Transcription
You probably type most of your chatbot messages. I know that I do. But you also may be missing out on one of the most powerful and overlooked features. I'm specifically talking about advanced voice mode, which can be used for creating custom AI partners such as personal sales coaches, an interview that can prep any person for a big opportunity, or a private sounding board to help you practice difficult conversations. I'll show you how to build these step by step, starting with the simplest. Let's get into it.
Now, the first place we should start is asking ourselves, what's the primary problem here? Why don't most people use advanced voice in the first place? Well, in my head, there's two reasons. First off is a lack of knowledge, knowing that this feature even exists and/or that this feature is provided by many different providers. And the other thing is habit building. So, this is a different type of interaction that we're going to have with technology. So, it's a habit that we need to build because over the last couple of decades, we've built the habit of typing in a keyboard, asking questions in a Google-like way, and a variety of other things. But now we need to transition to a more conversational way of interacting with technology because I do firmly believe that voice is going to be the primary method that we interact with technology going forward. And there's three reasons why, two of which are kind of intertwined. So the first one is that it's much more natural for us to talk than type. We humans were meant to converse. So it's much easier for us to converse with technology. And that then goes to the next point which is the cognitive load of talking is much lower than typing. Often times when we type, we're actively editing what we're trying to state, which can be useful sometimes, but often times if the AI is intelligent enough, it'll understand and intuit what we're trying to achieve. And then finally, this is going to be one of the most interesting: is that I could see AI being more ambient in nature. Now, what does that mean? It's basically saying that AI sits in the background quietly waiting for you to initiate it, similar to how Alexa or Siri works today. You have to use your trigger word to then get the AI to interact with you. The same premise would happen with more intelligent AIs. And I'll actually show you one use case in relation to having an ambient AI assist you in reading.
Now, as I mentioned previously, there are tons of providers that are already providing this technology today. So, here we have the four big players, all of which have advanced voice mode. I have Claude slightly shaded out because their version of advanced voice mode is not as good as the others, but they still have the existing feature. Now, of these three, I'd say that ChatGPT is probably the best when it comes to utility, but there is a lot of entertainment in Grok. If you want to go over there and play with their unhinged version or whatever else. And for this video, we're going to focus most of our attention on ChatGPT because that's what most people have access to today.
Now, with ChatGPT, there's obviously different plans, right? So, we have free, we have Plus, we have Pro, and then we have Business or Enterprise, different variety of plans. Now, I wanted to give you a quick snapshot of what type of access you have to advanced voice mode. So, if you have free access, you're using ChatGPT for free, you'll have limited access and a limited number of hours per day. Now, there's no specific numbers that were shared by OpenAI, but they're kind of vague on the ranges. So, they say limited number of hours for free. When you're using Plus, it's nearly unlimited per day. And if you're using Pro, same thing. It's basically unlimited per day. It's also important to note that the AI that's behind advanced voice mode is actually GPT-4o. That will likely change in the coming months where they'll then shift this to GPT-5, but for now, it's still using 4o. So, that's a good thing to just keep in mind.
And before we jump into the use cases, I want to quickly show you how we can get access to ChatGPT or at least how to use it. So if you're on your computer, you would simply go to the ChatGPT instance here. And on the right-hand side, you have this little speaker icon that says "Use voice mode." When you select this, you'll be able to talk to AI. So we'll give it a test.
>> Hey, can you hear me?
>> Yep, I can hear you loud and clear. Let me know what you need.
>> Awesome. I want you to quickly explain what entropy means in the context of the universe. To keep it very concise.
>> Sure. And in a nutshell, entropy in the context of the universe is basically a measure of disorder or randomness. Over time, the universe tends to move.
>> And there you go. So, that's how you use advanced voice mode. A few things in the context of this. I'm going to turn it back on to show you a few other features. So, here we have the mute button. So, when I turn this on, it's going to mute my microphone. And we'll talk about why that's important in a second later on. The other thing is here, if we go to settings.
>> Hey, what's up?
>> Feel free to ask me about anything that's on your mind. I'm ready to go.
>> Hold on. Okay. So, I muted it so you can't hear it. But basically, you can go through these different options to see different voices that you would like to use or interact with. So, um, that, uh, that's how that functions and that's how that works. That's how we use it. And you can see that we have the previous statements that were discussed verbally. So, you can see the things that I said and the things that AI said back to me.
Hey there. Quick pause in your regular programming. This video is brought to you by me. Two quick things. First thing below is a free link to a 30-day AI insight series. Completely free. You'll get insights in your inbox of how you can apply AI to your business and your work. The second thing is if you'd like to work with me, I have a series of offerings below to see if there's a good fit between the two of us. That being said, let's get back into the video.
Now, let's jump right into the use cases. I've broken them into two categories. One is the basic use case, something that's more simple, and then the other one's more advanced use cases where we tailor the AI to our specific scenarios. We'll start with easy and get more complex later. For basic use cases, it's mainly AI on the move. So either you're walking or driving. And here we have daily walks. So this is something that I do on a consistent basis where I go on a walk and then I talk to AI. And from my perspective, I'd say the easiest way to transition into a habit associated to advanced voice mode is getting your mobile phone, downloading the mobile app onto your phone for ChatGPT and using it that way instead of using it on your computer. So the way that it works is I have my phone with me. I usually put on earbuds, Bluetooth via my phone. I put my phone in my pocket, but I turn on advanced voice mode. Once I've done that, I can ask simple questions while I'm walking around based off of just things I'm curious about that day or what I'm reading. I could also think through specific ideas I'm thinking about in relation to my company or problems that I'm facing. And then this final one down here about learning about topics. The interesting tactic here you can play with is they do have learning mode inside of ChatGPT, but if that doesn't always work with advanced voice mode, what you can do is you can basically pre-prompt the AI when you initially turn it on and say, "Hey, I want you to talk to me in a Socratic method about this specific topic." So instead of it just spitting out ideas, it's asking you questions in relation to the idea so you can ensure that you're learning the topic more effectively. It's a bit of a minor tip, but it's actually really useful.
Now, in addition to daily walks, a lot of people have commutes, so they commute to work or elsewhere. For me, I commute to the gym. So on my commutes, often times I would listen to podcasts, but I've started to, I would say 40% of the time, I've replaced podcasts of passive passive listening to active problem-solving with AI. So what I'll do is I'll turn advanced voice mode on. I'll have it set up on my car when I'm driving and I'll be able to talk to it on the way there, talking about specific problems on the commute there and back. A very specific example is I was driving back from the gym and I needed to go somewhere where they would allow me to do money orders and I had no idea where that would be around that location because I've never done it in this location. So, I had to talk to advanced voice mode on the way home. When I did that, it was doing research on my behalf and was finding different locations. By the time I got home, I had a list of two that were open on that Sunday at that time so I could then get that money order.
Now, another one that I really enjoy is social interactions. So, often times my wife and I, when we have dinner, we have conversations about a variety of topics and it can get in areas where we don't necessarily know the answer to the question. We'll think through it together a little bit, but often times the more enjoyable part of this is pulling out AI, setting it on the table, and adding it as a third person to the conversation. So, put it into advanced voice mode and then have a conversation in relation to that topic you're interested in. And a recent one we were talking about entropy and I promise you our conversations aren't always this intellectual, but we were talking about entropy and we wanted to know exactly what it meant in the context of the universe. So we started asking questions of the AI to this and the cool thing about advanced voice mode and AI in general is you don't necessarily just get an answer back, but you can ask questions at the level of your understanding and have it explain different aspects of that answer to you.
Now, a pro tip in relation to the basic use cases as well as the advanced use cases is muting yourself. So, when you use advanced voice mode, there is the ability to mute yourself. And you want to do this so the AI doesn't accidentally stop thinking that you're trying to interrupt it to make a point. This could be because of background noise or something else. So, when you're having that conversation with AI, make sure you mute yourself after you're done speaking and then unmute yourself when you want to interrupt. Now, this is getting better with different forms of AI that are being released in relation to voice-to-voice conversations. But the turn-taking aspects of things is quite difficult and it's easy for us humans because we know when to interrupt, but for AI, it can be kind of challenging. So mute yourself to make things more seamless and interactive.
And now for our final basic use case, which is slightly more advanced but it's still interesting, where I've seen others and I've tried myself doing a reading companion component. Where you're reading a book and you've set your phone down, either you're sitting down and reading or you're pacing and reading. You sit your phone down and you have advanced voice mode on in the background in an ambient way. So this is the ambient use case I was speaking about where it's sitting in the background and it's waiting for you to trigger it. Now, when you're reading the book, you may run across a passage that's too confusing and you want a helpful guide to give you some assistance on understanding what that passage represents and what it means. So, you can simply say, "Hey, I'm reading this book and this chapter on this passage. Can you give me some context about that?" And it likely will have the answer you need for it. Another thing that I really enjoy is getting explanations explained to me at different levels of understanding. So if I'm reading some deep philosophical piece of work or something that's technical, I want the AI to then explain at different levels so I can ensure that I understand each one of the levels before moving on to the next part of the book.
Now, that's our reading companion. We're going to go on to advanced use cases where we're tailoring the AI to a specific scenario, baking a certain persona into it so it can help us deal with that scenario more effectively. Now, there are different ways we can customize the advanced voice mode. You have two choices. You can either do it with a custom GPT or you can do it with a GPT project. Both of which have pros and cons. So on the custom GPT side, it's sharable. So if you have a specific advanced voice mode scenario that you want to share with others, you can then make that sharable to others. In this example, we could do something around like a sales pitch. And I'll show you an example of that. Then we have instant context, meaning that the AI is going to immediately understand the system instructions that tailored it to that scenario, and it's going to jump right into character instead of you having to prompt it. Here you have to make sure that the context is there. It usually always is, but it's nice to double-check for GPT projects. And then the final thing on custom GPTs is the history isn't always saved. So when you're having a conversation, you want to stop the conversation and you want to be able to see the text from the conversation that was had, that's not always there. So if the history matters to you, then you may want to default to GPT projects because that is readily available. And then the final thing here is that it's private. So if you want it to be private, that could be a pro, but if you don't want it to be, that could be a con. So those are our options.
So now what I'm going to do is I'm going to walk you through an example for a sales training. So this is something I've seen a lot of companies get benefit from where you create a custom GPT or GPT project with a very specific persona and then you allow your sales reps to practice their pitch to ensure that they can get live experience without risking the loss of a customer. And the cool thing here is that you can bake different personas into the GPTs. So you could have a chatty buyer, you could have a skeptical CFO, you could have a hostile analyst, or whatever is most common for the service or product that you're pitching. Because oftentimes, you know, when you're pitching different customers, there's different types of customers you'll run into. Of those personas, there's probably one that's more likely than the others. So you should bake that into your GPT, allow your sales reps to practice that so they can be prepared for that customer interaction. And as I mentioned previously, there are clear benefits to this. So the first one is we're reducing the risk, right? Right? So, we're reducing the risk of churning our prospects by a poor sales pitch. The other one is confidence. So, we're increasing the confidence of our sales reps because they're practicing in a semi-live environment with a persona that could be somewhat hostile. And finally, they can practice 24/7. They don't need you to be around to practice. They don't need other sales reps to be around to practice. They can pull out their phone and they can practice at any time. So, those are some of the clear benefits.
Now, let's go into the example. So, this here is the prompt, the system prompt that sits behind our custom GPT. And in this case, there's a few things that I wanted to call out here. So, when you're creating your system prompt for an advanced voice mode interaction, especially in this sales context, you want to give it a persona that fits back to the personas that I mentioned. You obviously want to give it the context of the product that you're selling, the situation that they're in of what's happening in that interaction, and then the overall outcome you want from the AI at the end of the conversation. So, here we have the persona, and the persona here is a skeptical CFO. They're running a services business that's between 5 and 40 million ARR. And then you give it some constraints stating that you need it, the AI, to be more conversational, ensuring that there's a lot of back and forth between them and the sales rep to make it more realistic. After that, we go into the product, giving it some context on what's being sold to them. In this case, we're selling paper from Dunder Mifflin. If you're an Office fan, shout out to The Office. And then we're stating that we need to adapt or adopt this persona in this scenario. We're stating the situation. So, this is ideally just a five-minute conversation where they walk into your store, they're doing a door-knocking, so going from office to office, foot traffic, trying to then sell paper to you for your office supplies. And then the outcome from this is obviously for us to practice with the sales rep. But really, what we want you to do at the end of the conversation is we want you to rate or assess the quality of the sale based off of different criteria that we care about in our sales team. So, in this case, we care about credibility. We care about the sales rep uncovering pain points. We care about them presenting quantifiable value, as well as handling objections and a variety of other things. So this is a base system prompt that we can start with.
But you can improve any system prompt with different optimizers. And this is something I've talked about in different videos where you take a base prompt that's somewhat structured. You drop that into a prompt optimizer and then it's automatically going to inject the best practices for that given model into that prompt. So in this example, we have OpenAI and OpenAI has a prompt optimizer that we can use to improve our prompt. So I've already actually improved the prompt previously. So this is our custom GPT, our Dunder Mifflin sales practice rep. So if I go into edit GPT, you'll see behind here, we have a much longer prompt that has more structure and is likely going to provide a higher quality experience for our sales rep. All we would have to do to improve our prompt to make it look more structured and get higher quality out of it is we would take the base prompt that we had from this previous slide. We'd send it to this link here, which is going to be not there. Let me copy it out and show you. So, I take this link. I go to it's going to take me directly to the optimizer. Real quick, this is a separate platform from ChatGPT, but if you have a ChatGPT account that's paid, you can log into it and you can then get access to the optimizer. This is all provided by OpenAI. It's just their developer platform and this is one of the features inside of it. So, once you've gone to this, you can paste in your base prompt here. You select "Optimize" and it automatically improves it. If you don't want to go to this very specific link that I provided in the slide, you can then just simply go to, I think it's platform.openai.com. It takes you to a page like this. You'll go to "Dashboard" in the upper right-hand corner. You'll then go to "Chat." You'll select "Create" in the center. Then you have this chat field here where it says "Developer message." You put in your prompt here. So your base prompt goes here. Once you put in some text, the upper right-hand corner, this "Optimize" button turns white instead of gray. You can select that. That then optimizes your prompt. So it's just two different ways to get access to the optimizer allowing you to improve your prompts automatically.
So another advanced use case here is interview preparation. So say you have an interview that you have coming up for a job. What you can do is you can create a custom GPT that entails these type of jobs you're applying for. You can also take your resume and you can upload that into its knowledge base. So, what I mean by knowledge base, so if I come back here and I close this out, say we're working on a custom GPT in this case, if I scroll down, you'll see that down here there's a section that's called "Knowledge." So, you can upload your resume here to ensure that anytime the AI is asking you questions interactively as if it were an interviewer, it's referring to your resume to make sure these questions are targeted and relevant. Another way that we can make it even more relevant is in addition to having your resume, when we start interacting with this as a user, we can paste in the job description saying, "I want you to interview me based off my resume and this specific job that I'm applying for." And through that practice, you can then practice very targeted for that job, that company, and your resume. And the way that you would turn on advanced voice mode in this context is you would simply paste in the job description with your custom GPT already built. You would hit enter. It would come back with probably a first question. After you came back with the first question, you would switch on advanced voice mode and you would just interact with it that way because it'll have access to all the previous things that happened in that conversation.
Another really useful use case that I've seen a variety of companies get benefit from as well as friends and personal relationships benefit from is having difficult conversations with AI first to practice to then have the difficult conversation either with an employee, a spouse, a child, a friend, whoever else. You can apply this to a variety of use cases. So like I said, we have personal, but I'll focus on the company side this time where we have manager-employee dynamics. So often times managers, they want to give good feedback but they don't have either the practice or experience to do so effectively and they don't necessarily know how to deal with difficult conversations. And the best way to get good at difficult conversations is to have more of them. And the best way to do this without risking a relationship, at least in the start, is doing this in a simulated fashion with AI. So, we can set up a scenario where we're having a candid feedback conversation with an employee or giving performance issue feedback or a conflict resolution situation or something like that where we create a custom GPT or a GPT project that has a few things baked into it. So, first off, it has the situation associated to what we're talking about. So, I'm the manager, this is the employee, this is what we're working through, here's some maybe performance feedback that we've already had and put that in the knowledge base and a variety of other things. So when you have that conversation, you can have it take on the personality of the employee because you can bake that into the GPT project or custom GPT as well, stating that this person is either passive-aggressive or whatever else to ensure that when you have that conversation with the AI in advanced voice mode, you're practicing that very specific personality that you're going to have to deal with, both dialed up and dialed down in real life. So I highly recommend working on this, especially if you're a manager and you want to work on your skills of being able to have more difficult conversations. And on the personal side, I highly recommend people use this if you're a spouse, a parent, or whatever else.
And as a quick recap, I firmly believe that voice is going to be the future of how we interact with AI. There's many reasons why: mainly, it's more natural, lower cognitive load, and we can always have it on anytime we want to interact with AI around our lives. Next thing is we should start simple because we need to build the habits to use these tools effectively. So, starting simple with walk-and-talks, commutes, social conversations with dinner, or reading companions are a good place to start to know how to pull the tool out with your phone. Use it when you need it. And the more advanced use cases that you're likely going to get even more value from when you set these up effectively and correctly is you can practice sales with your team, reducing risk and increasing their confidence. You can prepare for interviews for bigger opportunities and prepare yourself to be ready for those conversations. And also you can have the ability to get better at difficult conversations both as a manager and/or as a spouse, a parent, or a friend. And these are just some use cases. These are just a small sliver of the use cases you can use advanced voice mode for. So I recommend being as creative as possible to get the most out of these tools. And that's it.
So if you enjoyed this, please reshare with your friends. And as I mentioned previously, two things. One, below is a free link to a 30-day AI insight series. Completely free. You'll get 30 insights in your inbox on how you can apply AI to your business and your work. The second thing is if you'd like to work with me, below is a series of offerings to see if there's a good fit between the two of us. With that being said, you should totally check out the next video. It's going to be around here that the YouTube gods thinks that you'll love. See you next time.