about is like a literally a, you know, a walled garden of well, you could talk about this, but it's got to be in the walled garden. Exactly, Nigel, and that's something we've so when we started at Power Connect guy, when we started, we saw a lot of hallucination. We were hallucinating all the time. We were pulling various information wrong telephone numbers from other utilities websites. But here's the thing, what we've what we found and by putting these proper guardrails in the framework in place and only using really in your own internal content and not pulling from like public facing information from like other websites sources. If you can do that right, then the halloos, then you can really minimize the hallucination. And really the, the, the the key goal is allowing yourself enough time for like testing it right. What we found was in All in all testing phases, how much old, outdated content and data exists within these platforms. So it was actually a good exercise to take a step back and make sure we don't have garbage in garbage out. But we are actually have, you know, good solid content and information before we give it out either to the agent or to the customer. Because one thing you don't want is you don't want to upset or frustrate your customers even more by giving out false, outdated information, right? And needs to be. It needs to be updated. And that's why the test phase is, is so important, right? And you have to have, you have to have certain executive sponsorships within, you know, making sure that if there's some, if we do identify something that we can notify the person so we can update that at least on our end. So just wanted to share that with you. Yeah, no data quality. It's an interesting take out and it's something it comes back to our model around execute engage, but you've got to get execution operation integrity's is absolutely key to building that trust. If I just jumped it back to Hillsborough a little bit, let's let's just bring this to life. So from an end user human in the loop, which is the agent am I, is it just essentially I've got more empowered human to talk to or, or can I see this at Hillsborough right now through an app or web screen or something? Where, where? Where were we at with that? So right now, so right now with Hillsborough County, they're using our AIH and assist. It's basically a very lean single one user interface where we've integrated to their Oracle CIS system which they're currently using. So really if you call Hillsborough, we can pull up the information instantly from Oracle with all the relative, you know, information that the agent needs to see. And if you say, hey, you know what? I have AI have a question about my usage instantly when you say that that's where we bring in net to grid, because it's all driven by AP is and that's that's the nice thing with new modern technology, we can provide those real time insights to the agent. And based off of that, you know, we can recommend the agent. You know, maybe this customer needs, maybe we need to offer them some sort of program, right, or maybe they have a water leakage even right. I mean, it's these kind of things where we can, where we can help the agent immediately on the phone. Right, so the moment sounds like it's about giving, you know, more power to the Asians elbow. You know, the agents got all these new insights is the next stage to move to a chat bot because it strikes that's riskier. That's that doesn't seem quite right. Whereas at least with the agent I've got that extra screening. Is that how you see it as a natural phasing? Correct. Yeah, exactly. So you were you were kind of breaking off, but I think I understood your question. So for like no problem. So I think so for the next phase, what we're going to do is exactly that we have a sole service portal where the customers can log in. We have a a digital AI bot. So same thing, same idea. Because if you get it right once you can replicate and reuse that same kind of information to the customer. So if the customer types in, hey, I want to know a little bit about my usage instantly same, same idea. We would, you know, we would provide recommendations based off, you know, the use the information we're retrieving from net to grid to assist the customer. We're actually taking it even a step further where I feel like chat bots at some point are just going to disappear. We're actually working on something right now, which is a digital AI avatar. So instead of typing in it, you're going to just be speaking to someone. Yeah, I know. But that's, I think that's really where it's heading to, right? Because that's where we're, we're really enhancing the user experience on how we can communicate to the end customer. And if we have that knowledge and the data, why not be able to have a normal conversation like I'm having with you, but with a digital AI avatar? It's. I show you I'm a human being. I'm not an avatar. No, you're a human being. I know that. I can't prove it though, right? I mean, just just building on that, Bram, you know, you've got a lot of experience in not just doing the AI, but also presenting this through insights apps. You know what, what's the next stage for the app itself? Is it, is it a dialogue that looks something like a ChatGPT dialogue? You know where I'm asking questions just like Emma. Yeah, I I think that's indeed the starting point we see already happening today. So indeed it's, it's really about this conversation, having basically having a conversation with your home, right. So to end user, you just talk, talk to their home assistant, let's call it like this. And you get not only the insights but also recommendations. What you see happening right now is that that same information is also being used in let's say your last customer journey part, this collaboration phase where you can even start giving commands to be followed up. So if you have some home automation in place that you can just ask your home energy assistance to optimise energy for you. And that is something that is actually currently being developed right now with a few clients of ours. So it goes beyond just. This sounds really interesting. Go on. Can you give me an example of that? Now in in the case of households that have for instance solar energy and producing energy at times of the day where you're basically not optimally self consuming that solar energy and the household member can ask for recommendations on how to improve the self consumption of solar energy. But if you have some products in the house that are connected, that assistance will will guide you in basically configuring your home system scheduling. So there you can for instance change your EV charger settings accommodating to self consumption of your solar energy. Gosh, how extraordinary. And and that's actually something that that is that is quite new in the in the atmosphere of let's say using Gen. AI to help you optimise this. But the controls are still with the end user. So it's not that this assistant will take over. It is assisting you in scheduling or optimizing that that configuration. Gosh. So OK, I think if you had to pick an industry, if I said to you, you know, I'm not going to say this, but if I did say give me an example of a really wildly innovative industry, the energy industry probably isn't the first one that comes to mind. So what you're describing sounds positively space age to me, right? My guess is there's going to be a confidence building phase while people get their heads around this, right? You know, we're sitting here talking about an extraordinary change where an end user can have a conversation either through with an agent or through some sort of dialogue, maybe even a spoken conversation or an avatar. Where do you start with this journey? It sounds so extraordinary if you if you were advising a retail energy provider right now and a lot of them are on the line and they've yet to adopt Jenny I. What's the one piece of advice? Well, ask each of you in turn. Brown What's the one piece of advice you'd give an retail energy provider about where to start? Yeah, I would definitely start with use cases that involve personalized information. So if it's, for instance, with the service we provide, if you provide personal insights on energy consumption, that's a great way to get also people either hooked to the service or to build, build trust that the that the outcomes of this service are really relevant for that household. And there we also indeed are very proud of the quality we bring into those agents. So if well, what Steve said, crap in crap out. If the source of data is not good, that's really important. The sources need to be of extremely good quality for these Gen. AI agents to be also delivering the quality service. So for us, it's really about start with good quality source data and also look for the the quick wins. So start small, start experimenting, but start experimenting also today in in the in the field. So don't build everything and hope that in one year we have something ready, but go quick and, and go fast and, and feel fast. And I think that's our clients, these experiments. And I think Steve has a lot of examples there. It can already be super helpful to have a small agent to do something great. Thanks. Steve, how would you answer my question? Where? What's the one piece of ice you'd give a retail energy company about where to start right now? Absolutely. Great question. I think actually too, what I could think is if you have good training content that your agents are already using to assist customers like me who call in and you know, asking what kind of programs rebates you offer. Maybe offer an internal knowledge base that they can use as like a digital AI training assist to assist customers on the phone. Another one that you could probably look at is scrape or crawl the content that you have from your public facing website today and just use a external web chat, AAI web chat using generative AI. You can use it with AWS bedrock. That's there's great because they provide a lot of large language models, test it and that's something that you can just kind of plug and play on to your content management system and assist, you know your customers 24/7 and any kind of language, French, German, Spanish, you know, especially here in the States where we have a lot of languages. And that's something you can kind of use right away, just reducing the call volume and assisting your customers. And then once you have that down, then take it into a phase approach, integrate it to your back end system, your CIS system, adding more automation along the line. Those are the kind of things that I would recommend energy retailer to to look at. That's great. You did mention one thing there in passing, which I also mentioned earlier on, which is this multilingual capability. Is that something come straight out-of-the-box with the LLM or do I have to do something clever? I mean, you know, we have a lot of languages in Europe, I tell you. And so, you know, it's a real big deal this and that can be. Well, tell me more about that because it sounds, you know, a bit of a free lunch to be honest. It is actually a free lunch. Yeah, it's one of the, it's one of the. That's why I, that's why we really like working with AWS because it's just, it's a language package that we just activated. You have to do the testing and make sure it's like, it's the output, you know, because we did it with conversational AI. When I I spoke something in Spanish and it kind of sounded a little bit too much gringo for me where you maybe have to fine tune. But here's the thing, you have to, you just have to test it out. But it's, it's one of the things that AWS offers. It's a language package that you can activate and you can start testing. OK, OK. And that's something that is is actually intrinsic to the technology. It's not something that's built in at the application layer. It's further down. Right. It's kind of. Yeah, OK, Wow. OK. This, this, this is all amazing stuff. I guess moving forward, the bit that is most contentious I find when I talk to energy companies, it's the chat bot thing. Chat bots, you love them all over them, right? That you know they how does this how does a Gen. AI chat bot compare with the what I will call the legacy chat bots we have today? Feel free either of you with that one. Yeah, no, I mean, yeah. So that's something we we see like if I go to a utility website, which I did yesterday, their chatbot it's it's very standard. So it's FAQ you, you, you give it a question, it's it's going to give you that answer. If you don't give it, you're not going to get the answer. The benefit with generative AI and something we can do so quickly and turn it around and actually give a customer to test this. When once we scrape and crawl all the information and, and, you know, we do the proper testing, the generative AI will actually, you know, the, the, it's basically the, the, the way you're having the interaction with the chat. It's, it's, it's totally different. It's, it's, it's more personalized. It's, it doesn't set, it doesn't read like your typical normal chat bot, like the old ones that you just mentioned. It's, it's something, something new. And, and the fact that you can type in Spanish, German, French, whatever language you want and it can, it can actually translate that like instantly within just a few seconds. That's awesome. But the best part with Gen. AI, if it does find the information on your website, it can actually guide the customer where they can find this information. So if I do want to know something more about EV or solar or energy efficiency programs, it can explain it to me in a complete new way, but it can also take me where I can find this information. Imagine if my utility had that I would have never had to pick up the phone and speak to 8 different agents and spend an hour and 30 minutes on the phone. I could have that that chat bot could have answered solved all my problems, right? So it's those kind of things that we that you can do right using a Jenny Ibot. OK. Can we can we come back to hard numbers because I've had some questions in about this. I mean, you're doing this live implementation at Hillsborough County, you know, can, can any of this benefit be measured? I mean, you know, the cost to serve is a is something that troubles a lot of my clients. Are we seeing any benefits like that Bram, anything around reduced cost to serve or is that is that not where the benefits are here? I think the the, the use of this service at let's say the the call centre it, it, it is not I think intended just to remove cost to serve. I think it's really to make the life of the agents easier and better because you will always have customers that want to call and speak to, to a real person, let's say. So I think that the the main use case currently today is to really improve the the session that end users have with an agent that it's a A1 time right fix that the call is shorter than the normal. And in the end that will result of course in in lower cost to serve. But I think the main goal right now is to indeed find the right fine tuning, if you will, of this of this agent support system in in this specific context of a of a call Centre. OK, Thanks. And one more question I've got is, is, and perhaps a key question to, to, to wrap up before we, we go into our poll is I'm going to ask you each in turn. So we'll do it the other way around this time. So Steve, you first, what's next for this technology, right? Seems to be being fast. Are there any new capabilities or integrations or anything, you know, new breakthroughs that you're particularly excited about in the, in the in the next 1218 months? Yeah, So something we're working on and we showed it last week at the Oracle Utility Conference in Nashville was around conversational AI voice is still like a very, it's kind of like voice is never going to go away. But if you can do it right, and if you have the content and the knowledge where you can assist customers, like no matter what time and you can integrate it to your back end system where they want to maybe know what their balance is, they want to do a payment, whatever that may be, right? They want to maybe move from A to B. We're going to see more automation down the road using conversational AI, agentic AI, which we didn't talk about today, where agents can now assign tasks, right, and help them in a different, complete new, different way. We're going to see more, more of that. And then of course, the digital AI avatar that I mentioned. So I think we're going to, we're going to see a lot more of that. It's, it sounds all futuristic, but let me tell you, we're it, we're, we're, we're very close. I mean, it's like, it's, it's pretty fun. And this is what kind of motivates and excites us. And before passing it on to Bram, one thing I want to say about Hillsborough, because we did start in January. One thing the users look and these you got to imagine these are CSRS, they're going off into retirement. Hillsboro is actually going to lose 20% of their staff members by the end of this year. So being able and to find new recruitment who want to come work for utility, it's hard but and I'll pick as a lot of them have that knowledge, but one thing we've done really well was we've taken their knowledge, we've embedded into the AI platform. We allowed them to be part of this change process and the fact that they're enjoying it and using it every single day and it's it's easing their life and they're not making the same mistake over and over on the phone, that's already huge. OK. Thanks for that. And and Bram, can I ask the the same question of you? Is there anything coming up that you're particularly excited about? Yeah, actually there there's a lot happening in this space. And as Steve said, it's really this this multi agent approach, right, that one agent controls other agents that that's really going to make a next shift in this whole Gen. AI space. I think that that's going to be used also on the internal process sites within utilities and energy retailers to optimise also workflows and these kind of things. But what I'm I'm very excited about as well as also to run simulation. So if you can imagine that you can make a a digital version of your house by doing, for instance, disaggregation services and with that same data use it to run simulations on so-called what if scenario. So it's really the the simulation of sort of a mini digital twin of your house and an energy retailer can do that for the whole customer base. So what will be next is indeed linked to energy forecasting of your whole energy portfolio. If you know exactly what is happening behind the meters of all your customers being analyzed, What what does it mean if they would introduce a certain tariff type? What if they start a certain campaign on selling specific assets that will be installed at people's homes? What does it mean for your energy usage profile as a as a company? These kind of let's simulation tasks. I see a big help by in those digital twin simulations, both on the grid side as well as on the energy retail side. OK. That's really, really exciting. Thanks for that. OK, I'd like to ask, come back to my poll and I'm going to ask the audience to answer the same question again. I didn't tell you I was going to do that because I'm interested to see if this conversation has affected your, your outlook. So again, you've just got about 2030 seconds to answer the same question I actually before. And it'll be interesting to measure how this has moved on. So well, we will see generative AI transform retail energy customer services in the next three years. So I just wondered if your views had shifted based on what you've heard today from Bram and Steve. There's no right or wrong answer, just a bit of fun. But as industry users, you know, we'll, we'll share these insights online as well, LinkedIn just to get the conversation going. But, and it's a conversation we intend to keep coming back to. OK, at LCB Delta, we're very much of the opinion this is too big to ignore that you can probably tell from some of my questions, I'm sitting here going, wow, is this real? So what do you think? Will, will we see generative AI transform retail energy customer service in the next three years? I'll give you 5 seconds more. So 5432 one, OK. And again, so OK. So what we've seen is a shift towards well, almost everybody thinks it's likely to be transformation basically to some degree 95% of of our audience today, which accords very much with LCP Delta view that we don't quite know this beast is going to look like, but it's too big to ignore. It's time I think start thinking about proof of concept, some small scale trials, some but but actually doing this, not just playing with it in the back office and giving you your IT team. OK. Any last questions from the audience? I've I've built some of those in. If there are any further ones coming in, we will deal with them offline. I'm very conscious the fact of time. I'd like to give you a few minutes back, but I'd like to leave finish actually with one last sentiment from Bram and Steve. So what's the one thing that you think people should be taking away from this session? Bram, you first. Yeah. I would say that that there's a lot out there already that there's been experimentations happening across the world also in other sectors like water and telecoms, banking. I think they're great examples also from those industries that can stimulate also I think adoption of these Gen. AI services for energy retailers. And definitely there's also a lot of reference designs if people want to build something themselves in a small fashion and do pilots, but definitely also suppliers and, and tech companies like power connect that have this available out-of-the-box. And, and I think it's, it's not something that's totally new and where you have to start from scratch. So that's, that's what I would like to give back to, to the audience and reach out to me and Steve, if you like. And we are there to, to guide you in your, in your transformation process using these great tools. Because I think it's still underestimated what the power is of, of this technology. It's still seen as a little bit scary technology. Maybe it's, it's really not scary. It's, it's really helpful. It's really helpful. Thanks. And Steve, what's your key take out from this session? Oh, Bram said it so well, I can only add what what he said, but yeah, I would say just really look what's out there. Take your time, especially again, because we've already had so much practice, you know, working with the utilities. Take your time in the test phase. There's a lot of free AI courses out there that you can use that You can already take either on AWS, NVIDIA, Anthropic Lang chain. Take your time and if you do have any questions, you know more than welcome to reach out to me and Brandon. We can kind of show you the possibilities. Yeah, I think, I think one last question is just popped up as you know chat bots, right. It would be cool to know if you guys could actually demo a chat bot straight out-of-the-box. If if somebody said can you show me a chat bot, could you do that? We could. You could, we could, we could even turn, we could even do a free POC within 48 hours. That's, I mean, that's, that's the kind of good practice. No, seriously. That sounds fun. Yeah. Yeah, I think. I think that's it. It's about convincing people this is real. It's the theme that I get from clients all the time. It's like seriously, there's a bit of that in the marketplace. But yes, I think that's the stage where it would be my key take out to wrap this up would be it's time to, to get our hands dirty and, and to actually do those POC's to get that proof of concept down and, and to have a go because it, it's something that's too big to ignore. So I'm going to finish off there. I hope you found that helpful. If you have any further questions, do come back to us. I'll share the slides afterwards. Those of you who asked questions that I haven't dealt with, we'll come back to offline and we'll also share a final copy of the video. OK. So thanks very much for your time and, and, and have a good day. Thanks. _1743402515391