Voice agents and the roadmap

0:00 And we're doing it with a real use case from one of our clients, a law firm voice agent system with inbound qualification, appointment setting, and call transfer bundled with an outbound voice agent that can collect overdue invoices. And here's what makes this different. I'm showing you how to make every piece swappable. The CRM, the calendar platform. So once you build what I'm about to show you how to build, you can adapt it to any industry, any use case, any CRM.

0:22 Stick with me and let's build. Let me share why I'm the one teaching you this. I spent the last 8 years of my life in the world of AI. I worked as a data scientist. I worked on data projects at Nike. I've been involved in machine learning projects. But about 2 years ago, I went allin on Voice AI. And since then, I've deployed over 30 voice agents for real clients across different industries. I've also run over 600 hours of live coaching sessions, building, troubleshooting, and helping people just like you get their agents working.

1:09 And I've helped over a,000 people create their first voice agent. I'm teaching you what actually works because I've done it. I've taught it. And I understand what has tripped people up. So, not only am I going to show you the right way to do it, but also what mistakes to avoid so you can move faster and get your first or next sale. This video exists because I want you to have what I didn't when I started, complete clarity on how to build and sell voice agents.

1:31 Before we dive in, I want to show you exactly what we're covering. I've built a complete road map for this master class. On the left side, we have the build stage. This is what we're going to cover foundations, different platforms that we're going to use, and the inbound and outbound use cases that we're going to build live. We're going to be using platforms that we actually use to deliver for clients. And on the right, we have the sales side.

1:55 Once you know how to build it, you'll need to know how to package it, how to price it, and how to close a client. We're covering all of that. And each phase has a checklist. You can take a screenshot of this right now. I'll come back to it, reference it as we go. And I'll be checking things off as we go so you always understand where we're standing. Here's a checklist for the sales side all the way from deciding what you want to build, what do you want to sell, who to sell it to, and all the process up to closing and delivering for that client.

2:18 This is a very comprehensive tutorial. I encourage you to pause when you need to. It's going to be a lot of information and you need to digest it correctly. Throughout the video, you'll see the topic that we're in and a loading bar. Ideally, pause after the loading bar completes because it means we're transitioning to a different topic, which is an excellent point to take a break. I also encourage you to build it alongside me. As you take pauses, go repeat the steps.

2:40 It's not only the watching, but the doing that really helps deepen the learning. And then finally, don't just put this in a separate window. If you need to take a break, go ahead. But I do encourage you to watch it, listen, and take notes as you go. Maybe that's physically in a notebook or in notion or in your notepad in your computer. But if you do those three things, it's been proven that you will absorb many times more than watching or listening alone.

3:04 And stick with me because by the end, you'll have everything you need to build and sell voice AI agent systems. Now, I'm going to show you how I actually build prompts, but I want to show you the magic behind the scenes. It will build what you need, which are skills. You can get all these templates on school. You can get the master prompts on school, but your skills is what I'm interested in. But I'm not here to show you how to create a demo.

3:50 I'm here how to show you how to create a full uh voice AI system that you can package and that you can sell. If you stay till the end of the video, I'm going to share with you how you can import these three agents that we created, the three voltage agents, all the automations in N8 and the three though highle automations for caller memory, for post call analysis, and for triggering outbound calls. You'll have these available so we can speed you up, but you'll still need the skills to make this relevant for your client and their business.

4:13 And that is one of the skills that I'm going to teach you today to ensure we're all on the same page about what we're building. So, what are voice AI agents? It's simply an AI that talks to people on the phone in real time. So, if you've ever called your doctor's office to book an appointment or received a call as a reminder of that appointment, those are two use cases where an AI could be talking to people on the phone in real time.

4:37 And some of these platforms for building voice agents have been around for a couple years. So why now? The tools are actually mature now. LLMs are much better are actually following instructions consistently. Voice generation is much more natural. The costs have become a lot more reasonable to the point where you know a receptionist could cost anywhere from two to five times more than running a voice agent. And the two things that are most important to me is that these solve major pain points for actual businesses.

5:23 It creates utility. Uh staffing is is harder now than it's ever been for businesses. 24/7 availability. You would need three full-time people to cover that 24-hour span. And that doesn't even include weekends. Uh this agent is available 24/7. And the brand consistency. This is something that as AI becomes more uh widespread uh brands are are requiring that the way that the that the receptionist that the human communicates with a user is consistent and that's usually not the case.

5:50 And the other big reason is how much this market is going to grow. So, it's estimated that by 2030 in the small and mediumsiz business market, uh, voice AI actually covers $45 billion per year, and you can have a chunk of that pie. And 2026 is going to be the inflection year in which uh, these voice AI tools are a lot more widespread and is not anymore the first time that businesses hear about this. And 2026 is going to be the inflection year where we go from early adopters or the first you know couple years of AI uh into that early majority late majority which is the biggest chunk of the market.

6:38 So now is the time to capitalize on the technology being ready. We're going to talk about niches. We're going to talk about what are the best most profitable industries to be a part of and then also the use cases for voice AI. Uh before we do let's understand the anatomy of a voice agent. The easiest way to understand this is picture a user and lead calling a business and instead of the business picking up uh our voice agent picks up.

7:01 We build on retail is where we recommend beginners definitely start and we've built every solution for clients in retail. We're very happy with it. So instead of the business picking up, the voice agent is going to pick up. So the voice agent is going to hear what the lead says with a speechtoext model. Then it's going to reason over that text with an LLM and then it's going to generate from text to speech with a voice generator and then reply to the lead.

7:26 Uh and the conversation keeps going that way. This all happens in about a second. So it's that's what we say when we say real time is uh this just happens so naturally and fast. You have many options for these reasoning models, the LLM and the voice generation. But there's a couple recommendations I'm going to give you, which leads us to what can you control? What what do you have control over with voice? Which voice do you choose?

7:50 We recommend using 11 lapse voices. These are clone voices, extremely high quality, but you have many options within 11 Labs uh of which voice you choose, and I'm going to show you a couple that I really like. Then you can choose uh the LLM, the brain that rees and generates a response. We strongly recommend GBT4.1. It is very good at following instructions and actually understanding the nuances of the conversation. Then we have the system prompt and the tools.

8:14 The system prompt is simply the instructions that guide the behavior. And the tools is what are you plugging your voice agent into? Uh calendar booking? Are you plugging it into a CRM? Are you plugging into a payment processor? That will depend on your use case. That will depend on your client. And we're going to touch more on tools later. First, we need to understand inbound versus outbound because these voice agents can pick up calls from customers calling the business.

8:37 And this can look like, you know, an air receptionist, uh customer support, can be for appointment booking. And some considerations that that we will certainly talk about is telephon. What happens when a business already has a phone number that they use? Um, another consideration is there's plenty of reasons why a person can call a business. How do you handle that infinite amount of reasons? And then the other aspect is that a lot of times when a user calls, you don't actually know the user.

9:22 You don't know their name. There's no history about them. So, you're going to have to handle some tricky things like taking down their name, taking down their email, and there can be some some things to take care of. And there's some things that can go wrong there that we're also going to address. Then you also have outbound. And outbound has its own considerations. This is when your voice agent is calling the user, is calling the customer.

9:46 This can be for for appointment reminders, uh follow-ups to other conversations. Uh they filled out a form and now you're calling them to give them more information, push the sale forward, and to reactivate old leads. I really want you to start thinking about this lead reactivation because when it comes to many many many businesses in different industries, this is something that they neglect and it's a it's a gold mine just sitting there in their CRM. Uh this is not the use case we're building today, but it is with everything I'm going to show you today.

10:08 It is something you can build as well. Some considerations for outbound is number cleaning. If you've ever received a call that says spam likely, that's why you need to clean your numbers. I'm going to show you a couple tools we use uh that help with that. And then since you are calling the user, you have to be very mindful of the tone that they use that you're using, the approach that you're taking. Don't waste their time and don't sound salesy.

10:32 They either are interested or not. And in 2026, we are all used to receiving these calls that we don't want to pick up. Don't be one of those. You have to be very mindful about the tone that you're using. And this is something to iterate on with your client. So, what are we building today? What we're building is an inbound agent and an outbound agent that are going to be working together. And our use case is for a family law firm.

10:58 So, the inbound agent is going to receive calls from people who need uh a lawyer. uh it's going to qualify those clients, make sure that the use cases is for a family law and um it's going to book an initial consultation into their calendar. And then we're going to have an outbound agent that is going to uh be for existing clients. There's work that has been done. Particularly what happens a lot is uh overdue invoices. So, we're going to call clients that have an invoice that they haven't paid and process that payment right there, right?

11:45 And we're definitely going to talk about security concerns. So, for example, taking credit card numbers over the phone and some better ways to do it that makes you happy and your client happy. Now, why did I choose law firms? Well, they have really high value clients. The lifetime value of each of their clients is really high. So, any service that they get to be able to serve more clients and serve them better can generally have a higher price tag attached to it.

12:08 Also, they tend to have very clear workflows. Uh, existing clients will transfer to their current lawyer. New clients, this is the qualification process that we follow. And they also have huge recurrent revenue potential. If you're making money for them, if you're helping them smooth out their processes and serve more clients better, then they're going to stay with you. So, think about how those characteristics of the law firms industry applies to other industries as well and other use cases.

12:31 Think inbound receptionist for doctor offices, the appointment booker. So many calls go unanswered because the receptionist is doing something else. They were not hired just to answer calls and they can't get to the phone when they're doing something else. And if they don't get to the phone, then the business can lose a potential client. Then some other use cases that are actually industry agnostic uh are what we call speed to lead. You're on Instagram, Facebook, on a website.

13:19 You fill out a form. You give your information. And usually you give your information because you want something in return. Uh, I want to I want to get a quote for this service. I want to purchase this right now, but I'm not sure how or I have some questions that I need answered. So, speak to lead is calling them right there, right? Then while they're the most likely to buy that they will ever be, the voice agent answers their questions, books an appointment, and moves them forward in the sales process.

13:42 This is fantastic for businesses running ads or focusing on SEO for the website. And then finally, also industry agnostic is database reactivation. There's so many businesses that have lists and lists. Maybe they've run ads, maybe it's website forms. There's so many sources where they get leads and then just don't call them or contact them 2 3 days later and they already went cold. But the list becomes so long that if you call them even if they're cold, even if just some of them still want your service or just happen to want their service now, um it is essentially free money for your client.

14:06 And we are going to talk about how do you approach different businesses with these solutions and how do you choose the right industry for you. And this is why we're building these swappable components. You can choose your voice, choose your LM, modify your system prompt, add new tools. And what I'm going to show you the rest of this video is going to make you super versatile as a builder. And as I'm building, I'm going to call out where you can swap pieces for your use case in your industry.

14:52 So, now that you understand how voice agents work and what we're building today, uh let's look at the checklist, why now is the perfect time, anatomy of a voice agent, inbound versus outbound, and our use case for today. Let's look at the best tech stack for you to start. And this is what I recommend. This is what we're going to be building with today. Retail AI for voice agents, NAD for automations, for integrating with the CRM, calendar booking, etc.

15:15 Then go high level is going to be our CRM where we're going to have our leads and then calv.com is what we're going to be using as a booking software. We're actually going to be connecting these two. So anything booked here also shows up in gohead level. This is part of where the subsable components come in. You could use a different platform for voice agents or for automations or for a CRM or even booking software. Now, although there are many platforms that make it easy for you to get started, this doesn't only make it easy for you to get started.

15:39 It makes it so you build something that lasts. This 2 years in, this is still what we use for almost every client. Before I show you how to set up each platform, I want to show you how we're connecting them. So, let's say we have an inbound call. The user is calling our business and the call comes in and it's actually not retail that's going to pick it up right away. We're going to connect NADN first and what we're doing is asking go high level because Naden is our integration platform or automation platform.

16:02 We're asking go high level hey do we know this user and if yes we do know that user we can start leveraging what we call dynamic variables. Hey first name hey Allejo welcome back how can I help you today? We can even redirect to different agents. So instead of picking up right away, we find out whether we know this user or not. If we don't know that user, then we'll know that it's a new person. We need to qualify and book them in.

16:49 So, thank you for calling uh Amplify Family Law. Are you calling about a new case? Uh it'll probably be a yes. And then we qualify and book them through cal.com. So, these are two distinct agent uh agents, and we're going to show you how we handle that. Then we have the the use case for outbound which is the payment reminder. So this actually gets triggered through go high level when there's an unpaid invoice for more than 30 days.

17:12 Then we will trigger a call that makes retail the voice agent call the user and then we have a conversation. It starts with hey Allejo I'm so glad that your issue related to issue is that is finally resolved. I see here that there's an overdue invoice. would you be so kind to settle that with me right now? They might be able to pay that right now. And what we're going to do is send them a text message through NAN with that payment link so they can settle it right there, right then.

17:35 Or if they say, "No, I can't right now. We will schedule a call back." What is the perfect flow for scheduling this call back? That will depend on your client. They might want to call three times in one day. They might want to spread those out. That's where you have to get out of what we call builder mode and into consultant mode. Understanding their business, figuring out what is their brand. Is their brand calling three times in one day?

17:58 Cuz maybe maybe they're, you know, transactional. They did a good job. Now they want to get paid. Period. Some other companies are going to be more, you know, let's call it user friendly and want to spread it out. Give them patience. Overdue 30 days. Okay. Well, let's check back in and pay 45. So, understanding your client and how to build something that makes them happy is also what we're going to talk about in the second part of the video in the sales stage.

18:21 For now, let's jump into setting up each of