AICSS

ChatGPT in CS (AI + Customer Success Summit Session)

Bex Sekar
  -  
August 9, 2023
  -  
2 mins

ChatGPT in CS: A Practical Guide to Starting Your GenAI Journey

Discover the transformative power of ChatGPT in the world of Customer Success. This engaging and interactive session will open your eyes to the potential of ChatGPT to revolutionize your daily tasks, customer interactions, and career.

By attending this live session, you'll gain:

  • A solid understanding of ChatGPT and its potential to elevate your skills and experience
  • Practical knowledge on integrating ChatGPT into your work for improved efficiency and productivity
  • Expert insights on crafting effective prompts to get the most accurate and relevant responses from ChatGPT
  • A live demonstration of ChatGPT in action, showcasing real-world applications for customer success managers
  • Inspiration to explore the future of AI in customer success and the opportunity to become an early adopter in your organization

Speaker

Mickey Powell - Co-Founder, Head of GTM at UpdateAI

LinkedIn

Mickey has spent the last 10 years working for public and private SaaS companies across sales, operations, enablement, and Customer Success. He’s a dedicated lifelong learner and loves uncovering new ways to see and navigate the world with technology. He currently leads Go-to-market at UpdateAI, a ChatGPT powered tool to automate call notes and insights for Customer Success teams through their Zoom integration. When he’s not learning he’s spending time with his family in Orange County, CA.

Watch ChatGPT in CS: A Practical Guide to Starting Your GenAI Journey

‍Transcript

This transcript was created by AI. If you see any mistakes, please let us know at marketing@matik.io.

Matik MC: So, hello, everybody, welcome to Matik AI customer success summit. You are currently watching the ChatGPT and CS a practical guide to starting our AI journal. I'm excited to introduce our speaker. As I introduce him. I will have him answer our ice breaker question, which is what would you use AI to automate in your life outside of work? So, Mickey Powell, cofounder and head of go to market at UpdateAI. I'll take it over to you.

Mickey Powell: Thanks Pascal. So I thought a lot about this because, you know, there might be an age where a lot of things are automated. But if I could wave my magic wand and have one thing automated today, it would be folding my laundry. That is, that is not my favorite task. I have to find something good to watch while I'm folding my laundry to make it somewhat bearable.

Matik MC: Always always good to be multitasking. Always good to be multi tasking.

Mickey Powell: Just stress myself.

Matik MC: Exactly. Great. Well, before we start the session, just a reminder for all the participants. Please add any questions you have to the Q and a section which can be found on Zoom controls, and we'll say the last 10 to 15 minutes to answer those questions. And with that, Mickey, go ahead and take it away.

Mickey Powell: Awesome. Thanks so much Pascale and thank everybody for being here. I'm super excited because I am just a big fan of ChatGPT and generate AI and getting started with it is definitely a fun journey. And for those that are on that journey, I want you to know that by the end, I'm hoping you're going to be super inspired to go faster along that route. So today, what we're going to cover is high level. What is ChatGPT, I'm gonna talk about how it has this potential to elevate not only your skills but your experience. And I'm going to give you some expert insights on how to craft effective prompts because that's how you use this thing. And then I do want to give you some practical knowledge about integrating this more into your daily routine. And then of course, I'm gonna do a little bit of a demonstration on how to apply this in the world of customer success. And the reason I care about customer success is that's where I spent the last 10 years of my career, I've been a customer success manager, leader, CS operations, and now I'm at update AI and we build tools for customer success managers. And for those that I would like to connect, I'm all over LinkedIn, happy to. But we're actually going to start today with a little bit of an exercise. So I actually am gonna have all of you… help me come up in brainstorm with some problems. And actually, Pascale, I see that I need to get them this link. So I'm gonna send it to you if you could give it to all the folks that are here. But what I want you all to do when Pascal gives you this link is I want you to come into this document and I want you just anywhere in here, add problems that you face in customer success, any problem you can think of, just grab a cell, add a problem. And I'm gonna give you a roughly 60 seconds to do this. So starting now, just anything you come across like too many meetings or not enough time… or… bad hand off info, anything you can think of, just throw it in a cell anywhere. We have about 35 more seconds.

Mickey Powell: I know we all experience lots of problems. Don't be shy.

Mickey Powell: I'm gonna give you 15 more seconds.

Mickey Powell: All right. Five, four, three two one. All right. So in 60 seconds… you came up with 39 problems that customer success that we face in customer success? Not bad. Good job. So now I'm actually going to have ChatGPT, generate as many problems as it can in 60 seconds. So I'm gonna tell it and let me Zoom in for you here.

Mickey Powell: I'm gonna let it go for 60 seconds. So while it's working… I'm gonna grab the kind of doing this like a cooking show, put it in the oven, go do something else, come back, grab it out of the oven. So I'm gonna grab these. I'm gonna open up a new ChatGPT window and I'm gonna ask it, summarize… these problems. And I just pasted in all the problems that you all came up with in that 60 second window. And I wanted to do this to demonstrate to you one of the applications of ChatGPT. It's very good at taking a bunch of unstructured written data and structuring it together. And you can do all sorts of things with it after that fact. But if you're dealing with surveys, if you're dealing with meeting transcripts, if you're dealing with emails, you can start to change and transform and categorize all of these different pieces of inputs to make it more structured easier to understand. So let's see what ChatGPT did it's still working away, look at it. Go all these different problems. We'll say that 60 seconds… I'll let it keep working but we'll jump back in here. So now we have a bunch of problems that ChatGPT came up with a bunch of problems that you came up with and we could use that later for anything that we want. But the two takeaways I want you to take away from this is number one from now and probably forever into the future. Whenever you're brainstorming ChatGPT is your buddy because it can come up with as many ideas as you want it to. They won't all be great. But remember with brainstorming, you just want to get everything out because even if one of them is perfect, that's all that matters. So now we've got our problems out a little exercise to get us started. Let's jump in. So we just talked about problems. But I wanna show you what Bill gates. If anybody knows Bill gates, what he said about ChatGPT when he saw it, he said, hey, this thing, this is as important as the graphical is a user interface, which is another way of saying the personal computer that's how important he feels this technology is to us as a human race. So what is this thing? What is ChatGPT? So I'm just gonna do a quick overview and I'm using ChatGPT. This is an overview for more of large language models in general. But I use ChatGPT as the heuristic. For it. So the first is that ChatGPT is a large language model and it's trained on roughly one trillion words that they've pulled from the internet and they've added safety limits to make sure that it's not going to produce races the hateful or bigotry or all these other things as much as possible. So it's kept on rails courtesy of the folks who are open. I, but how does it work? So in my research, I found a really simple explanation from a gentleman by the name of Steven wolfram. He's a computer scientist, mathematician and he said it's just adding one word at a time. So when we have the whole corpus of the English language, if you mapped it in a three day space chat is just drawing this path through what is the most likely path through these words, connecting all these dots together to form what sounds like a coherent sentence… here's. An example, something that we're all kinds of used to the whole adding one word at a time, things because humans, we actually do this. So this sentence, I like peanut butter and you likely finish the sentence for me in your mind jelly. So this is what he does is actually just a very fancy version of this. There's a lot more under the hood. But this is a really good way of thinking about it but it has a bunch of key features. It has way more than just the ability to guess what the next word is. The first is that it has a natural language understanding. You don't have to be super specific and precise with it. You can talk to it the way you would talk to another person. The second is it has memory now every single time you talk to it, the memory is confined to that session, but it remembers what you're saying throughout that session. And that is really critical. So that if it doesn't forget as you go along and I'll dive into that a bit more in a second and then lastly, but certainly not least what it generates sounds very human like even though it's not human. So we're used to like cry and Alexa and things like that. This is a very different approach. So memory, I just mentioned it, but I want to delve in a little bit more. This is what I think makes chat really shine, and how this is different than what we've experienced before. So the memory is and contained to that one session that you're working with it. But how it works is it only remembers about 3,000 words backwards in the conversation. So as you're having a conversation with ChatGPT, whatever. Is at the top once it hits that roughly 3,000 word limit, it's gonna start forgetting. Now, you don't really notice this all this much because it's still guessing with a lot of accuracy based off of the context. But I just want you to imagine the memory and they call it the context window. It's shifting down over time. So if you have a very lengthy conversation, it'll start to forget stuff at the top. So you're having that back and forth with ChatGPT and it's able to remember. And this makes it a much more fluid and dynamic and easy to follow conversation. But there are definitely risks. So I don't want you to be full. I wanna call out some of these so that as you're using it, you can keep these in mind. There's this term hallucination, and really, that's a way of saying that ChatGPT can make stuff up and human scan too, right? We can lie. We can offsite all of those things. The difference with ChatGPT is it doesn't really know the difference. It will give you an answer and it'll be confidently incorrect. So when you see something, you have to, of course, take the approach of this doesn't know if it's lying or not lying. It doesn't have that sense. So we have to, as an outside observer, always keep that in mind. And that's why I'm going to remind you that you are responsible for the inputs and the outputs of this. So if you're using ChatGPT and if it matters for legal or safety or ethical reasons, then you should definitely consult a professional if you cannot independently verify it. And then of course, there isn't just ChatGPT. There's a lot going on. So it's like, hey, how do I choose? Could I go after this? That? What tool should I use that competition really is starting to heat up. And my guess is it's going to be white hot for a long time. And here's the top four that I'm gonna call out in this large language model space. So of course, we have ChatGPT. Plus this is what I use daily. And then we have being and being is actually just ChatGPT under the hood. They use ChatGPT and being. And then there's Google barred. It's not as good right now, but I would not count Google out for the future. And then there's a company called anthropics. They have a model called cloud and this a bunch of X, OpenAI employees and it's gaining a lot of ground. It's definitely good. And it has some other little advantages. But for now I think you're safe sticking with ChatGPT. Plus if you can afford the 20 dollars a month, I think it's well worth it. I don't get paid for saying that, but I do pay for it. I do find a lot of value in it… but I know that we all have jobs and lives and things and we have to ask. Okay, but look what is what's in this for me? Because I have, you know, needs, I have things that I want to accomplish. And because I come from customer success, I know all too well that you're already overwhelmed you've got so much on your plate between putting out fires and taking on new accounts and writing emails and creating documents. And the list just goes on forever. And with all of this overwhelmed, there wasn't really any sort of help in site. And then chat shows up. And now it's like my gosh this thing is going to be massive. And there is actually a study done by OpenAI in conjunction with the university of Pennsylvania and they estimate that with our current technology of ChatGPT, we can get 15 percent productivity improvements across the board. So we could all be 15 percent more productive today by using these tools. And as those tools evolve, they start to get baked into the softwares that we use daily. Well, they think there's going to be 47 56 or even more productivity improvement. So when I think about what more can I do with my day? How can I be more efficient? How can I get more done? And somebody saying that I could do 50 percent more? That is massive. Because what makes a great CSM is that you do a lot of things, you wear many hats, your problem solving, you're teaching others you're building processes and so much more. And the amazing part about ChatGPT is that it is designed to amplify you. It is so interwoven into what we bring to the table and helping us that this tool will work across so many different things. And, I'm gonna lay that out a little bit before we end today… taking on a new task. So let's say you're having to do something you haven't done before. And this is one of the reasons why I'm so bought into this as a technology of the future when you wanna learn about something that you haven't done before which we do a lot in our daily jobs, right? Hey, I'll do this or I'll do. That even though I've never done it before. The old way of doing it is you would go to Google, you would try to find somebody's knowledge around it articles. You're gonna try to learn it, internalize it, make sense of it, bring it into your context and then apply it. And that process takes a long time. Google made it faster, right? You don't have to go to the library. You can go to Google. But now ChatGPT actually flip this around. It allows you to go straight into application by starting to get the work done immediately because it can help generate. Google cannot generate. But PT can help you generate. And even if it's not perfect if it gets you 50 percent 80 percent of the way there, that's already faster. And you're learning along the way because you still have to verify what it's putting out. You have to make sure it makes sense. So I see this as a new paradigm of how we're going to work, which is we can get started and learn along the way much more effectively with this super tool that's guiding us. Put another way, I think this is like chat is a new on the job training. You can learn how to do things as you're doing them with this expert guidance. So how do start doing that? How do you actually start tackling all these different things that you have to tackle and get them done faster with ChatGPT? So it all starts with a prompt. So that's all prompt is just that's what it's called when you put something in the ChatGPT. And the tricky thing about prompts is that their part science and their part art. And I'm gonna do my best to lay out what we and the community know so far about how to create expert prompts and the importance of them is that what you get out of ChatGPT is in large part tied to what you put in. So bad prompts, they're going to be really generic. And this is one of the, this is one of the things people point to and say, yeah, you know, it's boring. What comes out of ChatGPT is boring. You're right? If you put in a boring prompt, you're going to get a boring answer, but good prompts, not only the creative and helpful, but they also inspire your next prompt. Just like when you're having a really amazing conversation with somebody and you just want to keep going and the conversation flows really naturally. That's what I experience when I'm having a great conversation with ChatGPT. So just in your mind really quick take a guess left or right, which is a good prompt, which is a bad prompt. I should be pretty explanatory and I'll break it down for you. Good prompt on the left, bad prompt on, the right one too generic. You're gonna get a, very generic answer. So let's break down how to craft a good prompt. So the first is the foundation, these are things that you need to keep in mind when you sit down to use these tools. And there's four of them, number one is start with the end in mind. If you don't know what you're trying to accomplish it's, going to be very hard for you to convey that to ChatGPT and guide it towards the right output. So make sure you know what you're trying to do. The second is to be specific and concise, it's not only going to help your thinking. It's also gonna help ChatGPT, thinking quote unquote so that you are guiding it directly and you're not confusing it. The third is to provide context. So what matters in this situation, what's relevant, add in that detail. So that ChatGPT, and you know, what is good and what isn't good for this. And lastly is to be direct these large language models. Don't currently have feelings as far as we know. So don't be afraid to just tell it. Hey, I like that. I don't like that. Do this. We don't do that usually with our peers or friends, family members, because sometimes it can be abrasive, or it can come off as harsh and we are anticipating the feelings of how we communicate with others, which I think is a good thing. But for ChatGPT, you can pair that back. You can be direct. You can get to your output much faster. So let me show you kind of a basic example of a prompt. So write an e-mail for an executive so that they understand the value of my product and the product is an automated note taker. So actually have that, let's see I have that loaded up here. So I'm gonna send, you can see what it generates. I don't know about you. But if I sent this e-mail to an executive, I can pretty much guarantee they're not gonna read it. So that's a bad example, but that is very basic. So basic in not so good out. So what can we do to fix that? And the analogy I like to use is kind of like a DJ table. So when you're sitting down to work with ChatGPT, there's all these little knobs. And levers and things you can do with language to get better outputs. So, I imagine like, hey, let's play with all these. And I'm gonna break down the eight that you can play with. So the eight components of great prompts, first four personas, context constraints and structure, the next four task formatting, tone and iteration. So breaking these down one by one personas. This one I think is the most interesting but it's not fully understood how this works but you can actually just tell ChatGPT who to be now in customer success, this is especially powerful. So if you're coming up with a plan for a customer, you can take that plan. You can throw it into ChatGPT, and you could say, hey, what would the VP think about this? What would they ask? If you're a VP? You know, what would you consider? Okay, now, if you're a director or if you're an individual contributor, or if you're a C suite, or if you're an engineer versus a marketer, you can use personas to generate to evaluate and to do so much more. But you can tell ChatGPT what to be. And this is an especially powerful way for you to change your perspective and get a different type of response than you could come up with on your own. Most likely, the second is context. Now, you're probably noticing this is in the foundation as well as in the components it's because it is,, so, so important, you do need to tell ChatGPT what it needs to know to make its output relevant and accurate. The third is constraints. So don't just tell ChatGPT what to do, also tell it what not to do. That actually enforces creativity. So, hey, like the funny example I use here is if you were a duck and you're trying to escape a bank vault and you can't use explosives, how would you do it? So, ChatGPT, if you tell it what not to do, it will figure out how to accomplish your task without doing the things that you've removed from the options. And by definition, that means it's more creative. And then of course, the structure, this is all about frameworks and templates and common ways of doing something. An example that we're all really used to is success plans. That is a framework that we use to make sure that we have, you know, what are the measures of success? What are we working on? Who cares about it? What is it gonna do for the business? You can tell chat, GB T to operate within a structure. So if you're going to build a success plan, not only could ChatGPT help you build, it could actually evaluate it. So you could say, hey, this is how we do success plans here's. What good looks like? How does mine look compared to that? Tell me what is missing? What I could work on? And it will know because you told it that, that's the structure you're looking for. These are everywhere by the way, like budgets to do lists. We use frameworks everywhere in our life lives. So leverage those common approaches and you're gonna get better outputs. Then of course, there's the task really simply like what is it doing? Like, what are you actually asking it to do? Just spell it out. Pro tip. I usually put the task at the end of my prompt so that it has like one last reminder of like this is what I'm expecting. And then there's tone like how we say what we say, that matters, what adjectives that can change the output. If you're speaking to an executive, maybe you need to be more direct and professional. If you're speaking to a friend, you're more friendly and personable and light, those are all adjustments you can give to ChatGPT to be more appropriate for the situation. And then there's formatting, you know, really simply good formatting can help make it easy to digest to follow. So you can do things like, hey, I want this in a table or I want it in bullet points or want it in mark down or I want headings, you can even ask it for emojis. I use it a lot to actually say, hey, I'm writing this, what's some emojis or visuals that will go along with this. So leverage it for better formatting, make the output even easier for you. And then last but certainly not least is iterations. My average conversation with chat is between five and seven messages long. So back and forth. And that's because I'm guiding it to what I want. So you should anticipate that you should be, testing. I want this, I'm telling it this, it's given me that and then, refining until you get to that output and you can even ask it to evaluate that conversation, give you a prompt. So at the end of a conversation, you can say, hey, all right. How did we get here? And how could I do this faster next time? What would be a good prompt that covers these different areas? And I think about this whole iteration thing kind of like curly. So the first prompt is when you throw it. But then you're like, brushing it. Along the way to get it where you want to go but it's really fast so it's totally worth the time and the effort. So when you put it all together, it's gonna look something like this. So we have all these different aspects. This example I talk about, hey, you're a strategic CSM on Salesforce and you're building this product optimization strategy. This is going to yield such a better result than what I showed you earlier with the basic example. And let me find it. Here we go. So I have it loaded up, ready to go. I'm gonna hit generate. So again, I give it some constraints. I don't want them to upsell on this product optimization strategy. I want it to be a proactive plan. I wanna use the pyramid principle which is from Barbara minto, she used, she was at Mckinsey. It's just a way of kind of can vying like this is the executive summary. This is the problem here's. The solutions here's, the evidence to make it really nice clean formatting, and very logically reason. So now it gives us this output. So, hey, here's, the situation here's, the complication here's. The question we wanna answer here's. What we believe the answer is. And then here's the breakdown of the plan. What are we gonna do? We're gonna do a pre product feature audit. We're gonna do some engagement analysis. We're going to gather feedback, educate and train users with personalized plans, resource sharing and ongoing support. And then we're gonna monitor measure and refine. So we need to define the KPI'S track them over time, make sure we're getting feedback loops for that qualitative feedback and being able to bring that into our process to further refine this. And then of course, regular reviews and then our conclusion. So back to the go from a basic example that is not the right fit, the right tone, not the right approach, just a little bit of extra effort to get something that's much closer to done. Yes, you'll tweak that you'll add some things. But the amount of time it takes to do that goes from an hour to hours down to 10 minutes. So we just went over a demo how to put this all together. The eight components. You are going to get a copy of the E or if you haven't already the E book that I wrote on this, there's more details in there as well. And I do wanna kinda give you a warning is that you don't need to over optimize on. Prompts, remember, chat is really fast. So take your first best shot if you don't like the output or if it's close and you want to guide it, then just keep going with the conversation and don't worry you're going to be a pro in no time. I only started playing with ChatGPT in January. So you're going to be there before you know it and probably surpassing me. So now, how do we start to integrate this into our day to day because they can do so many things… in the future? I do believe that it's gonna kinda be integrated everywhere and it's gonna always be helping us always be anticipating, and assisting us with different types of tasks. But for now, you do have to sit down and use it and kind of know what to do with it. And I'm gonna give you an example and this is going to be a lot. So I know, take this all in. I want you to imagine that. I'm trying to build a webinar on communications tactics. I know a little bit about this from my career and having taught stuff like this before, but there's just so many different things that have to be done. This is a big project hours and hours to get this done. And if I were to use ChatGPT across all these different things, I would ask it a bunch of different things to help me. So I would want it to summarize some books, some like expert material. I would want it to combine maybe my notes, what I've already written down about what I know with that material. And then I would ask you to find common themes and patterns. Okay? You know, where is the content all starting to join together to make, you know, the most expert advice in one place. And then of course, I want customer success related examples and stories. So I would have chat help me generate those. So I can take this content and put it in the context of customer success. And then we need learning objectives. What are people going to get out of this webinar? It's very good at writing learning objectives and it will help me speed up that process. So so much. And then of course, things like creating an outline asking it, am I missing something? Is this easy? You know, hey, what might I need to add to this to make it good? And then we need a quiz. We need exercises, things to go along with it. And then we need to create slides with an agenda and content and a recap, and we need visuals for all of those. So there is a lot of information on this slide, the takeaway I want you. To take from this is, these are all independent things that ChatGPT can do. It's. Also things that can do in one session and it can save you time across all of these. So if you see this like orange and blue, this is to try to show you the like, hey, before ChatGPT, I would roughly take me this much time per task. And after chat GPT, you know, using it as a tool, it reduces that and it's reducing it by a huge margin. So it's not finishing everything for me. I still have to add my human intelligence but it is speeding up so much of this process. And I wanna kinda show you what happens when we put this essentially big prompt because I'm asking you to do a lot of different things. So I told that, hey, you're an expert instructional designer. You're coming up with a webinar on communication tactics and I want you to do all of these things. And I'm just gonna kinda let it go and start to generate. And like I said, it's not going to be done. It's not going to be perfect, but it is going to get me so much further so much faster than I could ever do on my own. And you could do this for your customers. You could take your own content. So let's say you have a customer that really cares about something in particular. You could go grab some of your support articles and you can put it together for them in a customized way asking chat to help you. So let's see how it did summarizing insight, I books influence never split the difference, how to win friends, crucial conversations, effort with experience. I've read three of these. I haven't read four and five, but ever split the difference is probably my favorite book on communication. So for anybody that hasn't read it, I highly recommended. I've actually read it twice common themes and patterns, understanding human behavior, effective communication, building relationships, ethical persuasion and reducing effort in portion. I think this is interesting because this is a lot of what customer success has to be good at. I know you folks probably are some examples and stories. These are great. An anecdote about CSM who use negotiation techniques for never split the difference. So it gives me like an example of what I would generate and I would actually ask it to go further on this like give me more specifics, what's the scenario, who, what on way, where here's the learning objectives. This is what the folks are gonna take away from the webinar. And then we've got our outline. This is high level. I would ask it to go deeper on all of these but this is a great starting point. Again, this would take me 10 minutes on my own, maybe 30 or an hour and it does it for me in seconds. And then, hey, we've got some assessments ease of understanding case studies. It actually said, hey, we're missing some things probably need case studies, interactive components and a feedback mechanism. Great. Thank you for calling that out here's. Some quick questions… exercise… and then our slide outline with suggested visuals. You can see it actually kinda skipped over here because it's like sometimes chat likes to kinda be almost lazy. It's like, hey, are you sure you want me to generate all of that? Again? I know this is a lot I'm throwing you all at once. The takeaway I want you to get out of, this is a very broad tool. You have a very broad job and what you do in customer success, this can help you in so many different ways. You just got to start playing with it. You've got to learn the underlying prompting and from there, this is the limit. So we just went over that demo. It saves you a ton of time. So maybe six hours down to less than two to get this ready. Maybe even more… and building that ChatGPT habit that's the name of the game here and there really simple ways to get started. If it's something important or annoying, you should try chat GPT or if it's gonna take you longer than five minutes to work on it, you should try ChatGPT. And then perfect is the enemy of good. Just get started. ChatGPT is so fast. I procrastinate like crazy if I'm procrastinating, I use ChatGPT to just get me started even if it's not good. The momentum helps me get motivation and then give yourself, you know, space to explore. It's a fun new tool. So go where your curiosity takes you. It's gonna be much more pleasant experience. And then lastly have fun. This is a once in a lifetime technology try, to get the most out of it. I'll give you an example. I've got a daughter turning five next week. We'd like to write stories together using ChatGPT. She really loves it because she gets to participate in the creation of the stories. All right. I know that was a lot. Thank you all for hanging in there. So let's recap. We talked about what ChatGPT is Lawrence language model and of course, how to elevate your knowledge while also just getting things done. There's so much you. Can do to help you how you can craft effective prompts using those eight components. We did a live demonstration across a couple of different instances. And then I'm hoping you leave today with inspiration on the future of AI and customer success. So with that, we can jump into the Q and a. And I hope of course, to connect with all of you in the near future.

Matik MC: Awesome. Mike. Fantastic presentation. Thank you so much for putting that together. I know I learned a ton. We had somebody in the Q and a actually say, you know, hey, I really wanted to say, I appreciate the content and the presentation style. So great work there. I got, some great ideas already on stuff that we're working here at Matik, but yeah, let's jump into the Q and a. So the first question, is there a risk in asking ChatGPT to summarize data that has confidential company information?

Mickey Powell: Though there is, and there always is just like there's risk in putting it into a Google doc or having it on a computer that is in air gap. For instance, there is always risk however ChatGPT, OpenAI rather has been very open about their data use policies and they have, and if you actually go to their data use policies like OpenAI dot com data use. The first thing they put up there is that they don't use your data just to train their models, unless you opt in, which is important, right? Opt in versus opt out. Now that's all to be said that if you're concerned or if your company has given you specific guidelines not to do that, then I wouldn't for the time being.

Matik MC: Fair, right. Good point. Another question. How do you make ChatGPT change their tone to sound more human, but still professional? I feel like it's got two modes, super professional robotic and to like slang every time I ask about to sound more friendly, the tone of way to informal?

Mickey Powell: Yeah, this is a great question. Number one, of course, you can tell it. You can give it guidance specifically on like no too formal. Two informal, be more like this, be more like that in the ebook that I put out. There is, an example I used of actually reverse engineering your writing style so you can actually give it a sample, you can have it evaluate that sample and give you a style guide of your writing style. So that next time you ask it to write something, you can say, write it in this style and it will sound more like you. You can actually get that copy of the book for free at ChatGPT update a.

Matik MC: And I've got a follow on question on that, like, so your style, right? You, you send that information, it learns. But you said that it also has like that 300 word limit. So, like will it forget your style eventually and you have to redo your style or how does that work?

Mickey Powell: Yeah. So OpenAI and this actually kind of dub tails into the next question of like, is the plus really worth it? If you have the plus plan, you can actually do custom instructions, which means that every time you use, it doesn't remember like the whole model doesn't remember, but it will bring in custom instructions to that individual session. So you can give it custom instruction to say this is how I prefer to write. This is what I care about. This is you know, the things that I work on and it will use that as it generates. So I suggest you know, generating the style guide of your style to keep to have and to use later but also to use ChatGPT plus give it those custom instructions and it, you'll get a better output. That's one of the reasons why. And I'm just gonna kinda roll into the next question. The plus is really worth it. You get access to plugins, you get access to GP T4, which is a much better model and one of the plugins I use daily that, you know, we just didn't have enough time to demonstrate here. It's called code interpreter and it can do basic and advanced to statistical analysis for you really fast. So I whole heartily think plus is worth it if you can afford the, if you can afford the 20 dollars a month. And if you can't if you go to ChatGPT being and use creative mode, that is GP T4 underneath. So you can use that for free. You just don't get access to like the plugins and things like that. Thanks Dan. I appreciate the kind words Sara, significant improvements in GP four. Yes. GB T4 is significantly better. I almost never use three point five anymore like maybe once a month, if I just really need something, really, fast like that's. The only difference is three point five is much faster than four in generating, but the quality is significantly better on four… Jonathan ask, have I seen CSM use chat agents and our plugins for their work? Not a lot yet. And that's mainly because CSM deal with a lot of very contextual issues and problems and challenges. So the context really matters now. I will say that update AI is an example of using chat within the workflow of a CSM. So update AI, for instance, when you have a call, we take that transcript, we use chat to generate your meeting notes, action items, what risks have come up? What type of risk they are, product feedback, what type of product feedback it is as well as important numbers. And then we also use it to draft your. Follow up e-mail after the meeting and prep you for the next day's. Worth the meetings. So that's where I kind of see, the market going right now is more built into your workflow as opposed to pushing you out of it. And that's what we are doing it at. Update. I… see car says, I'm not sure I'm comfortable with presentation, teaching us as chat to be for some reason books especially there's open litigation alleging chat there have pulled pulling data from books without compensating or telling the authors care brings up a good point. There is, there is an ethical question about, you know, what's and potentially a lawful question about what's published on the internet. Was it trained? Was it lawful? We are going to learn the answer to those questions over time as lawsuits play out in the courts and figure out, you know, does the law believe it was lawful? So, I have my own opinions, but I think we have to wait and see how that plays out.

Matik MC: Awesome. I think those are all the questions. Great questions. Thank you for saying that everybody. And yeah, I just wanted to give a big thanks to Mike, for taking his time presenting this, and sharing all this information in with us. If there are any further questions that come in… they will be posted on our hub that we sent out after the session with a bunch more information. But yeah, just thank you everybody for attending and thank you Nikki for your time.

Mickey Powell: Of course, thank you all. Have a great day.

Matik MC: Cheers, bye.

 

---

Click here to see all sessions & speakers, and navigate to recordings.

Join Our Newsletter

Don’t miss our next blog post & other great content!

ChatGPT in CS: A Practical Guide to Starting Your GenAI Journey

Discover the transformative power of ChatGPT in the world of Customer Success. This engaging and interactive session will open your eyes to the potential of ChatGPT to revolutionize your daily tasks, customer interactions, and career.

By attending this live session, you'll gain:

  • A solid understanding of ChatGPT and its potential to elevate your skills and experience
  • Practical knowledge on integrating ChatGPT into your work for improved efficiency and productivity
  • Expert insights on crafting effective prompts to get the most accurate and relevant responses from ChatGPT
  • A live demonstration of ChatGPT in action, showcasing real-world applications for customer success managers
  • Inspiration to explore the future of AI in customer success and the opportunity to become an early adopter in your organization

Speaker

Mickey Powell - Co-Founder, Head of GTM at UpdateAI

LinkedIn

Mickey has spent the last 10 years working for public and private SaaS companies across sales, operations, enablement, and Customer Success. He’s a dedicated lifelong learner and loves uncovering new ways to see and navigate the world with technology. He currently leads Go-to-market at UpdateAI, a ChatGPT powered tool to automate call notes and insights for Customer Success teams through their Zoom integration. When he’s not learning he’s spending time with his family in Orange County, CA.

Watch ChatGPT in CS: A Practical Guide to Starting Your GenAI Journey

‍Transcript

This transcript was created by AI. If you see any mistakes, please let us know at marketing@matik.io.

Matik MC: So, hello, everybody, welcome to Matik AI customer success summit. You are currently watching the ChatGPT and CS a practical guide to starting our AI journal. I'm excited to introduce our speaker. As I introduce him. I will have him answer our ice breaker question, which is what would you use AI to automate in your life outside of work? So, Mickey Powell, cofounder and head of go to market at UpdateAI. I'll take it over to you.

Mickey Powell: Thanks Pascal. So I thought a lot about this because, you know, there might be an age where a lot of things are automated. But if I could wave my magic wand and have one thing automated today, it would be folding my laundry. That is, that is not my favorite task. I have to find something good to watch while I'm folding my laundry to make it somewhat bearable.

Matik MC: Always always good to be multitasking. Always good to be multi tasking.

Mickey Powell: Just stress myself.

Matik MC: Exactly. Great. Well, before we start the session, just a reminder for all the participants. Please add any questions you have to the Q and a section which can be found on Zoom controls, and we'll say the last 10 to 15 minutes to answer those questions. And with that, Mickey, go ahead and take it away.

Mickey Powell: Awesome. Thanks so much Pascale and thank everybody for being here. I'm super excited because I am just a big fan of ChatGPT and generate AI and getting started with it is definitely a fun journey. And for those that are on that journey, I want you to know that by the end, I'm hoping you're going to be super inspired to go faster along that route. So today, what we're going to cover is high level. What is ChatGPT, I'm gonna talk about how it has this potential to elevate not only your skills but your experience. And I'm going to give you some expert insights on how to craft effective prompts because that's how you use this thing. And then I do want to give you some practical knowledge about integrating this more into your daily routine. And then of course, I'm gonna do a little bit of a demonstration on how to apply this in the world of customer success. And the reason I care about customer success is that's where I spent the last 10 years of my career, I've been a customer success manager, leader, CS operations, and now I'm at update AI and we build tools for customer success managers. And for those that I would like to connect, I'm all over LinkedIn, happy to. But we're actually going to start today with a little bit of an exercise. So I actually am gonna have all of you… help me come up in brainstorm with some problems. And actually, Pascale, I see that I need to get them this link. So I'm gonna send it to you if you could give it to all the folks that are here. But what I want you all to do when Pascal gives you this link is I want you to come into this document and I want you just anywhere in here, add problems that you face in customer success, any problem you can think of, just grab a cell, add a problem. And I'm gonna give you a roughly 60 seconds to do this. So starting now, just anything you come across like too many meetings or not enough time… or… bad hand off info, anything you can think of, just throw it in a cell anywhere. We have about 35 more seconds.

Mickey Powell: I know we all experience lots of problems. Don't be shy.

Mickey Powell: I'm gonna give you 15 more seconds.

Mickey Powell: All right. Five, four, three two one. All right. So in 60 seconds… you came up with 39 problems that customer success that we face in customer success? Not bad. Good job. So now I'm actually going to have ChatGPT, generate as many problems as it can in 60 seconds. So I'm gonna tell it and let me Zoom in for you here.

Mickey Powell: I'm gonna let it go for 60 seconds. So while it's working… I'm gonna grab the kind of doing this like a cooking show, put it in the oven, go do something else, come back, grab it out of the oven. So I'm gonna grab these. I'm gonna open up a new ChatGPT window and I'm gonna ask it, summarize… these problems. And I just pasted in all the problems that you all came up with in that 60 second window. And I wanted to do this to demonstrate to you one of the applications of ChatGPT. It's very good at taking a bunch of unstructured written data and structuring it together. And you can do all sorts of things with it after that fact. But if you're dealing with surveys, if you're dealing with meeting transcripts, if you're dealing with emails, you can start to change and transform and categorize all of these different pieces of inputs to make it more structured easier to understand. So let's see what ChatGPT did it's still working away, look at it. Go all these different problems. We'll say that 60 seconds… I'll let it keep working but we'll jump back in here. So now we have a bunch of problems that ChatGPT came up with a bunch of problems that you came up with and we could use that later for anything that we want. But the two takeaways I want you to take away from this is number one from now and probably forever into the future. Whenever you're brainstorming ChatGPT is your buddy because it can come up with as many ideas as you want it to. They won't all be great. But remember with brainstorming, you just want to get everything out because even if one of them is perfect, that's all that matters. So now we've got our problems out a little exercise to get us started. Let's jump in. So we just talked about problems. But I wanna show you what Bill gates. If anybody knows Bill gates, what he said about ChatGPT when he saw it, he said, hey, this thing, this is as important as the graphical is a user interface, which is another way of saying the personal computer that's how important he feels this technology is to us as a human race. So what is this thing? What is ChatGPT? So I'm just gonna do a quick overview and I'm using ChatGPT. This is an overview for more of large language models in general. But I use ChatGPT as the heuristic. For it. So the first is that ChatGPT is a large language model and it's trained on roughly one trillion words that they've pulled from the internet and they've added safety limits to make sure that it's not going to produce races the hateful or bigotry or all these other things as much as possible. So it's kept on rails courtesy of the folks who are open. I, but how does it work? So in my research, I found a really simple explanation from a gentleman by the name of Steven wolfram. He's a computer scientist, mathematician and he said it's just adding one word at a time. So when we have the whole corpus of the English language, if you mapped it in a three day space chat is just drawing this path through what is the most likely path through these words, connecting all these dots together to form what sounds like a coherent sentence… here's. An example, something that we're all kinds of used to the whole adding one word at a time, things because humans, we actually do this. So this sentence, I like peanut butter and you likely finish the sentence for me in your mind jelly. So this is what he does is actually just a very fancy version of this. There's a lot more under the hood. But this is a really good way of thinking about it but it has a bunch of key features. It has way more than just the ability to guess what the next word is. The first is that it has a natural language understanding. You don't have to be super specific and precise with it. You can talk to it the way you would talk to another person. The second is it has memory now every single time you talk to it, the memory is confined to that session, but it remembers what you're saying throughout that session. And that is really critical. So that if it doesn't forget as you go along and I'll dive into that a bit more in a second and then lastly, but certainly not least what it generates sounds very human like even though it's not human. So we're used to like cry and Alexa and things like that. This is a very different approach. So memory, I just mentioned it, but I want to delve in a little bit more. This is what I think makes chat really shine, and how this is different than what we've experienced before. So the memory is and contained to that one session that you're working with it. But how it works is it only remembers about 3,000 words backwards in the conversation. So as you're having a conversation with ChatGPT, whatever. Is at the top once it hits that roughly 3,000 word limit, it's gonna start forgetting. Now, you don't really notice this all this much because it's still guessing with a lot of accuracy based off of the context. But I just want you to imagine the memory and they call it the context window. It's shifting down over time. So if you have a very lengthy conversation, it'll start to forget stuff at the top. So you're having that back and forth with ChatGPT and it's able to remember. And this makes it a much more fluid and dynamic and easy to follow conversation. But there are definitely risks. So I don't want you to be full. I wanna call out some of these so that as you're using it, you can keep these in mind. There's this term hallucination, and really, that's a way of saying that ChatGPT can make stuff up and human scan too, right? We can lie. We can offsite all of those things. The difference with ChatGPT is it doesn't really know the difference. It will give you an answer and it'll be confidently incorrect. So when you see something, you have to, of course, take the approach of this doesn't know if it's lying or not lying. It doesn't have that sense. So we have to, as an outside observer, always keep that in mind. And that's why I'm going to remind you that you are responsible for the inputs and the outputs of this. So if you're using ChatGPT and if it matters for legal or safety or ethical reasons, then you should definitely consult a professional if you cannot independently verify it. And then of course, there isn't just ChatGPT. There's a lot going on. So it's like, hey, how do I choose? Could I go after this? That? What tool should I use that competition really is starting to heat up. And my guess is it's going to be white hot for a long time. And here's the top four that I'm gonna call out in this large language model space. So of course, we have ChatGPT. Plus this is what I use daily. And then we have being and being is actually just ChatGPT under the hood. They use ChatGPT and being. And then there's Google barred. It's not as good right now, but I would not count Google out for the future. And then there's a company called anthropics. They have a model called cloud and this a bunch of X, OpenAI employees and it's gaining a lot of ground. It's definitely good. And it has some other little advantages. But for now I think you're safe sticking with ChatGPT. Plus if you can afford the 20 dollars a month, I think it's well worth it. I don't get paid for saying that, but I do pay for it. I do find a lot of value in it… but I know that we all have jobs and lives and things and we have to ask. Okay, but look what is what's in this for me? Because I have, you know, needs, I have things that I want to accomplish. And because I come from customer success, I know all too well that you're already overwhelmed you've got so much on your plate between putting out fires and taking on new accounts and writing emails and creating documents. And the list just goes on forever. And with all of this overwhelmed, there wasn't really any sort of help in site. And then chat shows up. And now it's like my gosh this thing is going to be massive. And there is actually a study done by OpenAI in conjunction with the university of Pennsylvania and they estimate that with our current technology of ChatGPT, we can get 15 percent productivity improvements across the board. So we could all be 15 percent more productive today by using these tools. And as those tools evolve, they start to get baked into the softwares that we use daily. Well, they think there's going to be 47 56 or even more productivity improvement. So when I think about what more can I do with my day? How can I be more efficient? How can I get more done? And somebody saying that I could do 50 percent more? That is massive. Because what makes a great CSM is that you do a lot of things, you wear many hats, your problem solving, you're teaching others you're building processes and so much more. And the amazing part about ChatGPT is that it is designed to amplify you. It is so interwoven into what we bring to the table and helping us that this tool will work across so many different things. And, I'm gonna lay that out a little bit before we end today… taking on a new task. So let's say you're having to do something you haven't done before. And this is one of the reasons why I'm so bought into this as a technology of the future when you wanna learn about something that you haven't done before which we do a lot in our daily jobs, right? Hey, I'll do this or I'll do. That even though I've never done it before. The old way of doing it is you would go to Google, you would try to find somebody's knowledge around it articles. You're gonna try to learn it, internalize it, make sense of it, bring it into your context and then apply it. And that process takes a long time. Google made it faster, right? You don't have to go to the library. You can go to Google. But now ChatGPT actually flip this around. It allows you to go straight into application by starting to get the work done immediately because it can help generate. Google cannot generate. But PT can help you generate. And even if it's not perfect if it gets you 50 percent 80 percent of the way there, that's already faster. And you're learning along the way because you still have to verify what it's putting out. You have to make sure it makes sense. So I see this as a new paradigm of how we're going to work, which is we can get started and learn along the way much more effectively with this super tool that's guiding us. Put another way, I think this is like chat is a new on the job training. You can learn how to do things as you're doing them with this expert guidance. So how do start doing that? How do you actually start tackling all these different things that you have to tackle and get them done faster with ChatGPT? So it all starts with a prompt. So that's all prompt is just that's what it's called when you put something in the ChatGPT. And the tricky thing about prompts is that their part science and their part art. And I'm gonna do my best to lay out what we and the community know so far about how to create expert prompts and the importance of them is that what you get out of ChatGPT is in large part tied to what you put in. So bad prompts, they're going to be really generic. And this is one of the, this is one of the things people point to and say, yeah, you know, it's boring. What comes out of ChatGPT is boring. You're right? If you put in a boring prompt, you're going to get a boring answer, but good prompts, not only the creative and helpful, but they also inspire your next prompt. Just like when you're having a really amazing conversation with somebody and you just want to keep going and the conversation flows really naturally. That's what I experience when I'm having a great conversation with ChatGPT. So just in your mind really quick take a guess left or right, which is a good prompt, which is a bad prompt. I should be pretty explanatory and I'll break it down for you. Good prompt on the left, bad prompt on, the right one too generic. You're gonna get a, very generic answer. So let's break down how to craft a good prompt. So the first is the foundation, these are things that you need to keep in mind when you sit down to use these tools. And there's four of them, number one is start with the end in mind. If you don't know what you're trying to accomplish it's, going to be very hard for you to convey that to ChatGPT and guide it towards the right output. So make sure you know what you're trying to do. The second is to be specific and concise, it's not only going to help your thinking. It's also gonna help ChatGPT, thinking quote unquote so that you are guiding it directly and you're not confusing it. The third is to provide context. So what matters in this situation, what's relevant, add in that detail. So that ChatGPT, and you know, what is good and what isn't good for this. And lastly is to be direct these large language models. Don't currently have feelings as far as we know. So don't be afraid to just tell it. Hey, I like that. I don't like that. Do this. We don't do that usually with our peers or friends, family members, because sometimes it can be abrasive, or it can come off as harsh and we are anticipating the feelings of how we communicate with others, which I think is a good thing. But for ChatGPT, you can pair that back. You can be direct. You can get to your output much faster. So let me show you kind of a basic example of a prompt. So write an e-mail for an executive so that they understand the value of my product and the product is an automated note taker. So actually have that, let's see I have that loaded up here. So I'm gonna send, you can see what it generates. I don't know about you. But if I sent this e-mail to an executive, I can pretty much guarantee they're not gonna read it. So that's a bad example, but that is very basic. So basic in not so good out. So what can we do to fix that? And the analogy I like to use is kind of like a DJ table. So when you're sitting down to work with ChatGPT, there's all these little knobs. And levers and things you can do with language to get better outputs. So, I imagine like, hey, let's play with all these. And I'm gonna break down the eight that you can play with. So the eight components of great prompts, first four personas, context constraints and structure, the next four task formatting, tone and iteration. So breaking these down one by one personas. This one I think is the most interesting but it's not fully understood how this works but you can actually just tell ChatGPT who to be now in customer success, this is especially powerful. So if you're coming up with a plan for a customer, you can take that plan. You can throw it into ChatGPT, and you could say, hey, what would the VP think about this? What would they ask? If you're a VP? You know, what would you consider? Okay, now, if you're a director or if you're an individual contributor, or if you're a C suite, or if you're an engineer versus a marketer, you can use personas to generate to evaluate and to do so much more. But you can tell ChatGPT what to be. And this is an especially powerful way for you to change your perspective and get a different type of response than you could come up with on your own. Most likely, the second is context. Now, you're probably noticing this is in the foundation as well as in the components it's because it is,, so, so important, you do need to tell ChatGPT what it needs to know to make its output relevant and accurate. The third is constraints. So don't just tell ChatGPT what to do, also tell it what not to do. That actually enforces creativity. So, hey, like the funny example I use here is if you were a duck and you're trying to escape a bank vault and you can't use explosives, how would you do it? So, ChatGPT, if you tell it what not to do, it will figure out how to accomplish your task without doing the things that you've removed from the options. And by definition, that means it's more creative. And then of course, the structure, this is all about frameworks and templates and common ways of doing something. An example that we're all really used to is success plans. That is a framework that we use to make sure that we have, you know, what are the measures of success? What are we working on? Who cares about it? What is it gonna do for the business? You can tell chat, GB T to operate within a structure. So if you're going to build a success plan, not only could ChatGPT help you build, it could actually evaluate it. So you could say, hey, this is how we do success plans here's. What good looks like? How does mine look compared to that? Tell me what is missing? What I could work on? And it will know because you told it that, that's the structure you're looking for. These are everywhere by the way, like budgets to do lists. We use frameworks everywhere in our life lives. So leverage those common approaches and you're gonna get better outputs. Then of course, there's the task really simply like what is it doing? Like, what are you actually asking it to do? Just spell it out. Pro tip. I usually put the task at the end of my prompt so that it has like one last reminder of like this is what I'm expecting. And then there's tone like how we say what we say, that matters, what adjectives that can change the output. If you're speaking to an executive, maybe you need to be more direct and professional. If you're speaking to a friend, you're more friendly and personable and light, those are all adjustments you can give to ChatGPT to be more appropriate for the situation. And then there's formatting, you know, really simply good formatting can help make it easy to digest to follow. So you can do things like, hey, I want this in a table or I want it in bullet points or want it in mark down or I want headings, you can even ask it for emojis. I use it a lot to actually say, hey, I'm writing this, what's some emojis or visuals that will go along with this. So leverage it for better formatting, make the output even easier for you. And then last but certainly not least is iterations. My average conversation with chat is between five and seven messages long. So back and forth. And that's because I'm guiding it to what I want. So you should anticipate that you should be, testing. I want this, I'm telling it this, it's given me that and then, refining until you get to that output and you can even ask it to evaluate that conversation, give you a prompt. So at the end of a conversation, you can say, hey, all right. How did we get here? And how could I do this faster next time? What would be a good prompt that covers these different areas? And I think about this whole iteration thing kind of like curly. So the first prompt is when you throw it. But then you're like, brushing it. Along the way to get it where you want to go but it's really fast so it's totally worth the time and the effort. So when you put it all together, it's gonna look something like this. So we have all these different aspects. This example I talk about, hey, you're a strategic CSM on Salesforce and you're building this product optimization strategy. This is going to yield such a better result than what I showed you earlier with the basic example. And let me find it. Here we go. So I have it loaded up, ready to go. I'm gonna hit generate. So again, I give it some constraints. I don't want them to upsell on this product optimization strategy. I want it to be a proactive plan. I wanna use the pyramid principle which is from Barbara minto, she used, she was at Mckinsey. It's just a way of kind of can vying like this is the executive summary. This is the problem here's. The solutions here's, the evidence to make it really nice clean formatting, and very logically reason. So now it gives us this output. So, hey, here's, the situation here's, the complication here's. The question we wanna answer here's. What we believe the answer is. And then here's the breakdown of the plan. What are we gonna do? We're gonna do a pre product feature audit. We're gonna do some engagement analysis. We're going to gather feedback, educate and train users with personalized plans, resource sharing and ongoing support. And then we're gonna monitor measure and refine. So we need to define the KPI'S track them over time, make sure we're getting feedback loops for that qualitative feedback and being able to bring that into our process to further refine this. And then of course, regular reviews and then our conclusion. So back to the go from a basic example that is not the right fit, the right tone, not the right approach, just a little bit of extra effort to get something that's much closer to done. Yes, you'll tweak that you'll add some things. But the amount of time it takes to do that goes from an hour to hours down to 10 minutes. So we just went over a demo how to put this all together. The eight components. You are going to get a copy of the E or if you haven't already the E book that I wrote on this, there's more details in there as well. And I do wanna kinda give you a warning is that you don't need to over optimize on. Prompts, remember, chat is really fast. So take your first best shot if you don't like the output or if it's close and you want to guide it, then just keep going with the conversation and don't worry you're going to be a pro in no time. I only started playing with ChatGPT in January. So you're going to be there before you know it and probably surpassing me. So now, how do we start to integrate this into our day to day because they can do so many things… in the future? I do believe that it's gonna kinda be integrated everywhere and it's gonna always be helping us always be anticipating, and assisting us with different types of tasks. But for now, you do have to sit down and use it and kind of know what to do with it. And I'm gonna give you an example and this is going to be a lot. So I know, take this all in. I want you to imagine that. I'm trying to build a webinar on communications tactics. I know a little bit about this from my career and having taught stuff like this before, but there's just so many different things that have to be done. This is a big project hours and hours to get this done. And if I were to use ChatGPT across all these different things, I would ask it a bunch of different things to help me. So I would want it to summarize some books, some like expert material. I would want it to combine maybe my notes, what I've already written down about what I know with that material. And then I would ask you to find common themes and patterns. Okay? You know, where is the content all starting to join together to make, you know, the most expert advice in one place. And then of course, I want customer success related examples and stories. So I would have chat help me generate those. So I can take this content and put it in the context of customer success. And then we need learning objectives. What are people going to get out of this webinar? It's very good at writing learning objectives and it will help me speed up that process. So so much. And then of course, things like creating an outline asking it, am I missing something? Is this easy? You know, hey, what might I need to add to this to make it good? And then we need a quiz. We need exercises, things to go along with it. And then we need to create slides with an agenda and content and a recap, and we need visuals for all of those. So there is a lot of information on this slide, the takeaway I want you. To take from this is, these are all independent things that ChatGPT can do. It's. Also things that can do in one session and it can save you time across all of these. So if you see this like orange and blue, this is to try to show you the like, hey, before ChatGPT, I would roughly take me this much time per task. And after chat GPT, you know, using it as a tool, it reduces that and it's reducing it by a huge margin. So it's not finishing everything for me. I still have to add my human intelligence but it is speeding up so much of this process. And I wanna kinda show you what happens when we put this essentially big prompt because I'm asking you to do a lot of different things. So I told that, hey, you're an expert instructional designer. You're coming up with a webinar on communication tactics and I want you to do all of these things. And I'm just gonna kinda let it go and start to generate. And like I said, it's not going to be done. It's not going to be perfect, but it is going to get me so much further so much faster than I could ever do on my own. And you could do this for your customers. You could take your own content. So let's say you have a customer that really cares about something in particular. You could go grab some of your support articles and you can put it together for them in a customized way asking chat to help you. So let's see how it did summarizing insight, I books influence never split the difference, how to win friends, crucial conversations, effort with experience. I've read three of these. I haven't read four and five, but ever split the difference is probably my favorite book on communication. So for anybody that hasn't read it, I highly recommended. I've actually read it twice common themes and patterns, understanding human behavior, effective communication, building relationships, ethical persuasion and reducing effort in portion. I think this is interesting because this is a lot of what customer success has to be good at. I know you folks probably are some examples and stories. These are great. An anecdote about CSM who use negotiation techniques for never split the difference. So it gives me like an example of what I would generate and I would actually ask it to go further on this like give me more specifics, what's the scenario, who, what on way, where here's the learning objectives. This is what the folks are gonna take away from the webinar. And then we've got our outline. This is high level. I would ask it to go deeper on all of these but this is a great starting point. Again, this would take me 10 minutes on my own, maybe 30 or an hour and it does it for me in seconds. And then, hey, we've got some assessments ease of understanding case studies. It actually said, hey, we're missing some things probably need case studies, interactive components and a feedback mechanism. Great. Thank you for calling that out here's. Some quick questions… exercise… and then our slide outline with suggested visuals. You can see it actually kinda skipped over here because it's like sometimes chat likes to kinda be almost lazy. It's like, hey, are you sure you want me to generate all of that? Again? I know this is a lot I'm throwing you all at once. The takeaway I want you to get out of, this is a very broad tool. You have a very broad job and what you do in customer success, this can help you in so many different ways. You just got to start playing with it. You've got to learn the underlying prompting and from there, this is the limit. So we just went over that demo. It saves you a ton of time. So maybe six hours down to less than two to get this ready. Maybe even more… and building that ChatGPT habit that's the name of the game here and there really simple ways to get started. If it's something important or annoying, you should try chat GPT or if it's gonna take you longer than five minutes to work on it, you should try ChatGPT. And then perfect is the enemy of good. Just get started. ChatGPT is so fast. I procrastinate like crazy if I'm procrastinating, I use ChatGPT to just get me started even if it's not good. The momentum helps me get motivation and then give yourself, you know, space to explore. It's a fun new tool. So go where your curiosity takes you. It's gonna be much more pleasant experience. And then lastly have fun. This is a once in a lifetime technology try, to get the most out of it. I'll give you an example. I've got a daughter turning five next week. We'd like to write stories together using ChatGPT. She really loves it because she gets to participate in the creation of the stories. All right. I know that was a lot. Thank you all for hanging in there. So let's recap. We talked about what ChatGPT is Lawrence language model and of course, how to elevate your knowledge while also just getting things done. There's so much you. Can do to help you how you can craft effective prompts using those eight components. We did a live demonstration across a couple of different instances. And then I'm hoping you leave today with inspiration on the future of AI and customer success. So with that, we can jump into the Q and a. And I hope of course, to connect with all of you in the near future.

Matik MC: Awesome. Mike. Fantastic presentation. Thank you so much for putting that together. I know I learned a ton. We had somebody in the Q and a actually say, you know, hey, I really wanted to say, I appreciate the content and the presentation style. So great work there. I got, some great ideas already on stuff that we're working here at Matik, but yeah, let's jump into the Q and a. So the first question, is there a risk in asking ChatGPT to summarize data that has confidential company information?

Mickey Powell: Though there is, and there always is just like there's risk in putting it into a Google doc or having it on a computer that is in air gap. For instance, there is always risk however ChatGPT, OpenAI rather has been very open about their data use policies and they have, and if you actually go to their data use policies like OpenAI dot com data use. The first thing they put up there is that they don't use your data just to train their models, unless you opt in, which is important, right? Opt in versus opt out. Now that's all to be said that if you're concerned or if your company has given you specific guidelines not to do that, then I wouldn't for the time being.

Matik MC: Fair, right. Good point. Another question. How do you make ChatGPT change their tone to sound more human, but still professional? I feel like it's got two modes, super professional robotic and to like slang every time I ask about to sound more friendly, the tone of way to informal?

Mickey Powell: Yeah, this is a great question. Number one, of course, you can tell it. You can give it guidance specifically on like no too formal. Two informal, be more like this, be more like that in the ebook that I put out. There is, an example I used of actually reverse engineering your writing style so you can actually give it a sample, you can have it evaluate that sample and give you a style guide of your writing style. So that next time you ask it to write something, you can say, write it in this style and it will sound more like you. You can actually get that copy of the book for free at ChatGPT update a.

Matik MC: And I've got a follow on question on that, like, so your style, right? You, you send that information, it learns. But you said that it also has like that 300 word limit. So, like will it forget your style eventually and you have to redo your style or how does that work?

Mickey Powell: Yeah. So OpenAI and this actually kind of dub tails into the next question of like, is the plus really worth it? If you have the plus plan, you can actually do custom instructions, which means that every time you use, it doesn't remember like the whole model doesn't remember, but it will bring in custom instructions to that individual session. So you can give it custom instruction to say this is how I prefer to write. This is what I care about. This is you know, the things that I work on and it will use that as it generates. So I suggest you know, generating the style guide of your style to keep to have and to use later but also to use ChatGPT plus give it those custom instructions and it, you'll get a better output. That's one of the reasons why. And I'm just gonna kinda roll into the next question. The plus is really worth it. You get access to plugins, you get access to GP T4, which is a much better model and one of the plugins I use daily that, you know, we just didn't have enough time to demonstrate here. It's called code interpreter and it can do basic and advanced to statistical analysis for you really fast. So I whole heartily think plus is worth it if you can afford the, if you can afford the 20 dollars a month. And if you can't if you go to ChatGPT being and use creative mode, that is GP T4 underneath. So you can use that for free. You just don't get access to like the plugins and things like that. Thanks Dan. I appreciate the kind words Sara, significant improvements in GP four. Yes. GB T4 is significantly better. I almost never use three point five anymore like maybe once a month, if I just really need something, really, fast like that's. The only difference is three point five is much faster than four in generating, but the quality is significantly better on four… Jonathan ask, have I seen CSM use chat agents and our plugins for their work? Not a lot yet. And that's mainly because CSM deal with a lot of very contextual issues and problems and challenges. So the context really matters now. I will say that update AI is an example of using chat within the workflow of a CSM. So update AI, for instance, when you have a call, we take that transcript, we use chat to generate your meeting notes, action items, what risks have come up? What type of risk they are, product feedback, what type of product feedback it is as well as important numbers. And then we also use it to draft your. Follow up e-mail after the meeting and prep you for the next day's. Worth the meetings. So that's where I kind of see, the market going right now is more built into your workflow as opposed to pushing you out of it. And that's what we are doing it at. Update. I… see car says, I'm not sure I'm comfortable with presentation, teaching us as chat to be for some reason books especially there's open litigation alleging chat there have pulled pulling data from books without compensating or telling the authors care brings up a good point. There is, there is an ethical question about, you know, what's and potentially a lawful question about what's published on the internet. Was it trained? Was it lawful? We are going to learn the answer to those questions over time as lawsuits play out in the courts and figure out, you know, does the law believe it was lawful? So, I have my own opinions, but I think we have to wait and see how that plays out.

Matik MC: Awesome. I think those are all the questions. Great questions. Thank you for saying that everybody. And yeah, I just wanted to give a big thanks to Mike, for taking his time presenting this, and sharing all this information in with us. If there are any further questions that come in… they will be posted on our hub that we sent out after the session with a bunch more information. But yeah, just thank you everybody for attending and thank you Nikki for your time.

Mickey Powell: Of course, thank you all. Have a great day.

Matik MC: Cheers, bye.

 

---

Click here to see all sessions & speakers, and navigate to recordings.

Next
Request DemoSchedule Demo

Related Blogs

AICSS
See All Sessions & Speakers
Bex Sekar
  -  
August 30, 2023
AICSS
AI-powered CS: The Ultimate Guide for CS Leaders (AI + Customer Success Summit Session)
Bex Sekar
  -  
August 14, 2023
AICSS
Workshop: How to Use AI to Create & Repurpose Content (AI + Customer Success Summit Session)
Bex Sekar
  -  
August 9, 2023
© 2023 Matik, Inc.
Product
How It WorksIntegrationsAI FeaturesSecurity
Resources
BlogTemplatesAI + CS Resource HubCase StudiesHelp Center