Speaker 0: I will. Yes. Yeah. Okay. The last few minutes. Yep. So you’ll just pass it over to me, but I’ll I’ll I’ll message you.

Speaker 1: Alright. Sweet.

Speaker 0: Hello, MarDreamin, and welcome, everyone. We are excited to have you all joining us today. My name is Andy Marquette with Sercante, and I’ll be moderating today’s session. Before we get started, I have a few housekeeping items that I wanted to cover. Yes. These sessions will be available, um, they will be recorded and available to you on demand after the event. We will also be following up with them via email. So, um, just be on the lookout for those. If you have a question during this session, feel free to post it in the q and a tab above you see in the top right. And lastly here, use the chat. Um, there’s emojis, GIFs, and more. So, uh, we wanna hear from you. Um, so without further ado, let’s get started. I’d like to introduce you introduce you to our speaker today, Mark Good, who has awesome session ready for us, all about getting ready for the AI revolution. So go ahead. Take it away, Mark.

Speaker 1: Thanks so much, Andy, for the intro. Yes. My name is Mark Good, and we are here to talk all about one of the most important topics of the year 2023, which is the AI revolution. I just finished watching the previous session, um, with PJ, Nicks, and Danielle from Salesforce talking about, um, you know, what are all the AI solutions that we have, um, that are being deployed and what all is coming in the pipeline. They did some demos. Um, really fantastic presentation. So if you haven’t seen that yet, you’re gonna wanna go watch the recording of that. And and that sort of feeds into what I’m doing here. So definitely check that out. And but today, we’re gonna talk a little bit deeper into some of the topics that they brought up, um, about specifically generative AI. We’re gonna talk about how generative AI is changing the Salesforce landscape. We’re gonna be talking about how to get ready for that and the skills that you’re going to need to acquire to take advantage to take full advantage of generative AI and the new tool sets that are rolled rolling out. And we couldn’t do that, uh, or or we can’t dive into that first without thanking our sponsors for MarDreamin. So, um, big shout out to everyone here on the screen, and we’ll dive right into the material here. Uh, yep. I encourage everybody to engage in the chat. Sometimes you learn as much in the chat as you do in the sessions themselves. So, um, if you’re a GPT specialist or an AI enthusiast, head over to the chat and, uh, just banter around and, um, expand a little bit on what I talk about as we go. There’s a lot of, uh, you know, a lot of depth to generative AI and what it’s doing inside the Salesforce space. So, um, yep, engage in the chat, enjoy, and I’ll try to keep up with questions as we go as well and try to answer those while we’re driving through this. Uh, but first, we’re gonna talk about the current state of marketing with AI. We’re gonna talk about current Salesforce functionality, the future promise of Einstein One, which was which was announced at Dreamforce twenty twenty three this year, including things like Copilot, Copilot Studio, and the trust layer. We’re gonna talk about what makes all of this technology possible in general and, uh, give you a little bit of a a training session on generative AI there. We’re gonna talk about the skills gap that’s developing because of this new technology that’s been, uh, being deployed within the Salesforce CRM and within the entire ecosystem and including other, uh, ecosystems as well. So massive skills gaps that are growing. And then I’m actually gonna walk you through some really cool, uh, generative image examples for generative AI, which I think will speak to marketers everywhere, uh, and talk about the time savings and the creativity improvements that you get. Um, as you as you learn these tools, uh, and I think this will also speak to, um, you know, just the value of skill sets, uh, and gaining the new skills as we go. So what are our goals? We’re gonna understand what’s available, understand what’s coming, take a look at that ImageGen tool. We’re gonna be using Dolly three today, and I’ll walk you through some cool example. And then, uh, talk about the AI strategy overall and the skills takeaways. So we’re gonna talk a lot about the skills that you’re going to need as we move forward. I’ve got some links for you as well. So if you if you’re not on your phone, you’re watching from your desktop about your phone, I’ve got some QR codes in here that you’ll be able to scan and make sure just to bookmark those. Don’t leave the session. Come on back, but bookmark those for later so you can take a deeper dive, and I’ll have that to share with you, uh, in just a little bit. So first yeah. Uh, talking about Salesforce Marketing AI. Apparently, about three weeks ago, there were some things that rolled out to become generally available. Um, PJ and Danielle from Salesforce mentioned those in the previous session, so we won’t be diving into that. But as I learn those tools, uh, I’ll certainly be back for the next more dream and and take a much deeper dive into those. Um, things like, uh, marketing cloud product road map, uh, including a lot of really cool things there. But I did wanna show you this example that we have today from the, uh, the previous examples that Salesforce has provided as as to what’s coming for generative AI for Salesforce. And, uh, even though on the right, you see it says Einstein assistant, this is actually going to be Einstein Copilot, rebranded to Einstein Copilot for marketing. Um, there are features that are coming out for things like, uh, subject line generation, content creation, and AI for account engagement. But this is just an example of that Einstein Copilot feature, and what it’s doing here is helping us create or generate from scratch a landing play a a landing page inside of a marketing campaign. And so this is really and I’ll take a little bit of a deeper dive into this, but this is really a new way to engage with the CRM, um, in general is using a chatbot. So on the right hand side, you have that chatbot assistant, which is doing a couple different things. It’s creating images. It’s creating text. It’s adjusting tech text. It’s creating forms. And so as you go throughout the chat process, you can choose what items to add to the page. Um, you can, uh, edit items that you’re creating on the right hand side. If you generated an image and you don’t like it, you can ask it to edit those images, uh, and then you can insert that into the page however you like. This is this is an AI assistant. This is the Copilot tool, um, and creating a new landing page has never been easier inside of Salesforce using this Copilot tool. And Copilot I bring this example up because Copilot is not just going to be for creating landing pages. It’s gonna be used across the entire CRM. So all of the different areas inside the marketing, uh, cloud that you use, uh, AI chatbot so this chatbot style AI Copilot is going to be infused, and there are gonna be so many different tool sets that you’ll be able to use. We’ll talk a little bit more about Copilot and the skills that you can build with those in just a minute. But this, uh, generative AI for augmenting marketing, creating a land landing page, I think, if I remember, is expected to be rolled out in 2024. So that is what we have here as an example. So I wanna dive a little bit deeper and talk to you about the actual what’s important about generative AI as we’re watching these toolsets be developed and deployed within Salesforce, and this will start to give you a sense of the skills that you’re going to need to be acquiring as we move forward. So text generation is the first one. Just on the right there, you see, uh, just a screen recording of ChatGPT, uh, doing a prompt in response to some text. Uh, when people think of generative AI, this is one of the first things that they that they think of, these large language models creating text, being really good at, um, predicting text that comes that should come out of the model based on your particular prompt. Uh, and so that’s what we have here. That’s what everybody thinks about of generative AI. But if it goes it goes beyond that. We we saw in the previous example where we’re actually generating images as well. So image generation is another thing that you can do with generative AI. There are large line or large models like DALL E three, Stable Diffusion, Admidjourney, and there are tons of other tools that you can use to now generate images. And, uh, there are tons of different ways you can go about generating images for marketing and creating those assets that you need. Um, this actually changes the way that we think about creating assets overall. It changes the process that we use for creating assets, and it changes the speed and the level of creativity that we have for generating assets. Um, even in just this really simple example, which I’ll show you actually live in just a minute, um, the prompt that I used to get the image. You know, this image that you’re seeing on the screen is not too dissimilar from the image that we have here on the example for Northern Trail Outfitters. So I’ll show you the prompt that I used to get that. Very simple, very basic. And just imagine the the assets that you need for your marketing campaigns, for your emails, um, for your app you know, various assets that you need on your sites, all can now be generated where they couldn’t before. Now you’re gonna run this to Roblox. I’m gonna talk about the skills you need to to tackle that. But image generation is, uh, an important feature of generative AI. We touched on this, the new user interface. So when using Salesforce, no matter what cloud you’re in, specifically talking about marketing, we’ll be using that Copilot feature that, um, this new UI UI, we’re on the right side of the record or the right side of the object and chatting with the object in order to do things inside the org. This is a brand new way of thinking about engaging with your org. Uh, it’s a new skill set that our end users have to have as well as our administrators on the back end as well, configuring this type of new toolset. Again, think about the skills gap that’s being created with this. Um, and also, uh, you know, this is pioneered by ChatGPT or by OpenAI, the owner of ChatGPT themselves. They said the best way to engage with a language model with generative AI is to do it through chatbot. And that still has stuck around through today. Salesforce is adopting that model and that mindset through Copilot. And, uh, we get to see that play out, which is really fun. Human in the loop. So Salesforce preaches this. I do as well. AI right now, the state of generative AI is such that you get an 80% solution, and then as the human, you take it across the finish line, you check out, uh, you know, how how the solution looks, and you sort of tweak it and give it those final touches in order to get it across the finish line so that, uh, so that you can actually use the solution. So a human does have to stay in the loop. It just is what it is. Uh, for now, eventually, AI will get better. It’s improving over time. It’s getting iterated over time, and the solutions that you can get out of it will be more and more appropriate or correct to the the the task that you have it doing and and what you’re using it for. But for now, a human is going to stay in the loop. Okay. Switching over to talking about Einstein one. So this was the, uh, the massive announcement during Dreamforce. This was sort of a rebrand from what we heard was going to be marketing GPT and sales GPT, and that has all been rolled up into this platform called Einstein One. Uh, and there are some well, in this in this chart, we’ll just talk about some of the core things that we need to know about when we’re when we’re when using Einstein one or why it’s going to be why it’s gonna work so well for us. Uh, one is our data that’s on the left side. So that data cloud functionality, um, and there’s a ton of importance around data and making your data using best practice when you’re getting your data right for your org to make sure it’s able to be used by the AI effectively. We’ve got a a new metadata framework across all of the apps, including marketing, um, just to make sure that we can use AI effectively and all the apps can communicate effectively using the same or or the appropriate metadata framework here. And then, uh, all of our CRM apps are included under that. And on the right hand side, you’ve got, of course, Einstein, who is the the pioneer for our, uh, generative AI functionality here. At least that Salesforce uses, uh, Einstein. So we’ve got our Einstein trust layer, but that that’s being developed and deployed in order to make our data safe. We’ll talk a little bit more about the trust later in just a moment. But the Einstein one platform, um, it’s integrated. It’s intelligent. It’s automated, low code and no code, and it’s open. And that platform is what’s going to be the platform for Salesforce driving us forward using, uh, AI overall and specifically generative AI. So I dive into the Einstein Copilot studio because, uh, that’s the chatbot that’s going to be in Salesforce. It’s going to be the home for a lot of things that happen now, including, um, that first example that I showed you, creating that landing page, um, and so much more. And inside or as a part of Copilot Studio, there are these two core things that are going to be happening that we have to gain skills around. One of them is prompt builder. And I think, uh, what you’re gonna see here in just a moment is an example of prompt builder, what it looks like, and what it’s expected to look like inside. And prompting is is core to being able to use AI effectively, especially okay. Just for an example, talking about generating images. I’ll show you some examples in just a moment. The prompt is what results or is what gives you the results that you get from AI. And, uh, cultivating or tweaking or adjusting your prompt in different ways changes the result. And so you, as a human, you are sort of responsible for the results that come out of AI, although you can’t entirely control them. Again, you can get 80% there, uh, but your prompt is is the core, uh, is the core skill. Prompt engineering would be the core skill that you need to learn in order to get the most effective and best solutions out of AI. We’re gonna be using that skill set inside prompt builder to create reusable templates and connect our prompts into our org so that they could be used over and over again by users to get what they’re looking for. Um, the other thing is skill builder. So skills are going to be things that we create um, at sort of, like, tasks or actions that we create that our Copilot that chatbot can do behind the scenes for users at the click of a button or at the request of some text to do a specific thing. So we’re gonna have to learn the skills to use the skill builder to give AI the skills it needs to do what we want it to do inside the org. And this skills builder opens up the world. It or it opens up it opens you up to allow you to create, uh, you know, just, uh, taking a look at your own internal process and create things that Salesforce isn’t providing directly, but allows you to adjust and do things that, uh, that you need to do inside the chatbot with Skill Builder. So taking a look at prompt studio, here’s an example of what prompt studio looks like. And in this case, uh, in this image, we’re building an email template prompt. So and and we have email templates. Right? That’s normal. But in this case, what we’re doing is we’re asking AI, um, using a prompt to build up build us, uh, or create an email as a result of our new template type. And our new template type is more of a prompt style template, and we’re infusing or injecting, inserting the, uh, the data from the actual records themselves inside the prompt. And inside the prompt studio, well, you have the playground at the bottom. On the bottom left, you have, um, what whether calling a grounded prompt now. Once you add your data from your org into the prompt and you have your, um, your prompt, uh, created and you run it, your AI is gonna be grounded, uh, with the data inside your org. And you can, on the right hand side, generate your response so you can test your responses from your prompts. So as a prompt engineer working inside the prompt studio, you’ll be creating, uh, various prompts using the right, uh, data tags for those prompts and then testing those prompts to make sure that they’re working effectively. On the right hand side, you’ve got your configuration. You’re choosing language. You’re you’re choosing things like the style of the prompt. Um, you’re choosing the the sender type for the prompt, the recipient type, the length of the, uh, the the length of the generated response, and then you can choose different language model providers as well. And so the skill for learning prompt builder is going to be extremely important or prompt studio is gonna be extremely important. And, um, your the effectiveness of your users hinges upon your ability as an organization to go out and deploy prompts effectively. Talking about that Einstein trust layer. So when you do build a prompt in your prompt studio and you deploy that prompt and users are using that prompt, well, how does that work with the Einstein trust layer? So on the right on the top left hand side, you can see that a prompt gets generated. Uh, you can see that it goes through secure data retrieval, dynamic grounding, so that’s grounding that prompt with the information from your org. You’ve got data masking as well where you’re taking, um, personally identifiable information or sensitive information, and you’re masking that so that that doesn’t go to the language model provider. The language model provider has an agreement with zero retention. So you see the zero retention box there where the customer data is not going to be kept, um, when it when and if it does go to the language model provider. Of course, generation where AI creates the the response. In the case of our example here was that text from the email, and then that goes through the toxicity detection layer that’s looking for, um, toxic results. And then you’ve got an audit trail so that you can audit, you know, who’s creating prompts or who’s use utilizing prompts and the the prompt and the results from that prompt, and that gives your your, um, your back end users the ability or your front end users the ability to adjust prompts, update prompts, and just make sure that we’re getting the best results that we can from our prompts, and we can tweak and monitor those over time. And so that’s our trust layer overall at a very high level. It’s gonna be important to understand and know and be able to utilize the trust layer effectively, um, and use it to adjust your prompts as you go forward. So that was a lot. I know. Uh, there’s a lot going on. There’s a lot of changes that are happening on in in the Salesforce platform that are sort of driving the way the future of what Salesforce looks like and driving the way that we’ll be, uh, adjusting and learning how to do things within Salesforce. But I wanna talk to you a little bit about the actual tech itself, and I’ve put together a couple slides here to teach you just really high level briefly about generative AI and and try to help you understand the tech and help you understand why it works, what it’s doing. And this way, when you’re going out and you’re trying to learn how to use the, uh, you know, the Salesforce functionality like prompt studio, you have a sense of what’s going on behind the scenes. So AI is that umbrella term. Um, this is a machine that can think like a human performing tasks like reasoning, planning, learning, and decision making. Underneath the AI umbrella, that artificial intelligence umbrella, you’ve got machine learning, and this is, uh, the act of using data to create predictive models. Under the scope of machine learning, you have deep learning, which is what we call, uh, when you’re using neural nets with AI and they’re learning to perform a task that you’re not programming the AI to do. So you’re allowing the AI to learn how to do a task using neural nets. And, uh, finally, beneath the deep learning category, um, sub to that, you have what we all talk about, which is generative AI. That’s the big one. That’s the one that everyone’s, uh, super excited about in 2023 and beyond. So this is a, um, this is AIs that are creating, in general, creating content. And the content can be text, imagery, code, voice, videos, and more and so much more. So where does all this magic happen? How is it happening inside those neural nets that I talked about under the deep learning category for generative AI? You’ve got these nodes and these connections between the nodes that have these weights, and the nodes and weights and the neural nets, um, all go all all come together to form what we call the large language model, the LLM. That’s the architecture. And just on the left hand side, you see that server room running these language models. And language model providers like OpenAI are out there, uh, running these requests, the prompt requests, and the responses that come back from those, um, through computers all day long. And, uh, that that’s it. That’s where the magic happens. So it’s kind of a black box. You know? You just put some text in, some text comes out, but this is what’s going on behind the scenes. What does that process look like? You’ve got your input or your prompt on the left, and I’ll show you some example of that in just a moment. This can be in the form of text, images, videos. Now we can upload files. You can add code. All of that can be a part of your input or prompt. Um, for a while, this has all been very static. If you’re inputting text, you can only input text. If you’re inputting images, you can only input images. We’re getting to the point where we have this thing called multi modal where you can do multiple things at once and then run that through the LLM and get multiple outputs at once. So in the center, you’ve got the LLM prediction. Again, that picture of the neural net with the nodes and the weights, um, making decisions um, based on how it was trained. You’ve got two functions happening there. It’s input encoding and output decoding. These are the predictions, the output decoding, and the predictions lead to the output on the right hand side. Again, it can be text, images, videos, files, code, whatever your heart desires. Okay. The skills gap. So as you can see, a ton of new skills are, uh, are, um, sort of needed to tackle this new thing that we have called generative AI. 53% based on Salesforce’s research, of employer employees don’t know how to get the most value out of generative AI. 62% of workers say they don’t have the skills to effectively use it. And then 70% of business leaders don’t believe their teams have the right skills, and so that’s why I exist. I’m here to talk about and I’m always talking about the personal and the team actionable AI skills that you need to take advantage of what we have today. Now, um, you wanna learn fundamentals. You wanna learn actionable AI. So I’m gonna talk through these bubbles here really quickly. One of the, um, these are things that you can do to sort of stay ahead of the curve or stay knowledgeable about generative AI, what’s coming, what’s here, and what what to do. First thing is to keep up. Uh, you wanna stay up to date with Salesforce AI news and announcements as they roll out. That’s really hard to do even for someone who is that’s all I do. That’s all that’s my only focus. Um, just watching that session moments ago made me realize there’s some other generally available things for marketing AI that I need to learn about. So, uh, I post that as news on YouTube, so you can follow that, um, at AI Force Training on YouTube. Uh, learn fundamentals. That’s the next thing you need to do. You need to learn the generative AI fundamentals. Those are available via the Salesforce Trailheads. I’ve got a quick link right there for the ad that’s that’s a QR code, by the way, AI generate a QR code for the AI fundamentals, uh, trails. So I give my opinions there and help guide you to walk through those trails, um, so you can scan as well. And I’m gonna drop a link in the chat for all these as well. Professional training, you wanna use tools like ChatGPT and the tools that Salesforce develops professionally, obviously. Um, taking advantage of tools like ChatGPT for Work, um, actually helps you learn the skills that you need for prompt building, um, for Copilot Studio, for Copilot Skills Builder. And so we teach that at AI Force training, um, as a part of our professional training, uh, GPT specialist program. And then finally, implementing specialized solutions. So there are already tools today that you can use to implement specialized solutions for AI inside the org. There’s one specifically that we train on at AI Force training, so I have a link to that one there as well. But then also as Salesforce rolls out their tooling with Copilot Studio, we’ll be teaching that, and you’ll need to learn that as well. So let me just briefly over here, uh, see if I can drop all these links in the chat so everyone can take that as well. Okay. Hopefully, that made it to you. And I know I’ve been going really quickly here, so I haven’t been able to keep up with the chat. So if you have any questions, hope they’ve been answered so far as I go through this, but, hopefully, we’ll have some time at the end to do that. Uh, team company team or company AI strategy. So I talk about the AI strategy. Every company needs one. It’s super important. I’m not gonna go through all these bullet points in the screen. You can get the slides after or just take a quick screenshot. These are the types of things that you’re gonna wanna think about, um, as a team or as a company or even as an individual if you’re approaching your company and they don’t have a strategy yet. These are the kinds of questions you wanna look at and answer, um, to build your AI strategy. One of the quick hit ones that I will talk about is hidden use, that first one. What tools are everyone using? Everyone’s using AI in some kind of way or at least some people are already within the team, but not everyone’s talking about it. So, you know, as management, you wanna learn what your team is using, uh, cross, you know, cross check what’s working well, what’s not working well, what tools do we like, what tools do we trust, uh, what tools don’t work well, and, uh, that should become a part of your AI strategy. Okay. So some of the fun stuff here. Uh, so hopefully, you got that strategy slide and you can take that back to your team. Uh, thanks for the comment, Justin. This is amazing. Awesome. Well, I think you’re gonna think this is even more amazing. Think about the, uh, Northern Trailhead example I gave up top where we’re creating that landing page for images. Well, uh, you can already do this today using Dolly three inside of ChatGPT. So I’m gonna show you what it looks like to go through a little bit of a process of creating, uh, an image that you can use as an asset for your for your company, um, using Dolly three, which is available via ChatGPT. So check out this prompt. Uh, this is my prompt here, and then this is the response from DALL E 3. Create a photo photorealistic image of a female hiker in the mountains from the top of a hill. She’s facing away from the camera and and the rest of the prompt. And look how, uh, how realistic and wonderful and just appealing the images that it created, um, that match with my northern trail, you know, uh, gear, uh, if that’s what I’m selling, if that’s my company selling. And I can absolutely immediately use these images as a part of my asset library. Um, so the images have been generated. You get a couple different image options. Now a lot of you are saying, like, well, okay. Yeah. But I need something more specific. I’ve I’ve tried this maybe before, and I’ve gotta generate something very specific or something on brand. Because, by the way, this is just a very general example. It’s not on brand. Well, here’s a fun little example, and this will build a little bit. Here’s the AI Force logo, and, uh, this is what you do to get your company brand to come out of these images. You give it an image that you already use, uh, and you have GPT describe the image itself, and we ask for specific things. I like to ask for style, color profile, font descriptions, and colors, and you can ask for all different kinds of things. And it’s gonna output a text of, uh, of those things. So here now it’s in text describing these things. Once you kinda capture your company brand or your company’s image or your company, uh, uh, color profiles and whatnot, then you can feed that into the image generator. So now I’ve pasted that, um, all down here, color profile, typography, and I’ve asked to create an image of a three d design following the original logo as inspiration. So now I haven’t given it an image. I’ve only given it text, and it’s a description of the other image, and it’s created this sort of three d design for the logo that I gave it. Well, that’s cool. I like that. That’s logos. That’s nice. But I want an actual image. Well, the way that we just used our prompt engineering skills to create that design, here’s another example. Here’s a prompt. I just wanna generate an an image of a question mark, very straightforward, but I do wanna make it match the color style and theme of my company. So I’ve fed in RGB color set. Here are you know, the description of the fonts a little bit. Here’s a description of the style, and I’m trying to match my company’s brand. And I was able not AI four. It’s a different company, but I was able to do that by generating these images. And the coloring is right. The darkness in the background is right. The popping yellow is right. This all looks good. But what about the text? Oh, that’s awful. You’re next what? I don’t know. And ready for six that’s awful. Right? Well, Dolly three has actually gotten good at text. So guess what I do? I say spell the words correctly, and immediately, the images are regenerated with correctly spelled words. Who knew? Uh, again, images on brand. This is perfect. I love the branding. I love everything. But now I wanna get another version, and I want the top three questions because I’m gonna do a blog post about the top three questions for my company. And here we go. Top three questions just like that. Uh, I I actually use this. This is fantastic. Um, and I’ll I guess I should head back to the presentation now. That is an example of using Dolly three for images. You can use that today. You can just play into your heart’s content, but prompt engineering is key. Getting good prompts to get your images is what you need, um, what you need to do. So we talked about what’s available, what’s coming, uh, use DALL E three. We talked about AI strategy, skills takeaway. I hope I packed a punch just in time. Thank you so much everyone for attending. I think if we have a second here, we’ll, uh, we’ll dive into some q and a maybe. Maybe. I I wrapped up really quickly.

Speaker 0: Great session, Mark. Uh, I did see one question come up in the beginning. It said, uh, I love, uh, using cheat, uh, chat GPT, but I don’t know how to get started. And I I think you kind of covered that during the session, but that was what I saw on there.

Speaker 1: Yep. That’s right. Yep. We’ve got at Air Force, we’ve got