May the Agentforce Be With You: AI in Energy Services

We've all heard the buzz around Salesforce Agent Force. What is it exactly and why is it so important? And what does it mean for your organization as well? And we're going to dive into also how does this AI impact the energy services sector? We're going to answer all these questions and more on this episode of the CG Hour. We titled it May the Agent Force be with you: AI in Energy Services. Let's start the countdown. Welcome to the CG Hour where every other month we come to you featuring experts and leaders talking about the latest hot topics in technology and business. If this is your first time tuning in, we are coming to you live on YouTube as well as LinkedIn and the replay will also be immediately available afterwards on both those platforms. And for those of you out there in the comments of the feed, be sure to introduce yourself, tell us what you do and let us know where you are tuning in from. We always pull from questions and comments and also bring it up live here on the show. Be sure to introduce yourself and ask questions of our guests throughout the show. My name is Fanny Dunigan, I am the host of the CG Hour and we are here talking about AI in Energy Services. And because it's May the fourteenth and we're all Star Wars fans and nerds here, we thought we'd title it May the Agent Force be with you, AI in Energy Services. So let's jump in. I want to first introduce you guys to the guests and our expert panelists here on the show. Over here on my right is Mike Hatch. He is the Enterprise Architect at Salesforce. Welcome Mike. Thanks Fanny. It's good to be here. Awesome. Please tell our audience a little bit about your career. You bet. I'm an architect. I've been in the architect profession for almost thirty years now. Started early on in a software startup and went to Microsoft, worked for Microsoft in Seattle area for almost twenty years managing the ERP systems that they use there. Then moved into the business team around Xbox and was involved quite a bit in Xbox Red Ring of Death, if you remember that, making sure Xboxes got repaired and had the good fortune of being there when Microsoft went direct to customer with stores and commerce and a number of things, building their own computer with Surface. I got to be in China when the first of the Surface computers was coming off the line, which was kind of a bucket list moment. That was pretty fun thing to do. Yeah. And doing architecture work in information technology, went from Microsoft into REI, worked in commerce and retail for REI for a few years, and then relocated down here to Dallas to work for Fossil, that's then when I started working for Salesforce about five years ago. I had the good fortune joining them. I found Salesforce after I had left Microsoft and got pretty interested in how Salesforce is solving problems that I was building solutions on top of SAP for a number of years. So kind of been around the ringer of integration and commerce and ERP and various kinds of systems and it's fun now to help customers think about their enterprise architecture and the solution landscape and how Salesforce could fit in, and that's kind of what I'm doing now. Awesome. So There's so much buzz around it. Yeah. I can't wait to dive in with you. Yeah. The world's changing. Yes. And then on our far side here, that is Jonathan Goldstein. He is the SVP at CG Infinity. Jonathan, welcome to the show. It's great to be here, Fanny. Yeah, thank you for returning to the show. You've been on several episodes. But please tell our audience for those that don't know you a little bit about your career. Absolutely. So started in IT very early, I think around the same time. And kind of worked my way through business analysts to project management, managing PMOs. And then ultimately found my way into retail energy at North American Power and Gas, and that's how I met CG Infinity. I was their customer and I had the great opportunity to join the firm and be one of the founding members of the Energy and Utilities practice about seven years ago. That's awesome. Thank you so much. So before we dive into the episode and kind of raid all these questions on our guests, we wanted to play for you a video that Mike brought for us to just kind of give you an overview of Salesforce Agent Force. So let's play that now. Awesome. Thank you for tuning in. I see a variety of people already in the comments, Matthew Goldman, Molly Gabby, Lacey. Welcome everybody. Thank you for tuning in to this live show here all around AgentForce. So Mike, I wanted to kind of start with you, right? AgentForce has rolled out. What does this mean for IT professionals and what's changed over the last year? Yeah. First of all, I would say what's changing, of course, is the disruption that the generative AI tool set has brought to the industry. We are chasing after now what that means, right, broadly in technology because it's changing the relationship that we have with technology. Often I'm talking to customers about the conversation that we're having with technology. Now maybe I watched a little too, to use a different reference, Star Trek. Growing up, I loved the idea of being able to just tell the computer what you want it to do and then it would know how to do that. Probably why I stay in the architect profession because marrying kind of that business strategy problem with the how does the technology solve that business problem is is a fun and difficult place to be. You know, how do we use technology? And AI has created the ability for us to have a conversation with technology. Now we can speak in human language and get responses in human language. It understands and can interpret data and interfaces and, it doesn't need a user interface. Yeah. Right? So it's like, now I have a new way, a new modality of interacting with technology, which is exciting and disruptive. And AgentForce is how Salesforce has been responding to that even more than a year. Right? Now a couple years of development. You know, Salesforce started in AI with predictive AI technologies ten years ago. Right? So it's not really new. The only thing that's different now is the way that it's the language models have have added this new ability to interpret human questions and language and respond in a very personalized natural way. It seems like magic. Right? Any sufficiently advanced technology seems like magic. Right? And and Salesforce with Agent Force, that's the way that we've been layering that technology into the platform in order to provide a way to use Salesforce in that way. We'll talk about that today, but And you actually brought a slide for us. Let's pull up slide one, which is kind of the overview of AgentForce. Can you kind of walk us through briefly through this? Yeah, you bet. This kind of represents how AgentForce is becoming kind of that front facing way of interacting with technology, but layered on top of all of the great business solutions that Salesforce has been doing for twenty years now. Right? Whether that sales and service is kind of the foundations that Salesforce started with, but now across revenue and commerce and industry specialization, lot of the work that I'm doing now, I'm I'm focused on energy and utilities as an industry space and industry clouds that layer on top of those core services to specialize within that industry space. And we're doing that in healthcare and hospitality and media and entertainment and manufacturing and automotive and energy and utilities. And so AgentForce now is this layer that becomes the way to expose those business capabilities to conversational tools, whether that's chat based tools, that's what you see most of the time now. Right? It's how we've been interacting with ChatGPT or whatever is asking questions. But also it's very quickly now penetrating into voice, so I can just speak to it like we would do at home with Alexa. And other other ways of interacting with it. Again, all those conversational ways of interacting with technology and incorporating it into the Salesforce platform in a way that allows you to you create and manage your own solutions. We're creating agents out of the box. So there are energy and utilities agents that are predefined around energy needs like billing questions, right, or field service questions and technicians that are interacting with the technology, but also then creating AI tooling so that you can extend those out of the box agents and create new capabilities or your own agents to solve for various needs. Yeah, you mentioned agents. I know we also have a diagram for agentic maturity levels as well. Walk us through that too. Yeah. I mean, the the gray level zero really is kind of what we were doing before the language models came along. Right? If you tried to do bots and you had to try to determine intent and that kind of thing and you had to hard code what it was doing, but it was very fixed and repetitive things. In fact, I would say that's how we've been managing technology in the large up until now. If you had a use case, had to think about where where's the data? Where what is the logic around the data? What's the user interface that I'm gonna build? And then the IT team would have to wrestle with testing and iterating on top of that. Agents change that paradigm. Language models change that paradigm significantly because now I don't have to think through the entire use case to deliver capability. I can plug APIs or data into the agents and immediately allow those agents to interact with that information in a dynamic way. We'll talk about that today. It's it can reason about what needs to be done and select the best path to solve for whatever the user might be asking for. And so that the nature of that in level one then is like what we've seen as we've all started using these tools. Right? Is information retrieval, ask a question. In fact, that's pretty much what Internet search is now. Right? You're you're It's kind of a chat GPT model right now. Right? Yeah. Right. Or if you're in Google and you type something in the address bar, your question, you're the first responses and and AI generated response. Right? Information retrieval. But then how do you get in level two and three is about orchestrating across different domains where there might also be permission sets. So you have different agent roles that have permissions to do certain things so that they kind of stay within their boundaries. Right? This kind of Oh, yes. Talk about that some more. You definitely want the guardrails. And then all the way to the right is what is happening with agent to agent interaction. That's where kind of the industry is innovating right now. We just introduced some announcements with Google partnership where Google has created some agent to agent protocols. So if you're in the box Salesforce agent needs to then hand off to some agent that maybe is outside of Salesforce, that agent to agent orchestration can be happening and that's where you're on the far right at the maturity level where it's going. Certainly a lot going on. But before we get totally into that level five, I'd love to hear from you, Jonathan, right, as a consultant, right? How do you kind of see the difference between how implementations were rolled out previously and now in this kind of agent force world? I think we're seeing a bunch of shifts. As you mentioned, data is the king, right? So in the media world, content is king, but I think in the enterprise world, data is king. So the structuring and maturation of enterprise data such that it can be made available to these agentic architectures is becoming preeminent. And I think we're also seeing a huge shift in the discussion of kind of an AI first thought process of like, AI assist in this? Can agent force can we build can we what can we do with agent force, right? Because a lot of the companies we interact with, they've already either invested in Salesforce CRM or they're expanding their investment and AgentForce becomes part of, okay, well how can I extend this to add even more efficiency in my operations? Yeah. Well let's get tactical, right? Mike, can you kind of list some use cases of Agent Force that you're seeing across your customer base right now? Yeah, you bet. The most obvious ones are the places where those bots that I talked about, where we've been trying to create support services bots that would be able to answer questions that customers are asking, but that was still pretty difficult. Now agent service is something that for call deflection, for supporting answering questions from customers, so that I don't have to go to the human agent right away. And the fascinating thing is, it's very personalized, right? So we're in the past interacting with that agent. In fact, I actually was interacting with an agent this morning because they're installing fiber optic in my neighborhood and they broke the cable line for my existing provider. And so then they patched it, but it wasn't working right. So I was like trying to tell them, hey, you need to come look at it again. And I had to go through this. Did you reboot the router? Did you do this? You know, like that classic bot experience. Right? I just wanted them to know, hey, we already know that we had a technician in your area yesterday. I'm like, yeah, I talked to him and, you know, like be personalized. That's one of the things that the first level. Right. It's not even And context aware. Right? Yeah. Yeah. It's very context aware. Oh, I see you have the service with us. Oh, there was an outage yesterday. I'm sorry about that. You know, here's what do you need, right? And then very personalized getting quickly to and it helps the customer feel connected, right? And I think that's a transition that's happening because I think we tend to avoid the bot experience because it feels very robotic. Right? Like I was describing. But now these agents can be so much more interactive and personalized without having to spend so much effort to code up. Well, what was the intent of the customer? What were you what do they mean when they say this? Those agents can figure out what that meeting is and then pick from options. That's the first layer. And it gives companies an opportunity to surface enterprise data in the way they haven't before. So like I think of the very famous Saks Fifth Avenue demo, right, that Mark gave at Dreamforce, right, where everything just works, of course, because Mark's doing it. But the return policy, this kind of sticks in my mind. Was like, if you have a return policy system that you can surface and make accessible to AgentForce, then now you can extend that customer conversation to where like, oh yeah, I can handle your return. And the return policy is checking like, they within the window? What kind of return do I have to do? And they're doing a lot of the decision work that a person would normally have to be on the phone to like work through and figure out, keep in their head. It's creating a very cohesive customer experience. And I think we could do the same thing for energy. Like where's, I just signed up. Did my application get approved? Just that ongoing conversation that we have in the energy space. Yeah, in fact, if you show that rainbow slide again too, think there's something interesting to talk about there, which is That was slide one. The data conversation with agents is blending structured and unstructured data. Right? So from a technology perspective, we used to to think very consciously about, okay, my what's my relational data model? And I mentioned that's still true. But there was all these documents like policy documents, right? This made me think of this unstructured information. And these agents can pull information out of the unstructured as well as the structured. And in the middle here, see Data Cloud, which represents Salesforce approach to scaling data conversations. But in terms of Agent Force, that's the place where I can take a library of documents that I already have policies or maybe equipment instruction guides for that technician. And it's immediately available. And you can I do this demo all the time because it's easy to take a document of a policy, upload it into the library, and then ask the agent questions about that policy and immediately get back responses or have technicians be able to ask questions? Well, I'm using unit number one, three. I need to do this. How do what's the process? They could say, oh, this is Well, I think it extends again, kind of coming back to the energy model, right? If right now the model is if I have an outage, as a customer I'm calling to say, what's the status? Whereas if I'm interacting with an agent force system that is context aware of the field service operations, the status of the grid, the status of my account. Now when I'm interacting with that bot, could be like, field technician's on its way. It's this poll that's affected. We have a team already out there. We'll proactively notify you when energy is about to get restored. So it's just that taking the customer experience, I think, to another level that we haven't really been able to in the energy space. Well, like you said, it's going from reactive to proactive, is huge, huge. For those of you in the audience, if you're already using Agent Force, let us know and let us know what you're using it for, what kind of use cases. We would love to hear from you. I want to shout out Tanya Kapoor in the comments here. She also watched that AgentForce two point zero keynote and she is very excited to know how humans and agents can drive customer success together. And Mike, I know you mentioned already, right? Just kind of like those kind of personalized conversations. Any other examples of that? Yeah. I think we've hit on the kind of a service oriented scenarios and field service scenarios, technician knowledge and that kind of thing. The other one that's been very hot off the presses, I was at Distribute Tech here in Dallas here a couple weeks ago and talked to a lot of customers about this billing, like billing usage. Why is my energy bill so high? Right? And now the agent can immediately reply and say, oh, yeah, I can see that your usage was this much and this much over the last three months. And the weather in your area has spiked up. So your AC is probably kicking in. So then it's pulling weather data too. Right. Now it can go across different scenarios and reason about that with the customer. And then one of the things that we're doing is clean energy programs, right? So I, as a provider, I might have like a smart meter and the more that I can get customers to adopt some of that to then start helping with some of the green energy, clean energy efficiency, the agent can immediately say, here's some advice, maybe reduce your AC usage during the day or whatever, how could you do that? Simple guidance. We also have a program here where if you replace the meter in your house, they see you are using meter number Y today. Here's a program now you can get a new meter. Would you like me to schedule that support, Or offer that to the customer, right? So now the idea of predicting the next best offer, which we've been doing for a long time, becomes part of the agent scenario, right, to continue the conversation. Also allows the energy company to extend product offerings. For example, there's an energy company where, like, if you let us control your meter or let us control your thermostat rather, we'll offer you this product. And it might be a little colder in the winter and it might be a little warmer in the summer, but your energy bill is going be much more affordable. Are you willing to do that? And so it's creating a conversation between an company and an energy customer that Because even as a consumer, right, then it's like I didn't know I wanted that, but now I'm being offered something that benefits me that I didn't even know that I would benefit from. That's fascinating. Now, Mike, obviously as with everything AI, it comes with caution sometimes and privacy concerns, security concerns. What is being put in place to ensure that? And I think you brought something for us as well with the Einstein Trust Lead. Yeah, wanna comment before we go to that is address the question about with these agents, a lot of people are concerned about, oh, this is gonna take over my job. Right? I think everybody's kind of thinking about that. And that is not what we're experiencing as people are adopting these agents. I think it's similar to other technology shifts that we've seen in the past where it seems like technology is gonna be taking over. These agents become kind of human augmentation, right? Relieving you of kind of the busy work, the things that are time consuming of going and collecting information and trying to figure out how to respond. And letting the human agent be much more effective and personal in the way that they're interacting when I'm interacting with. So like before we talk about the security thing, wanted to kind of mention that because that's something I think people are concerned about, right? But what we're seeing is that applied in a graceful and conscious way, they're augmentation. Agreed, agreed. And handle those kind of like out of the box scenarios, right? But the unique scenarios, the humans can handle that and then leave the AI agents for more of the repetitive, predictable I think it frees people up to be thinkers, right? And abuse them kind of with information at their fingertips so that they're having a much more intelligent dialogue with the customer. They're not fiddling around like, Hey, let me go search something up real quick. That's a very smooth conversation because information is being brought to their desktop in real time that might not have been available before. I'm thinking of like the agent force call summaries, the account summaries, like the pushing information to an agent's, human agent's desktop that, they would have had to search for or rifle through, find notes, right? It makes them a lot more effective. And I think to your point, it does not take their job away. It actually makes their job easier and it frees them up to take care of more customers on a given day than they might be able to without it. Yeah. Talking about the deeper kind of trust and security question that you were asking about and kind of the slide we have to talk about that. From the beginning of generative AI Salesforce has been thinking about how do I make sure that I'm applying trust with these new tools? We've been leading the AI thought leadership for a long time in that sense. But Salesforce has created what we call the Einstein trust layer, the agent force trust layer, probably now is the language for that. And, it is about ensuring that the prompt to the question that's being asked goes through a series of steps to ensure data protection and reliability. Right? So retrieving data in a secure way, grounding that information on my enterprise data, then masking it. Like if I have sensitive data that I just don't ever wanna go out to the language models, I'm hiding it. Also masking it, maybe the person interacting with the agent doesn't have rights to see some identifier or something like that or a credit card number that it gets masked when it's going back to the person asking the question as well. You can create policies for that. And also then zero retention. So Salesforce on the right hand side here you see models, right? That represents the language models. Salesforce is not building our own gigantic language model. We're supporting connection to the language model that you would choose to use. So we support all the the models that are out there. You can bring your own model and then select that as a default language tool. But then zero retention with any of those tools is something that we're setting up those agreements with them. So we're not training OpenAI's tools on your business data. Right? It's protecting your business asset and your business data, your customer information. And then on the way back after those models have generated responses, there could be some toxicity. They're trained on the Internet, so you're not exactly sure, like maybe some some erroneous type of responses might be getting into the the language responses. And we're within the Einstein Trust layer detecting that toxicity and preventing that from penetrating the business interactions or the customer interactions that you're having with our tool set. So that's how we talk about trust and the layers within the technology that Salesforce has been applying. So that when you're using AgentForce, you can rely and know that it's protecting your business information, Protecting your customers. Jonathan, anything you want to add to that from what you hear from your clients around trust? Yeah, think, so where my head was going was, I really love that you pointed out that it does not have to be the open GPT LLMs that we hear about, right? And of course, many utilities have governance mandates as though they're not free to use OpenAI LLMs, right? But if they are able to develop their own LLM, then that can still take advantage of some of the things that Salesforce is bringing to bear in terms of customer service, customer sales, support through field service. It doesn't it's not prohibitive. So yes, you can use GPT in a very secure method, right? They don't need to know my particular address in McKinney in order to generate a non hallucinated response, right? You could know a customer in McKinney with this kind of generation history is asking about this, right? Like you don't need to know, the LLM doesn't need to know all the details. But if the governance mandates for a lot of these public utilities requires that they not take advantage of generative AI models that are open, there's still a method for them to take advantage of this AI wave. Right. It's fascinating to see all those layers, Mike. So thank you for sharing that diagram with us. For those of you out there, let us know any kind of questions you have for our experts here. Now is the time to get your questions answered about AgentForce. And while we do that, and while you are typing those questions in the comments, we are going to play for you a segment that Jonathan helped prepare to dive a little deeper into explaining what Agent Force is. Let's roll that now. Salesforce's tools are about really refining all that apparatus, but doing so within a secure framework. So I think at a high level what Salesforce has done is truly remarkable in that we have had GPT for a while now. We've had the concept of LLMs for a while and machine learning for a while and neural networks for a while. What's happened outside of Salesforce I think is that at a technology industry level is the all these technologies kind of coming together and not only being like shape shifter and able to harness computing power that we never even thought of fifteen years ago would even be possible but also making it commercially accessible and not only accessible to IT groups but accessible to anyone and so that's kind of the promise of generative AI. Now what Salesforce did was realize like okay you still have some fundamental, barriers that that you should have right there's still a security concern you don't want to just copy paste your intellectual property into a GPT you know you don't you don't need the LLM learning from that right so your personal data exactly your client's personal data but we also know that those the more intelligent we can make a prompt the better the output is going to be. I'm writing an article right now called don't debug the code, debug the prompt and the concept behind it is that where we're going is really a very prompt based environment where where everything is going to start with well what's the conversation you're starting to have with your agentic architecture and that's what Salesforce has done is really given some end user tools that help you build the prompts, help you ground the prompt in CRM data so that the prompt is very inclusive, protective, but will generate the output that you need because we are still in the world of garbage in garbage out. And now the stakes are a little bit higher because you can't always tell when a generative AI framework is hallucinating or it might take a while to be like oh that's completely off script. So Salesforce's tools are about really refining all that apparatus but doing so within a secure framework. So you have what's called the Einstein security model where before it even gets to the LLM, before it even goes out to GPT, there's all these steps that have to happen in order to get there to protect your data. That's what's really impressive I think about the sales force offering. Thank you, Jonathan, for that great explanation as well. When we were prepping for this show, you had a great term that you used around how we should think about AI. And you said that think of AI as a team member now. I want to talk about that for the next segment here. Maybe you can start us off, Jonathan. How are you structuring project teams now if AI can handle parts of development and coding and QA? What is the how are roles shifting now on project teams? A couple ways. I think for the developers and QA engineers, their job or their role I should say has really stayed the same. You need that developer QA mindset on a big enterprise technology project. What AI offers them is a way to move a little bit faster, try things out in a forum that's not going to **** out project time, right, and try to iterate a little faster than normally they would be able to. Because these GPT frameworks, like they're able to generate code at a speed with which no human is able to. Now you still need that mindset to be able to look at what AI generates and critically assess. Did it build it right? Are there changes that we need to make? So there's that. But I think the other shift, which they and. And the other shift is members of our team that had very defined roles, BAs, right, I think. Business analysts. Business analysts, it comes to mind, right, where their job has really been about textually defining what the requirements are, getting that into a repository somewhere. Well now they can be kind of part of the team of like, I've gotten a framework this far, can you, the development team, can you take it a little bit further? And so now we're seeing some interaction where BA is part of the development process, not part of the pre development process. So that's I think a huge opportunity. Mike, anything you want to add to that? AI as a team member? Yeah, I agree. Well, some of what you're talking about too is in the first part we talked a lot about AI is interacting with technology. AI and language being able to speak to the technology in natural language is how I'm telling the technology what I want to have done. In other words, I'm programming with natural language, right? That's what a prompt is. And the agent force framework instructions of do this when they say this, do that. Don't do this. It's still a natural human language thing. So I'm not surprised, right? The BAs are now taking a much greater role. And I would say that's kind of where we are at this point in time with AI is that technology management is still about take that prompt, make sure it's governed well and deployed operationally in a way that makes sense for the business. I very quickly see this moving into my use case development. My business requirements management is now much shorter. Because that natural language that my business person is gonna use can go more directly into the environment to then create the solutions that I wanted to do. Yeah. And to your point about your earlier point about enabling, right? I think it enables that process to move faster because now I can use AI to document, to generate the documents that normally would have taken a couple days, if not a week or so, to get a business requirements document out or get it into a format where I can move it into my Jira or my Azure DevOps. So, yeah, it's all those enabling capabilities that AI is bringing. We have a great question from the audience, Jason Winningham. Thank you, Jason. He said he was reading yesterday that novices are now getting into building program and coding through the help of AI. So will humans with education and coding still have an edge, you think? Or what should humans do to prepare themselves to have an edge in the market working with AI? Thoughts? Oh, absolutely. So, will humans, just repeat the question one more time. Humans have an Humans with education in coding, well they still have that edge. I've had this conversation with our CTO quite a bit and we both agree. Absolutely, that coding mindset, nothing can replace that. The person who understands how code should be is the one that's gonna ensure that the solution that we deliver to our customers is not saddled with technical debt and and potentially moving code into production that is more difficult to maintain, more difficult to support. Fundamentally, that's what our job is. And so I think the developer, the person who can reasonably look at the code that comes out, or I would say critically look at the code that comes out, will be the one that comes out. Now the person who will be left behind is the developer or the QA person who says like, I'm not going to embrace AI. That would be a mistake. But I think the technical people in our midst who see AI as an opportunity will reap many rewards as we move forward. Awesome. Mike, I know you brought for us kind of the on the AI side, right? There are certain attributes that agents will have. Can you walk us through that? Yeah, this connects to with what you were talking about, Jonathan, as well. I'll pull up slide four here to talk about it. The attributes of an agent are what is the role that that agent is supposed to do? What's the information that it has access to? What are the actions that it can take? What are the things that it can call out to and execute? And then those are maybe not that different than what we've done classically in development practices, maybe express a little bit differently. But guardrails, I think is an interesting one, right? Because when we built technology solutions up until now, the developer was on the hook for writing the code structures in such a way that accomplished the business outcome that I want to achieve and I test for that. So it's a rigorous and very structured approach, right? With this language models because it's human language and because it's not with the same output, always get the same input, always get the same output. With the same input, I might get different outputs every time. I now have to put guardrails in place. So there's this kind of new concept I think for the developer, for the person managing the prompt to say, again in natural language, don't do these things. And we don't necessarily code that way today, right? We don't like say, here's what I'm not gonna do but everything else is Yeah, we say, we put in the prompt use defensive coding standards and list out the defensive coding standards that you want the model to use. And so that yeah, when it generates code it's commented, there's error handling built in. Yeah, great. And if I would comment on what you were saying in answer to Jason's question? Yes. About humans. I think the language that you've learned and that you know probably is less important because the AI can interpret and understands those languages whether it's JavaScript or Python or whatever. I think the language is less important, but it can't replace the logical steps that the human element of is still saying this is what I want those outcomes to be. And the Roger Penrose is famous for saying artificial intelligence, intelligence is the wrong word. It's kind of artificial cleverness, right? That's great, like that. These language models just know what words go together. They don't really know what the words mean, right? And so the human element even in the development space, think is still so important. It's more language agnostic and more human language centric than it ever was before. But no, I don't think it changes. Well, to Jason's question, the novice can be as much a risk as they can be a benefit. That person's like, Oh, I can code this. Yes, you can. May not know all the scenarios though. Exactly. So we do have to be careful. Mike, we also got a question from DeWalker in the comments. How does AgentForce support agent training and onboarding? Agent training as in? I'm guessing it's human agent training. Human agent training. Yeah. That's a fascinating question because the way a human will interact with technology and education about how to use that technology becomes part of the agent experience as In fact, I've watched a few YouTube videos where the training, the folks who make training videos about how to use software, AI is being used to say, oh, oh, no, you clicked on that. Go back one. You you need to click on this thing, and then that's how you get to this part. And so AI actually is being used to support how do I use the software. It's like reactive training. Right. Almost, right? Right. And so there's an evolution of that happening. And agent force could be used to help guide the user to the right things, or at least get them to the right information more quickly. So again, think it's less about the UI first approach than it is becomes what's the right information that the agent is gonna direct that person to to help them answer a question or be knowledgeable. One of the use cases that I do all the time, I used to have to create business models to map out. Well, I'm gonna go talk to this customer. Here's what their business is like. And I would draw some pictures and try to understand who their customers were. And now I can ask agent force that question and say, is the business model of this company? How do they make money? Whatever. And also it'll tell me what their existing footprint is. Right? Because inside my Salesforce org that we use internally at Salesforce, we know what software that customer has. So it's those kinds of things where I used to have to do discovery and kind of find it. The agent tools are immediately just delivering that to me so I can go more quickly to, okay, what's the thing that this customer might need? It truly is like accelerated decision making, right? Yeah. Okay. Jonathan, with all these tools now, what would you say is kind of like an overarching tech stack that organisations can start to look at and groupings and things like that? I think the tech stack itself hasn't changed. Like we're still going to use Versus Code as our we're still going to need an IDE. Now what's happening, I think what's exciting in the AI space is we're starting to see companies that have IDEs be purchased by the companies that have the models. That just happened with OpenAI purchasing Windsurf, or at least announcing the purchase last week. So that's pretty exciting. But our IDE, whether it's GitHub or whatever you choose, will still be the same. What we're seeing is extensions that then bring in Gemini and bring in Copilot and other enterprise AI models that can then couple with your coding methodology. But now it's that natural language. I need to code this, can you help me? And it just generates the code. So it's kind of in the same space, I think. I don't think what we use to code will change, but how we use it will absolutely shift. So Mike, as a customer that wants to get started, right, what would you kind of say is the preliminary steps for customers and clients to get started with Salesforce agents? I'd say two things. One would be understand what use case you want. I mean, that's the conversation we're having with everyone now is it can do pretty much anything. It understands human language. It has access to your data. What is it that you're trying to achieve? Where's the most value? And that's not different than every customer conversation I've been having up to this point. The difference being now it's more focused on what's the role, what you want the agent to do, what channel are you gonna expose it to, right? Is this gonna be customer facing or is it employee or partner facing, however you wanna use it. So understand the use case and narrow it down because you can do a lot. Right? You can do pretty much anything. So get focused and then start prototyping that particular use case and learn and grow agility becomes ever more important. The pace of change now even with generative AI has accelerated yet again. If you didn't think it was fast enough before, now, like, I I heard the other day that the agent force team is releasing every week. Wow. Right? We have our three releases a year that we've been doing. We're kinda talk about that a lot. But the agent force team is iterating very quickly. I I that's not all penetrating the product releases I think every week, but the pace of change. The other thing I would say is the thing that I really like about AgentForce is it is giving a framework for doing that exploratory work. Create a prompt, create an agent, start experimenting with it. And the agent studio tooling lets you do that and test against that in ways that help you kind of prove out what you're trying to discover. In fact, you can even have the AI tools generate all the use case the test cases for you. So like, well this is what I'm doing, here's my prompt, generate a bunch of test cases against this and run them against the prompt and then tell me whether or not you think it passed or failed. And so the iteration even becomes more quickly. Would be a tactic. Essentially Salesforce becomes the IDE, if you will. Right. Yeah. And the other thing I would say is operationalizing agents in the world that we're in. This new technology has penetrated in a way that it leads us to kinda start building new things because I can go quick and I start building something. Taking, you know, a a build approach to an API that you might be using Gemini or whatever, and making that into something that's production ready is got significant new new additions to it. How do you monitor it? Where are you logging that information? How are you validating? And agent first framework provides that operational framework as well, so that I can deploy something into a production landscape very quickly. And I have the Salesforce ecosystem helping manage that for me on the outcome side. So that's something that I think is overlooked. Sometimes customers I'm talking to, yeah, we've already got a plan. I've got a developer, he's building something with OpenAI or something like that. Okay, great. Where are you? That's great innovation. Don't stop. But how are you operationalizing that? Got it. Well, Jonathan, as an expert in the energy services sector, how would you kind of build on top of what Mike has just said as an additional advice for energy services companies that want to get started with these agents? Where I have found myself is because of the speed of change. If you have a project with a limited budget and a time that something needs to be delivered, something will shift in the AI world before you deliver that product. So you have to pick a horse essentially and just say, I know this model might be outdated in six months, but I'm gonna use it for the purposes of this project. And just so I can keep my team focused because, I mean, just within, I started a project in March using AI and within that time span there's been another LLM introduced. So, you know, it's impossible to, I think, for everyone to stay on top of the change and continuously shift your project methodology to embrace it. You can be on top of it in terms of the knowledge, but the application of it, everything takes time. You do have to kind of pick a horse, I think. Yeah. And then just stick with it and not chase after another shiny. Yeah. Because there's gonna be thousands of shiny objects everywhere. Yeah. I think the innovator's mindset is important there. Agree with you. And this this tool set in technology is finding its footing. Right? Recently, the model context protocol stuff that's been coming out about how to hook tools inside agents has been coming out in the market and there's a lot of hype about it. And this agent to agent protocol, that's where the kind of innovation is happening. Once I've defined an agent, how do I hand off between them in a secure kind of way? So yeah, I think it's gonna continue to evolve quickly because this technology is being applied in a business sense. It's exciting and we need to figure out what's the right adoption approach, right? Yeah. Well, this show isn't possible without CG Infinity and their sponsorship. So before we continue and kind of look towards the future of agents, we wanted to play for you this video around the experts in energy and technology at CG Infinity. So let's roll that video I may look thirty five years old, but in reality, I was involved in these meetings when these standards were established back in the late 90s and the early 2000s. In the retail energy industry, one of the larger challenges is getting the right data where you need it to be to quickly run a multitude of processes, especially within mid office. First of all, I'd have to say why it's very important that your customer count is accurate because if it's not then you aren't scheduling or buying the correct amount of energy and that will end up costing you money. So with energy pricing you have this aspect of there's a market based component that is shifting often in and of itself. You have transportation costs, you have your own cost to serve. It's going to feel very jumpy and roller coaster, but in the end, I absolutely believe AI is going to create more jobs than it's going to displace. Retail energy companies are generally low margin companies. One percent of fraudulent customer could actually take away all the profit for the year in retail energy domain for that particular company. The utility has no idea that this is supposed to be your customer unless the transaction goes out. So now you've got data in different systems that aren't necessarily talking to each other. But to run these processes, you need bits of, if not all of this data in one place for this calculation. The company that can get contracts out to that customer in the speediest way possible with still observing the process internal processes that they need to review a contract and make sure that risk and portfolio and price and customer account details are all managed effectively. But we know the industry and we have helped our customer minimize their fraud impact. Within our team, we've got seasoned professionals with decades of experience under their belt in multiple facets. I know watching all that and also seeing some of the comments in there, people are saying you guys give great explanations. And I think that's what's needed right now, is almost like translators and interpreters of this human AI world, right? And you guys are kind of bringing those things together. So thank you for that. Mike, I want to bring back that slide where slide two where we talk about the agentic maturity levels, right? Because I'm super excited about all the agentic capabilities, but something that still makes me a little nervous as a consumer is the agent to agent interactions. So what would you say to those kind of fears or concerns out there about agent to agent interactions? Yeah, I would say first of all, this is a place where kind of innovation is happening as it starts to become reality for many businesses. I have agents, I have those roles defined. They're answering questions, retrieving information, orchestrating across different domains of business, whether that's like you were describing earlier. I have some energy billing question, but that becomes then some sort of maybe field service appointment scheduling thing. Could maybe think of those as two agents within a Salesforce framework. Right? So those agents know how to do certain things and their guardrails are defined and they hand off between each other. But many of my customers are all saying, yeah, like all of my system providers are coming to me now and saying, I've got an agent that's coming with this for HR purposes, for example, right? And so you don't really want to have to have the user switching contexts and the agents can be aware of each other and know how to hand off. And that's where this agent to agent, multi agent orchestration is happening. Whether it's within the Salesforce ecosystem or it's across your application landscape, where the tooling and the agents that you're providing make that seem much more seamless. Salesforce has always been about creating a seamless single pane of glass, customer engagement layer, if you will. And this is a continuation, I think, of that in the agent interaction world. Thank you. Jonathan, I mean, what do you see in the energy services world of this kind of agent to agent interactions? What scenarios do you see for The kind of opportunities. The biggest opportunity that comes to mind, you know, the mid office, which is often where pricing and supply is managed, is extremely dependent on data flowing to them. And so you have front office capabilities that are showing you what your pipeline looking like. You have your back office, which is showing kind of real time consumption, data. So and then in the middle, you have this pricing desk, which is fundamentally responsible for keeping that business a going concern, making sure that the margin targets are being maintained and that they're staying in the money as it relates to energy supply. And how amazing would it be if they didn't have to be guessing all the time or waiting for systems to start talking to each other. If back office could talk to the mid office and be like, hey, consumption was a little bit higher than we expected. Or this data center is burning hot and we're going to be out of the money if we don't adjust our risk model. I can see that use case as kind of preeminent in the space and could be a game changer. And I guess as a consumer then I'm hoping to get cost savings out of that because everything's being anticipatory now too, right? Yes, absolutely. Yeah, awesome. I saw a comment here from Mike Reeves. Thank you, Mike. He says, There is a fear that AI agent force may replace the customer service reps. How would you address this concern with potential customers? Do one of you to Sure, I can start. Again, computers are not human. They can't feel, they can't understand how somebody else is feeling, right? They can only respond to what is verbally being traded, right? Which has no emotive context. So you still need a human to be on the phone. Now, what we can do is make that human enormously more prepared for that call so that they know why this customer might be a little irate or what is prompting this customer to call, right? And that just makes them better at their job. So I don't see how it would replace. I think it does create a front end, right, where a gate, where, hey, is there more information I can push to a bot that customers who are willing to interact with a bot can ask a bunch of questions and get the same answers that an agent would provide? That that that I can see is a risk to call volume, which could be a risk to call staffing numbers or call center staffing numbers. I can see that. But I think in the short term, these tools, would not be responsible for jobs going away. They'd be responsible for people doing a better job at their role. Yeah. I see a lot of customers that have call centers at scale wrestling with the fluctuation of the resources, the human people. Like they come in and take the job and then it's difficult and they leave and then I'm retraining and there can be a hundred percent turnover in some businesses that I've talked to, right? So I think that that is probably in part how easy it is to do the job and how can I get out the right information? And if it's cumbersome and hard, people aren't gonna stay there and do that, right? So that'd be another way to think about that question, right? As I think, of course, right sizing my customer service center, but making it more personal and making it more exciting for the people who are doing the job. The ones who are always senior in those call center roles are always the one who have the compassion, who are thinking about what the customer needs and answering those questions. You're going to empower people to continue to do that and let the rest of that grunt work go to the agent to find and be personal in that conversation at the same time. Well, as we kind of wrap up here, building on everything that you both just said, I want to end this conversation of AI with a discussion on culture. My final question. And that is, at this rate of development, what kind of organizational culture should leaders be building so that it's ready at the human level for AI? Jonathan, you want to start? Sure. I'm thinking back to a conversation I had with my team a few weeks ago, where I encouraged them to be curious. And I think that is what this AI world kind of enables us to be, is innately curious in a very non risky environment where you can just try something. Just see if it'll work. The worst that happens is throw away code and we go a different direction, but at least we didn't waste days building it, right? We wasted an hour. And maybe in that hour we learned something about a problem that we wouldn't have otherwise assessed. So I think in terms of culture, what we're trying to do at CG Infinity, I think we innately attract curious people. So, but it's really, at least, I think we're trying to create a culture that's based on be curious and don't worry about being wrong because being wrong is an opportunity to learn and AI allows us to course correct at a speed we've not had before. Curiosity and experimentation basically, right? Mike, from your perspective, cultural? Yeah, from a culture perspective, I would say the things that are important probably are just as important today as they were before. I've always been a big fan of five dysfunctions of a team, Patrick Lencioni. It's a little dated now, I bet in Internet time. But demonstrating what you said, the leader being able to say, I don't know the answer to the question, let's find out together. I'll be vulnerable with you and let the people that are doing that work know it's safe. Hey, okay, great. Can and use the tools and experiment with a new use case and then come up with and share those ideas and provide the avenues to learn and grow into it because this is coming. AI is just gonna disrupt the technology industry. That's why Salesforce is responding with agent force and it's coming after your business. Your competitors are gonna be doing it. And if you're not in front of it, so create that innovator's mindset, learn and innovate and explore and grow, and find ways so that it can impact your business. It would be the cultural thing. Not different I think than it's always been. I couldn't think of a better way to wrap up the show is this build a culture of curiosity, experimentation, innovation, learning, exploration and growth. So thank you Jonathan, thank you Mike. Thanks Fanny. For all your insights and advice. Thank you all out there in the LinkedIn and YouTube world for tuning into this show. I want to call attention to our next episode, which will be August twenty seventh at noon, where we'll chat more about AI technology in the energy services as well. So thank you for tuning in. As always, I wanna to remind everybody of CG's great tagline and philosophy they live with, and that is people first driven to transform. So we'll see you next time, and thank you for tuning in.