Can Voice AI Really Make You $10K in 30 Minutes? Discover the Secrets! | voiceaiwrapper

Hi everyone, thanks for taking a look at our video. So before, uh, I get into the core of the video, a quick introduction about who we are and what we do. So we are VoiceAirWrapper and we are a platform that agencies use to white label voice providers like VAPI, Retail, ElevenLabs, etc. VoiceAirWrapper allows them to create their own client portals under their brand and also control pricing as well as automate client onboarding. So, um, in this video, um, I'm gonna cover some basic facts about creating agents. I am going to use VAPI as an example. This is a principle which you can apply to any provider, but this is going to be useful if you're starting off in the voice AI space and you're unsure about prompting and how to configure agents. This might also be useful even if you're an experienced voice agent builder and just make some of your workflows a little bit easier. So let's dive in. Quick summary about what we're going to cover. We will talk about creating, using AI to create your own voice AI agents, which is kind of great because your voice AI agents are basically a set of programs that understand markup language. And AI agents, that is AI applications which are LLMs, are great at creating content in markup, and it's easier for the agents to understand. So a couple of things that we're gonna do on this call is we are going to use VAPI as an example. We do work with other providers, but this is— we're just taking VAPI as a use case here, and we're going to create a— Inbound assistant for VAPI. We are going to then take that assistant and white label it with Voice AI Wrapper. So a couple of things which are important to remember. One is how your agent, voice agent, behaves and talks, and the, the quality, the output of the agent depends on first the configuration. When I say configuration, it's the Uh, the LLM that you're using inside your voice provider. So if you're using VAPI, so I'm talking about the LLM model that you're using, I'm talking about the transcriber that you're using, right? So even within the transcriber, there are different models. So all of these— so your agent is essentially a sum of parts, right? So just using the best transcriber or the best voice or the best LLM is not enough. You need to make sure that that entire stack, starting from your LLM to your transcriber to your voice, and everything works in harmony with each other, followed by of course your prompt. So, um, yes, these products are, are no-code products, which means you can just type normal English and you can still get output from the agents. But these agents are nothing but a set of programs, and, uh, the same fundamental that works, uh, in case of Google search or JSON structure works in case of agents as well, which means if the instructions you're giving your agents are well-defined and indexed, they are able to respond better and they will behave exactly in the way you want them to behave. So it's important to be able to write your prompt in Markdown and in a very organized and indexed fashion. And we'll show you how to use voice AI— we'll show you how to use AI applications to create your prompts. Lastly, of course, are your guardrails. So beyond your prompt, there might be certain nuances, certain things that you want to avoid, right? So for instance, you don't— LLMs are free flow. They can— they'll just go wild, right? So if you, if you don't want them to make false claims or, you know, maybe say something which is, uh, which is harmful to your brand, you need to make sure that you Add that piece of information, audit instruction, in the prompt, and that's what we're gonna show you. So step 1, we are going to look at what do we mean by configuration, and for that we will head into the VAPI console. Okay, we are inside VAPI right now, and like I said, the objective of this video or this demo is to create an agent for Voice AI wrapper. So I am going to name this agent Voice AI wrapper, and I'm going to select customer support specialist. There's a template which is ready-made, which is great because it gives me a head start. Yeah, and then inside this agent, I can just go ahead and change the name just to make sure this stands out. All right, so now if you, if you look at this agent right now, there are a lot of settings inside the VAPI configuration, right? The first thing is your LLM. So you have a bunch of LLMs and then you have a bunch of LLM models, right? Each model has its own characteristics or its own specialty, right? So for instance, just because you have 4.1 doesn't mean that you cannot use 3.5 Turbo, right? So similarly, if you have a scenario where you want the agent to do calculations, then 4.1 or 4.0 is way better than using another model from the same provider, OpenAI. Yeah. So now how do we know all of this? Either we know that we learned that by hit or trial, hit and trial, or we ask AI to tell us what is the best model suited for my type of agent. Yeah. So that's what we're gonna arrive at on this call, on this demo. And then as you scroll down, you have the prompt. And like you see here, VAPI has already provided a template, which is great. But this template might not have all the inclusions that I need for my agents. So I'm gonna show you how to take this and improve on the prompt that is already provided. Next up, adding files is extremely important. So you don't want to front-load your prompt with all information about your product or your service. You would want to add that in form of knowledge base files so that agent can quickly refer as and when required and not get overloaded with a huge prompt. So now again, when you add files, most people, uh, would just upload a PDF or a Word document. Um, converting it to a text file which is in Markdown is ideally recommended, uh, is recommended by us. Why? It's lightweight. 2, Markdown is indexed and it's organized, so it's very easy to search. And, you know, by nature LLMs are faster when it— when, when you have structured data for them to look up. Moving forward, you have tokens. Again, this means how many tokens are being used. The agent will keep on processing in the background. The ideal range is between 100 to 250. So depending on the scenario and the agent, you would want to control that. Then you have additional controls like which voice provider you use. Yeah, and even with inside the voice provider, you have have several voice types that you can choose from. Again, inside the voice provider, you have different models you can choose from. This is also important, uh, because depending on whatever use case you're handling, uh, the voice provider model might be optimized for that use case. Again, you can use an LLM, uh, to do the research and tell you which one is ideal for your scenario. Yeah, moving on. This is pretty standard. You can tweak and optimize this as per your liking. Now, another important setting is your transcriber. So again, there are a bunch of transcribers here. Deepgram usually stands out in terms of quality, but again, depending on your use case, you might need a different transcriber. And after picking the transcriber, the transcriber model is also very important. So for instance, if you have a medical client, If you select NovaTree Medical, this is optimized for medical terms. It recognizes and it catches medical terms very easily. Yeah, so just selecting the transcriber provider is not enough. There's a specific model that you might want to pick to make sure that your agent is optimized to handle the kind of conversations that you want it to handle. Um, so yeah, so these are the basic functions. Rest of them are about configuring your, uh, your agent with a calendar or any other third-party tool. Then you have other functions. So these can be controlled by you, but the major items are your LLM, your transcriber, and the voice provider that you use and their corresponding models within these primary providers, right? So, um, next up we are going to create a set of basic instructions which will help us figure out what kind of configurations to use here. And we're gonna use AI to tell us what configuration I should use with this agent. So let's get to that. So now I'm gonna put together a set of instructions which I will then send over to an AI assistant. To execute. So again, going back to my purpose, my purpose is to create a customer service voice agent for Voice AI Wrapper, and I'm going to add this to my site. And then anybody comes into my site and clicks on the agent, it should be able to take questions and answer questions about Voice AI Wrapper. So I've created a Word doc where I'm clearly defining what the role of the, uh, agent, what the role of the AI should be when it's creating, giving me the suggestions for the configurations that I need to add to my AI agent. So again, I've used very basic indexing structure here. So I've defined what role the AI is supposed to take on before it starts giving me answers. I've also defined what, what is the goal of this entire exercise. Then I've given it 3 clear tasks, right? And I've told it that you're supposed to complete task 1 before you move to task 2. The first task is, uh, based on the agent type that I described. Now I'm going to add an agent description, uh, at the bottom where I'm going to tell it that my agent needs to be a customer service agent which is answering questions about Voice AI Wrapper. So based on the agent type description, it's supposed to tell me what settings to use in WebAI. Yeah, this includes the LLM, the token size, the temperature, the voice provider, everything that I showed you in the last clip. Now, once it's done with that, task 2 would be to create a knowledge base. Now, the agent is going to, or the automation is going to scrape my site and create a knowledge base in a markdown format so that it becomes easier for me to just upload it into the Webby knowledge base. Once it's completed that, I want it to then proceed with task 3, which is to create the optimized prompt for me. And for that, I will give it the sample prompt which VAPI has already given me. So if I go back into VAPI, it has given me a prompt. I just want to enhance this prompt to accommodate whatever additional requirements that I have for my agent. So I'm going to take all of this and I would run it Any AI product. So I can use— you can use Manus, you can use GPT, you can use Claude, you can use anything of your choice. So let's use this in one of the AI products that I use and let's see what we get. Okay, I am going to use one of the AI systems that I use frequently, and then I just dump all my instructions here And let's see what it comes up with. Okay, it's doing its thinking. Now, the edge or the advantage that I get here is it's doing a lot of research on best practices and articles and knowledge base published by Wappi. So it Saves me the time of doing research and my own due diligence. All right, so now it has done its research and it's starting with task 1. So the first task was to tell me what configuration to use, and it's quick to the job. It's recommended GPT-4.0 Mini, and there's a reason, because it's good with latency. It's also Gave me the token size, 250, and it gave me a reason as well. They'll explain what temperature setting I should use. And excellent. Okay, so I— let's use this. And it's also given me a cost prediction. So let's do this. And then like I instructed it, It's going to wait till I, uh, ask it to proceed with task 2. Uh, and if we remember, task 2, which I also forgot, is, uh, is to add a knowledge base, right? So we are going to apply what suggestions it gave me to our agent inside VAPI. Uh, let's do that. I've added the, uh, the settings to VAPI and, uh, I've Just updated the assistant. So now I'm gonna move on with task number two, which is yes, task two. All right, I'm happy with task one. So I'm I've just given it instructions to proceed with task two, which is to go to our website and extract information and create a comprehensive knowledge base in markdown format. And I'm going to use that. And upload it in the knowledge base section here. So let's wait till it's done. Now it's generated the knowledge base, and this is at least latest and updated. So it has all the key information that a customer service agent is supposed to know about us while taking on calls from callers, and I'm happy with this. So I've just asked it to create this in a downloadable I'm going to take this and add it to VAPI right after it's done creating the file. It's done with task 2, which is to create the knowledge base file, which I've added to VAPI. Now we'll go to task 3, which is to create the prompt. Now create the prompt and the Existing prompt is. Okay, so we've given it the existing prompt inside VAPI's template, and it already knows about the purpose of the agent. So let's see what it comes up with. It started generating the revised prompt, and It also included the things that I wanted to— what I wanted it to watch out for. So for example, it also added a touch of its own for skeptical prospects, and it added that guardrail for not making false claims when to escalate. So it did that all on its own based on what I instructed it. So now I just take this and I head over to my VAPI prompt and I just paste it. And I'm going to remove the part which is not necessary. Okay, and I'm going to rename this to reflect the name of the agent. Let's say we want to call her Anna, and we're going to call this Voice AI Wrapper. All right, so now we've made changes to the configuration based on the suggestions given to us by the AI agent. We've also updated and enhanced the prompt, and we use that to update our VAPI agent. And I'm just going to publish this, and let's take a call and see if this works out.

Hi there, this is Anna from Voice AI Wrapper customer support. How can I help you today?

Hey Anna, what does Voice AI Wrapper do exactly?

Great question. Voice AI Wrapper is a platform designed to help agencies like yours launch their own voice AI services quickly and easily under your brand.

We work with dedicated clients, and do you guys offer any security for— I mean, are you guys certified security-wise?

Absolutely. Voice AI Wrapper is designed with security in mind. We offer enterprise-grade security, data encryption, and options for SOC 2 compliance preparation.

Well, that's great. Thank you.

You're so welcome. If you'd like, I can guide you on how to sign up for our free 10-day trial or answer any other questions you have.

Okay, that's it. So we just created an agent. We configured the agent. We created the prompt in less than, I don't know, maybe 10, 15 minutes and without any grunt work. And all that I need to spend time on is to talk to the agent and find out whether the final output suits my purpose or not. Now, um, once we're done with all of this, you can then head over to Voice AI Wrapper and white label the whole thing. Um, let me show it to you in a few minutes. All right, I'm logged into my Voice AI Wrapper account, and once I'm done creating my agent in VAPI, I would just go in here, add my VAPI keys. This will sync my agent inside Voice AI Wrapper, and then the assistant that I just created is going to show up here, right? Along with the phone numbers and the knowledge base and everything that I created. Now I can go into campaigns and I can create an inbound campaign, associate this agent, and when this agent starts taking calls, all your calls are going to show up here along with the call logs, recordings, plus I'm also going to get analytics on the incoming calls. And if I were to offer this to a client, I would— I can go ahead and, you know, add on my own domain here, white label the whole thing, add a client account which will have their own client portal, and have the clients log in to the client portal and view analytics for the inbound campaigns. So we have more demos and tutorials on how to white label with Voice AI Wrapper. You can check the rest of the videos on this channel to understand how Voice Airapper allows you to launch your white label business. I hope this video was helpful. Thank you, and let me know what other videos you'd like to see.