Leveraging AI

155 | AI for Efficiency: Real-World Automation Strategies to Boost Business Efficiency with Heidi Araya

• Isar Meitis, Heidi Araya • Season 1 • Episode 155

Share episode

How do you scale your business without drowning in inefficiency?

Running a growing business comes with a unique set of challenges: endless phone calls, lost leads, and complex processes that feel like they’re running you, instead of the other way around.

In this episode of Leveraging AI, we have Heidi Araya, process improvement guru and AI advocate, who has spent 25 years helping businesses of all sizes optimize their workflows. In this episode, Heidi shares a case study of a handyman who went from overwhelmed to streamlined by implementing an AI-driven receptionist and automation tools.

If your 2025 business goals include working smarter, not harder, this episode is packed with actionable insights and proven strategies to help you optimize processes, delight customers, and reclaim your time.

In this session, you'll discover:

  • The step-by-step process of mapping and optimizing inefficient workflows.
  • How to identify the "biggest pain points" in your business that AI can solve.
  • The secret to building a personable AI receptionist that delights customers.
  • How automation tools like Make.com transform processes across industries.
  • Why maintaining a human touch alongside AI is a competitive advantage.

 Heidi Araya is a process improvement expert who has worked with companies ranging from startups to Fortune 500 organizations. Leveraging AI and automation, she helps businesses unlock efficiency while keeping the human touch alive.

About Leveraging AI

If you’ve enjoyed or benefited from some of the insights of this episode, leave us a five-star review on your favorite podcast platform, and let us know what you learned, found helpful, or liked most about this show!

Isar:

Hello and welcome to Leveraging AI, a podcast that shares practical, ethical ways to leverage AI, to improve efficiency, grow your business and advance your career. This is Isar Meitis your host, and I've got one of my favorite topics for you to start the year of 2025. Now, those of you who have been listening to this podcast for a while, I No, two things about me and what I teach when it comes to AI. The first thing, that I'm a process freak. I really think that running an efficient business has to do with having proper processes and knowing them and allowing people to run through them in the most efficient way. The other thing that you know is that I'm teaching in my courses that one of the basic things you have to do in order to automate things with AI is to know the processes, because if you don't know the processes, it's very hard to know. What and how to automate. So with these two things in mind, I'm very excited about the show. And I'm very excited that this is the very first show of 2025, because this is exactly what we're going to dive into. Now, our guest today, Heidi Araya has been in process. management across multiple sizes of companies in the last 25 years. She's done process improvement in companies, large and small across varying level of sophistications. And now she's helping businesses do this with the help of. Of AI, which again is one of my favorite things to do because it actually does absolute magic for businesses. So if you're looking for magic in your business in 2025, and if you're looking to, for ways on how to understand what's happening in your company right now, and how can you apply AI to the relevant steps of the process, and you Can define what those relevant steps are, as you will learn shortly from Heidi. This is going to be the perfect episode for you to start the year with. So I'm very excited to have Heidi as our guest today. Heidi, welcome to Leveraging AI.

Heidi Araya:

Thank you so much. Thanks for inviting me. So excited to be here today.

Isar:

Same here. I, like I said, I'm a total geek about this thing and I have multiple processes myself, but what you're doing is on a completely different level. So I think you came up with a specific example, right? Cause this is not just a generic thing that we're going to talk about. This is going to be a real client with a real problem that we're actually going to review, right?

Heidi Araya:

Yeah, absolutely. Yes. I had someone approach me. He saw me at the One of the chamber events and said, I was talking to somebody and they said, I can't help you. Your process is the problem. And he said, based on what you said, I think you might be able to help me because you have this background and process improvement and AI. How can I use AI? and improve my process at the same time. So then we started having some conversations about what problems he wanted to solve in his business. So yeah, would you like me to dive right in?

Isar:

So before we dive into like sharing screen and stuff like that, let's work in a minute, what's the process, like the big process, like what did you do with him step by step in order to really approach this problem? Because I think that's one of the biggest issues people have. They don't even know how to get started. And then we can dive into the steps and see exactly how to do them versus what to do.

Heidi Araya:

Great. Sounds great. So his problem was that he couldn't answer all the phone calls he was getting. He has a ton of inbound phone calls and he can't manage them all. He can't respond to them. He was like, I just let leads go because I just, I can't even answer all the phone calls. And then they leave some vague message and I have no idea what they want. And so he was saying, It's just struggling to manage the phone calls among his other business and he had multiple channels. So it wasn't just phone calls. They are text messages, emails. He's got a web contact form. He's got WhatsApp messages, Facebook, and and of course the phone calls coming in. So he has got, he had multiple screens set up and of course he has to minimize and maximize and when I, so I went to his office. So the step number one would be understand. The challenge that someone is facing or that you yourself are facing, AI is only here to solve some problem. And so I met with another guy the other day. I said, what is, he's like, how can I use this in my trolley car tour business? And I said, tell me what problem you've got. I don't have a problem. This is just a retirement business. I'm okay. I said, let me take a tour and then I'll tell you. but anyway, in this case, you have to really understand what problem you're trying to solve or what challenge you're facing in your business. So you can begin to have that conversation. So start there. And then once you have anchored on what the problem is, now you can say, tell me about your current process. So that's what we did. I, at first I went to his, so I started with a zoom call and talked about a little bit. I went to his office, I took a big old, board with stickies and I sat with him for about an hour and a half and I went through and I asked him questions to what happens is, okay, this happens. And then what happens next? And then this happens. Okay. What if this other thing, and this, so it's a, it's an art rather than a science. And so you have to talk through and think about and explore what happens if not, or this. So that's the kind of thing that we did. So after about an hour and a half, I had the board and I took it back to my house. I actually then ported it into Miro, which I'm going to show you the exact board that I put that I had afterwards. And I still had some outstanding questions when you're thinking through and you're realizing, Oh, we didn't cover that. We didn't cover that. That's all natural. We had a lot of conversations along the way. and then once you understand the process, you could then understand. What are you going to do to solve the challenge that you came up with? Cause now you have an understanding of the process. And so then the conversation moves to, all right, so what is the problem in this process that you're going to solve? So for him, he wanted to solve the problem of, answering phone calls was his main one. We could hold, we could put. Basically, a huge portal together and have them all integrated in one. He said, let's just start with the phone calls. That's my biggest pain point. That's another suggestion that I have for folks. You can't solve every single problem all at once. It's going to be hard to transition your way of working all at once to some brand new way. you could do it. And that would be good and it might save you some money, but sometimes the best thing to do is just start with one problem. You're trying to solve the biggest problem and then you can add on and see and evolve as you go along. So that's exactly what we did. I created a, an AI receptionist and then when people call the number, it goes to the AI receptionist. The AI receptionist asks questions about the job for this handyman. tell me more details. What city are you in? What is your name? And then send me photos to either my email address or text me the photos so that by the time he's able to review the messages, he has already a good idea of what the estimate could look like. So it saves him a ton of time.

Isar Meitis:

We have been talking a lot on this podcast, on the importance of AI education and literacy for people in businesses. It is literally the number one factor of success versus failure when implementing AI in the business. It's actually not the tech, it's the ability to train people and get them to the level of knowledge they need in order to use AI in specific use cases. Use cases successfully, hence generating positive ROI. The biggest question is how do you train yourself? If you're the business person or people in your team, in your company, in the most effective way. I have two pieces of very exciting news for you. Number one is that I have been teaching the AI business transformation course since April of last year. I have been teaching it two times a month, every month, since the beginning of the year, and once a month, all of last year, hundreds of business people and businesses are transforming their way they're doing business because based on the information they've learned in this course. I mostly teach this course privately, meaning organizations and companies hire me to teach just their people. And about once a quarter, we do a publicly available horse. Well, this once a quarter is happening again. So on February 3rd we are opening another course to the public where anyone can join the courses for sessions online, two hours each. So four weeks, two hours every single week with me. Live as an instructor with one hour a week in addition for you to come and ask questions in between based on the homework or things you learn or things you didn't understand. It's a very detailed, comprehensive course. So we'll take you from wherever you are in your journey right now to a level where you understand. What this technology can do for your business across multiple aspects and departments, including a detailed blueprint of how to move forward and implement this from a company wide perspective. So if you are looking to dramatically impact the way you are using AI or your company or your department is using this is an amazing opportunity for you to accelerate your knowledge and start implementing AI. In everything you're doing in your business, you can find the link in the show notes. So you can, you just open your phone right now, find the link to the course, click on it, and you can sign up right now. We're giving off the course. And it's valid through the end of the month. If you're listening, you are still in luck, go and sign up quickly. Also, we have a limited amount of seats. It's on a first come first serve basis. So grab your seat now right now the promo code is LEVERAGINGAI100 And now back to the episode.

Isar:

First of all, this is really exciting, right? For, because there's, the problem you're describing that he has is a good problem, right? Having too many leads is a good problem to have, but it doesn't matter what the problem is, right? You may have a problem of not having enough leads. You can reverse this process and say, okay, I can do outbound calls with an agent like this to go after potential leads and so on, but I'm putting that aside for a second, I want to touch on a few things you said on, on. On things that can even help people accelerate some of that process. So you said the number one thing is identifying the problem and working through the kind of like really clarifying the processing. The minute we're going to share a screen and we're going to explain everything that's on the screen. For those of you just listening and driving, don't worry about this. But for those of you who want to watch this is also going to be on our YouTube channel. And there's going to be a link in the podcast to the YouTube channel. So you can, in the show notes, find where to go and see it. But what I do with some of my clients to help them map processes, I just ask them to do the process and preferably several times and just explain, narrate what you're doing. So just explain to me in your own words on the screen, or if you're working in a warehouse or cutting trees or whatever it is that you're doing. Just walk me through, explain in your own words, as you're doing the thing that you're doing there. If you have several employees doing this even better, have them record themselves, explaining as if they're narrating what they're doing to other people, and then you take these recordings, you upload them to an AI. and he will do, I do this usually with Claude and I'll explain in a minute. Why? And I asked. Claude to create a checklist from the process, a detailed checklist. multi leveled, as many levels of eat things as necessary. And then I ask it to ask questions that will help clarify the process better. And then I give it back to the people said, okay, here's the checklist. Here's the questions. What's missing from the checklist. What are the answers to the questions? I feed it back to the AI same chat. So it has the memory of everything that happened before. And I get a pretty good checklist. Then what I do, I. In Claude, what Claude knows how to do, which as far as I know, none of the other tools knows how to do is create flowcharts and it creates really good flowcharts, even multi layered swim lanes, like whatever flowchart you want. Now it's never perfect, but after. prompting back and forth six or seven times, you'll have a good solid starting point that will take you 10 to 15 minutes with Claude and will take you about an hour and a half doing it on your own. So you just save an hour and a half. And then what Claude does, it actually creates a code in the back to create that flowchart. So you can grab that code and paste it into third party tools, like Lucidchart, and like Mermaid, both are tools to create flowcharts. where you now have a very solid starting point and I can add colors and notes and whatever other things you want to add to make it a better flowchart. So this is just a few ideas. If you're not, by the way, my recommendation is to hire somebody like Heidi to help you through the process. But if you don't have access to a person like that, you don't have the money, you don't have the time, whatever the case may be, what I just said is a very good starting point for you to do things on your own. So. I'm closing this very long parentheses and let's dive straight through the process and see your magic and how it works.

Heidi Araya:

Yeah, I will do that. Before that though, I also want to share that a lot of problems in processes are across people. So this is great because he's a, he has six people working for him, but mostly he's handling this process and it's an end to end process that he's handling. However, there You often have processes that it's okay for you to have your own process and your process is fine. But the biggest problem is across departments and across people. So just be careful you're not honing in too much on a single narrow, process when you should be looking at the end to end process. I'm sure that's, I'm sure that you've come across that too. Okay. So this is the actual Miro board that we worked through. Again, I ported it from paper because I did exactly what you said. And I went to his office and I watched how he was working and took down the notes at the same time and chatted through it. And so I just want to highlight here that the sticky note colors are anything yellow is the handyman's manual action. The blue is from the customer side. And then I had some questions like orange. and then I have some other random questions here. So this looks very, very messy, but you shouldn't worry about how neat or messy it looks. You can always clean it up later. The most important thing about mapping a process is just getting it. Getting started. Start with that first sticky. It's going to seem overwhelming. And here in this case, I, again, I just move these around later, but you know, I see his text messages coming in. He's got emails coming in. He's got website contact form, Facebook messenger. I think next door messenger too. He said WhatsApp messages and phone calls. He wanted the phone calls to be handled gracefully 24 7. That was what we had started it at the end decided, but at the end, but in the beginning, we were just trying to get this process here. He said, I get a call or something and I say, hey, I need a job done. And the customer, he asked the customer to send the name, the city, the images. Customer sends the info he needs more details. Then he has to decide the subcontractors to send the job info, then he has to text them, then he's got to, get estimates from them, he's got to negotiate the price, he's got to, decide the price for the customer, so he covers his own expenses, send the quote to the customer, and then wait. Customer says okay, customer says no, or maybe the customer never responds. He then has to, customer says okay, he selects the final sub. And he has this additional process, which was decide which LLC it goes under, because there's that little complexity there. once he selects the sub, say Jane, he goes to create a work order for himself, and then he schedules a job, negotiates time with the customer and, text the customer saying, Hey, we'll show up tomorrow about two o'clock. then he has to tech this, the sub with the info. Now I'm going to pause here. This already is an extremely complex process. And he was saying. I'm always worried that I'm going to send someone to the wrong job or I'm going to do something where like they show up, two people show up at the same place by accident or I'm sending them to the wrong place or I tell them about the wrong job. So this is becoming a real nightmare and he has to continually just check his work. So he's explaining this to me as we go through, if customer says no, then maybe the price is too high. And then he tries to negotiate. if there's no response, he just, follows up a few times. So you can see here, he's how many times do I follow up? if they haven't replied. these are all conversations that come up as you're talking through, tell me about your process. Maybe you're watching them go through the process yourself. And, but imagine the customer says yes, you've got to negotiate a price, it's successful. And then the sub arrives on site to start the job. This is just to show that it's not his action. It's just the sub arrives on the job. The sub finishes the job. Now, once a day. He has to remember to ask each customer how it is going and I have used him So I know he does do that. He was checking in. How is it? How's it going? You know with my guys over there Then he has to communicate with the sub because the sub's also telling him what's going on So there's this loop there that he has to get back jobs finished. He has his messages. I asked the customer if they're happy. Customer replies. He responds to the feedback and then he asked the customer immediately how they want to pay. He always gets the feedback first before he asks how to pay and then he sends the invoice. Then the customer pays and then he asks for Google review and then he checks it and then he replies it and then that's his basic process. Um, so yeah, any thoughts? Yeah. And

Isar:

I want to pause you for just one second. this sounds crazy complicated, but if you go through any process you have in your company today, And really map it to all the steps. You will end up with something very similar. Like we have very few two step processes in our businesses. Once you start really breaking it down to what action, what's the mechanism that happens beginning to end step by step. You will end up with 15 to 25 steps. And you will end up with all these what if questions. what if the customer didn't answer? What if they didn't pay within 20 days? What if The thing that I ordered for the job didn't show up on that. there's all these what ifs in everything that, that we do. And even if you're handling stuff that doesn't have anything in the physical world, I can handyman. So you do accounting, you would still have all these what ifs, right? There's again, invoices that are not aligned with what the PO says. there's whatever. How to marketing don't even get started. There's so many what ifs in marketing. So any aspect of the business has very detailed processes. So starting with this, actually mapping the process is critical because that will lead us to step two, right? Now we can start analyzing what we can automate. So what happens now?

Heidi Araya:

So step two is I went through the process and I So I figured what could be automated of this. There's certainly some steps that he wants to himself and one of the ones he told me was like, okay, the WhatsApp is not that big of a deal. I want to keep that, but maybe we can automate other things. So I went through and I changed the color from yellow to this. Blueish purple color and thought, we can automate a lot of this, ask the customer to send name, city images, texting the subs that could be automated. He, of course, he needs to decide the price and the estimates, it could be once he did, though, the customer could be quoted automatically, have something go out once he has the price, follow up X number of times, that would be an easy scenario to, to do. It could even automate creating a work order, texting the customer. That would be easy. Texting the sub, that would be easy. I would never make a mistake. remind the sub about the job. It's say it starts three days from now, then they're going to work on it. then of course, asking the customer how it's going one time per day. That seems like good automation. Ask the customer if they were happy. So these things I thought are good. And in conversation with him, he said, for example, Heidi, sometimes the invoice The number is different because we're there and it might have cost more or less or maybe they asked for some additional work. So I can't send that automated. I just want to do that myself for now. when the customer responds, if they were happy, like I just want to reply to that feedback myself. So I certainly don't want to automate that, asking the customer how they want to pay is still, I've got just this easy thing I send. So he, he then shared with me the things that he still wanted to be the human touch in this. And I guess that's the other part here. Just because you can automate something doesn't mean that you want to, or that you should. And so he was very into this personal touch. He said, this is my personal touch. Like I, so yeah, I can automate some of it. And yet I want to keep some of it for myself. just for that personal touch.

Isar:

I want to touch on a few very important things that go back to the courses that I teach because they connect very nicely to some of the things you said, the first thing is the human aspect of business relationship, both inside our business, in our ecosystem, suppliers, subs, et cetera, as well as our clients and potential clients is going to play a much bigger role in an AI future. And those of you are not seeing the board, it's actually really interesting. So the board looks like a sticky note board, like a big white board with lots and lots of sticky notes. And we started with many yellow ones. And now in the new board, there's a lot more purple ones or about 50, 50 purple ones to the yellow ones, which means we're going to automate about half the processes that human factor is going to make the difference between you and the other company people are going to choose. And the reason for that, everybody, eventually in the beginning, it's a Huge benefit. Like you will gain market share if you learn how to automate, because you will need to spend less money to deliver the same amount of work and you'll be able to do more work, meaning you'll be able to be more competitive and you'll grow, but in two to five years, everybody will figure it out. This will be using Microsoft office. And Every, all these automation processes will work seamlessly for everyone. And so what's going to make the difference between you and the other guy? Human relationships. So I love that you, that your customers understand this, and I love that you touch on this point and I love it that you encourage and enable that to your clients while you're building automation. So that's one thing. The other thing that I want to touch, and it's actually not reflected here, but it's a very important concept when it comes to building automations with AI. And I call it, I have my five laws. Of success in the AI era. And one of them is actually called, stop thinking process and start thinking outcome. And what I mean by that is, yes, you have to start by figuring out the process. Like you have to, because otherwise there's no way around it. But once you understand the process, there are many AI tools that will allow you to skip three, five, 20 steps in the process, because they can do all of it bundled into one solution. and then you may not need the process anymore. The problem we have as business people that we're trained to think through the sticking up boxes, right? Like we, so many years, they told us to work the process. You've got to improve this thing by 5 percent and then this thing by 10%. And then your overall throughput grows by whatever your bottleneck is. That's the way it works. Yeah. The reality is now you can sometimes skip multiple steps. Like the example I always like to give is customer service. The goal of customer service is not better intake form. It's not a better alignment system to which ticket goes to which kind of agent. It's not, what automation system you have to make the phone calls faster or better IVR. It's not the goal of customer service is happy customers, right? that's right. And so today there are AI agents that can do customer service actually better than most humans. And so you don't need 25 steps. You need the good Yeah AI system that does customer service. That's it. Now that's not simple. It takes. A lot of money to put in place, blah, blah, blah. but if you can afford it, you can circumvent 25 steps in the process. So I'm saying that because yes, you can do step by step things, which usually is a lot quicker and it'll still get you a big value. But the real value comes from looking at the bigger picture and trying to understand what can AI plus automation replace.

Heidi Araya:

Yeah.

Isar:

And it doesn't have to be a single step.

Heidi Araya:

Good point. Yeah. it resonates with me also because I'm big with, value stream mapping. And although we're not touching on this topic today, it's what you shared, right? It's the outcome that you want to have happen. It's not necessarily the process is not the outcome. what is the purpose of the process? Interestingly, you talked, you touched on customer support. So certainly not this automation because he's getting leads, but the purpose of customer support, say you have a thousand incoming customer support calls would be to reduce that. Of course you can even handle that with AI. So I can handle all those customer support calls, but shouldn't your goal be to wonder why am I getting so many customers calls and. Can I go upstream and do something to save those, people from even having to deal with the customer support altogether. and yeah, what you said really resonated with me as well.

Isar:

I agree a hundred percent. And I will add one more thing to that because it's related and it's important for people to understand on this topic. Maybe the biggest gift that AI gave us out of the box without paying a lot of money to important fancy tools is the ability to do qualitative data analysis, which relates exactly to what Heidi just said, right? Let's use your example. You get a hundred, a thousand. Customer service tickets to analyze what is the common threads between those thousands is not doable by humans. Just not. Because these humans are busy answering these things, like you cannot go back and say, okay, let's look at a thousand, read all of them, try to find common threads. AI does it in seconds.

Heidi Araya:

Yes.

Isar:

And then you're like, Oh, okay. These are the three main things that cause us to deal with. Issues with customers. Let's go fix those upstream. Like you said. So it's another thing that you can do with AI as part of automating and analyzing stuff in your company. But let's go back. Let's go back to our handyman. And if you're not sure,

Heidi Araya:

just one last point. Yeah. If you're not sure how to fix those things that the AI found, you can always ask the AI, how can I address this in my XYZ business? yeah,

Isar:

absolutely.

Heidi Araya:

so interestingly. I, this handyman process when I said, I can actually automate 75 percent of this manual process. And he's I don't want to automate everything here. And, he actually said, I'm afraid. So if you AI, if I everything, I'm going to have so many leads that my guys will be. overwhelmed and I don't want them scheduled out a month because part of my value proposition is immediate service. You call me and he, I have seen that because I have used him. I called him today and he's I can be there tomorrow. I'll, and so he, he takes pride in being different than the rest. He said, I can only AI if I've got a few people at the ready, but I have to vet them and make sure that they're quality folks. I'm not ready. So I'm going to do that a few months down the road, but how about maybe we start with the voice agent here, just a, an AI receptionist. I told him I can build you an AI receptionist who will answer the call, get more details. And I said, I worked with them on the details. what is the city that they live in? What is their name? What kind of job it is and get details about the job. And then crucially, he said, I waste a lot of time because by the time I get the voicemail and I call them back, then I have to ask them to text over some photos of the job so that I can give them a quote. And what if I had it waiting for me there? So the AI receptionist that we chose to build is going to handle all of that. Intake so I'm going to share very briefly the AI receptionist that I built in make. com or sorry in Vapi. I'm going to share the AI receptionist that I built in Vapi and I named it This is a Handyman Hero. So I believe you wanted me to cover something here. it's our,

Isar:

Yeah, I think it would be interesting. So I know, first of all, I don't know if everybody knows that these exist, right? but there, yes, they're pretty good voice agents that sounds human, that can ask you questions and have a proper conversation. So it's not like the old IVR was like, if you want to make an S please press one. If you like, it's not, it's actually having a conversation with you like a human. Person would, and it's, in many cases, hard to know that it's not a human having a conversation with you. But I think the biggest question people have, even people who know about it is how do you control what it's going to say? How do you provide it the right information to ask questions? the right questions to collect the right information and not to say dumb things that it shouldn't be saying. yes, we'll be there in six minutes. Just

Heidi Araya:

don't,

Isar:

we'll wait for the ring in the door or something like that.

Heidi Araya:

Yep. Actually people have told me, so this is, I'm showing my screen. That's VAPI. it does create voice. Assistance and voice agents. So you can build here an agent using the prompt. So you, just like you might do in ChachiBT, you build a prompt and you give it constraints, about what it should be doing and what it should not be doing. So for example, you can tell it if anything else other than you can only reply to business, questions related to the handyman hero business, anything else, say, I'm sorry, I can't help you with that. I have some interesting test scenarios there where someone, actually, I was testing it and I said, the cat needs to go to the vet and my own personal assistant said, this is in testing. I'm sorry, I can't help you with that. This is only related to Heidi's business. So I can't take the message and just refuse to take the message. It was really funny. So anyway, that's why testing is important. And that's why even though I can write a prompt for the voice agent, and you think it's so simple, look, there's just this interface here. There's actually quite a bit of complexity around training the AI voice assistant. So they don't keep talking forever about a topic that someone wants to run up your bill. And so part of the constraints here are telling it its role, the skills that it possesses, its objective, and, hey, if a caller asks to speak to Hector, who's the handyman, then politely inform them Hector's in a meeting, and you give them the tasks that they should be performing, extract the reason for the call, details that may be important for Hector to know, especially sending photos of the job to Handyman Hero and, the text number and, And, yeah, Always get the city of the collar. And so it has some rules here too. Again, So you're greeting the caller warmly. Actually, I don't know if this has it, but the output improved drastically when I told the AI voice assistant that he loves his job. And he's so proud to work for a top ranked handyman business in the local area. so that actually improved the output. And, so you give them examples, steps to follow. So the greeting, how they should be collecting information, give them examples of what good looks like, ask the caller if there's anything else you can help them with, and then closing the call. And so then it helps more if you have some sample conversations. So here you go. I'm Joe Handyman here as AI assistant, all calls are recorded, Hector's in a meeting right now. How may I help you today? So you give these kind of examples here. For the assistant as guidelines, that's going to help it even more stay, as you say, like on the reservation and not go off and, out to the cafe down the street and just make up stuff. So that is, that, that said, if you get a phone call, that's an unexpected scenario, you have to learn to train the AI to adapt to that as well. So just. There's a lot of training that goes into that. And testing is huge there. So just get random people to call, ask it off the wall questions. One time I took, for my own Ingrid assistant, I said, tell me about Heidi likes butterflies. So tell me about butterflies. And this went off and talk, start talking about butterflies. And, so I was like, okay, so that's part of the training. Cause it got around it by saying Heidi likes butterflies. and then it was happy to talk. So the other thing that I learned over time is to. Have a call cut off. So if somehow someone gets around the training that you did, then have a maximum call duration. If someone can't tell you what they need in three minutes, whatever the average call duration that you have established, then you should just cut off the call. And that's going to save you from them from going too far down the line. So yeah, I wouldn't just have like really long conversations allowed in your AI receptionist because most calls are just shorter. So I've learned over time that those are things that should be done.

Isar:

So quick question, that I'm sure two quick questions, I'm sure a lot of people are asking one is what is the cost of this thing? And I actually know the answer, but I want you to answer and two is what's the practical setup? Does it connect to my business phone number or do I need a new number or how does it meet a website? Like, how does it work?

Heidi Araya:

Great question. The cost depends on the setup that you have, and you can vary the model and the kinds of voice you have. And so some voices and models cost a little bit more. you can see here a model for a mini, and the providers open AI. But if you tweak those, it might be, 11 to 16 cents per minute or so. And. what was the other question? By the way, the

Isar:

cheapest they have is 0. 05 per minute. I think it's worth paying a little more, but that's like to, for people to get the general bandwidth, it's like 0. 05 to 0. 20 a minute of call.

Heidi Araya:

yeah. So it is charged by the minute. Yeah. And sorry, what was the question? Other question you asked?

Isar:

how, does it connect to my regular business phone number?

Heidi Araya:

You can forward your phone number. To the phone number that you would get here so you have to buy a phone number here It's a couple dollars a month and then you forge your regular phone number here And then it can accept the call. So that's how I do it. that's all I'll add.

Isar:

So yeah, so the way this works, this gives you a phone number that you pay for and it's negligible. like you said, like it's a couple of bucks a month for that phone number. What you can do, especially if you have a Google phone number. So Google phone numbers allows you to set like as your business phone number, you can have a Google phone number and then you can set what if rules and you can say what happens in different times of the day, in different hours of the month in different, like whatever you want. if there's a call waiting, like you can define all these things in the Google voice and then you can decide which calls goes here, which calls goes to you. So when you're. Wife calls or a kid's call. The call still go to you and not go to here. Like you can do all that stuff. Just another two cents.

Heidi Araya:

Yeah. Okay. So

Isar:

now we have a voice agent. What happens now?

Heidi Araya:

Now I have a voice agent. So what happens? So someone's going to call the number it's forwarded, but how then does the handyman get notified that he's got this call? That is the problem that, make. com automation solves. So make. com is an automation platform. And. It's become very popular, and it's very easy to use, and in my opinion, it's easier to use than Zapier, which I don't love that platform, but you can think of them kind of similarly. There are lots of automation platforms, and this one just happens to be the one that I like. it's very beginner friendly, but what you would do then is just make sure that you connect the voice assistant to the make. com automation. And I'll just pause before talking about that connection to say. There is nothing new with automation. It's been around for many years. If you remember the old phone trees, it's still a lot of businesses have those today or the, they'll say dumb chatbots, not the AI chatbots, but it, they're around a decision tree. So someone has programmed them to say one, two, three, four, five. If the person is calling customer support, one sales, two billing, three. Then you're like, but my option isn't in the list. So you get frustrated. That's what AI has now added because instead of just having this pre programmed decision tree, we've programmed the voice agent with some boundaries and guidelines, and we can now take AI and natural language processing to understand what the person is trying, is calling about. So AI lets us automate a lot of things that were never automatable before in the old school style automation and just makes it a lot, for, to reuse a term, right? Intelligent, smart. so now we have ability to do these things that have never been done before.

Isar:

Yeah. And two words about make in general is like a canvas where you can decide what's going to trigger a process. And then the different steps of the process, each step of a process have an input that comes from an external source or one software to another software. And then in every step of the process, you can look at all the outputs of all the previous steps and pick any information you want from them to make the process move forward. And what really I like to call it is the glue. of the internet, right? It allows you to take information from any software and move it to any other software. It's not really any software, but it's most likely that most of the software you use is already in there. And if not their workarounds, but for a beginner, it's very easy to connect your email to your voice agent, to your CRM, to your customer service ticketing system, to your Google docs, like whatever stuff that you're using on your day to day, a very easy to connect. So let's work through this process.

Heidi Araya:

And you need to know the, you need to know the process and what connections you want to make, right? Because otherwise it's just like a tool here that you will be like, why do I even need make. com? No, you need a problem that you want to solve. So come with a problem. The problem that we had here to solve is that's great that I built the voice receptionist, but how is he going to be notified? Of the calls that happened and so what you do here is, like you said, there's a trigger that happens. A call happens. Someone's just called and the voice assistant has answered for Handyman Hero. And so we have a webhook created. So here's a demo webhook. once you get this, you can just, add one and you create it and then you would go back and add it into VAPI so that VAPI now understands now we're connected and Here, I'll just go through this flow first. Then once you've received this, you have to then tell it. What do you, what are you trying to do here? So this is going to take the recording of what the customer said in the call. It takes the transcript. Now we asked it for the phone number and the output, and we're going to tell it to follow the transcript. So again, this is a prompt. We have a prompt in OpenAI that we're telling it, What do you want to take? What do you want to do with this transcript? And the transcript is Creating a summary. So that's what this automation does. It takes the transcript, it creates a summary, and then it's going to send an email to his email, notifying him that a call just happened and some of the key details that are needed so he can make that call back. so start with the caller's name, add the phone number next to the name. Add the city because that's very important for the handyman to know where they're supposed to be going, and then wrap the summary in these HTML tags for fun, we'll use the appropriate emojis. We didn't specify which emojis. So they, they're sometimes different and, adding it to strictly adhere to the output format. And don't make up anything along the way. So again, this is our prompt, right? Ensure you preserve the order of messages faithfully. Don't include the transcript if the caller hangs up without speaking because sometimes it would just hallucinate a transcript when that didn't happen. so again, these are just things that you will need to learn to prompt. And so here you can, see, That we're pulling information from the webhook that's receiving information. So without, going into too much details here, you can see that the webhooks pulling information. We're actually taking that information. We're telling it to include it in the transcript, and then we're giving it some little example, saying, you know, hi, I'm this, assistant, here's John, et cetera. So this is the automation. this is the AI part of the automation, right? And then the thing that happens here is just simply sending it to the email address.

Isar:

I think this is absolutely brilliant. And I want to touch on why this is so brilliant, because this is like having a human receptionist, right? The human receptionist will take the call. Hopefully he or she will take notes and we'll write it in a standard way that then it will be sent to wherever it needs to be sent. In this particular case, The handyman's email, but this in the same exact ease could go to your CRM and be assigned to the relevant sales agent based on the region based on the size of customer based on Whatever the case may be in the same exact thing. It could update a accounting platform in the same exact ease it could create a Google doc or a Microsoft word document or add a line to an Excel file or whatever the case may be. So I want to do a quick summary because I think what we did was, sounded really long and complex, but the outcome is absolutely brilliant. We started with looking at a problem. The problem was multiple steps in a sophisticated process that took a lot of time and actually lost leads for a, an actual business after mapping the process, we identified the biggest problem from. The business perspective, and you identify the biggest problem by one of three things. One is what loses you the most amount of money. Two is what doesn't make you the most amount of money. So you can either increase your top line or reduce your costs in the process. Or three, the thing that people hate to do the most. These are the three ways you identify stuff that is valuable for the business. And then in this particular case, it was answering the phone. It was solved through a voice agent. But then. That solves half the problem because, okay, now somebody takes the calls, but how do you know what happened? If not you taking the calls, then there's an automation process building make that takes all that information. Synthesizes it pulls out just important information. It creates a standard structured email that's being sent to the relevant person. And now this thing runs on autopilot. This can run. Once a day or 6, 700 times a day. And it's still going to work in the same beautiful level of efficiency as if you had the ability to scale your team of employees from one to 600 overnight, which is obviously not doable in real humans. And so it's. As I mentioned in the beginning, this is absolute magic. Heidi. first of all, this is absolutely brilliant. If people want to follow you, if they want to work with you, if you want to learn more about what you do, what are the best ways to do that?

Heidi Araya:

BrightLogic. ai or my website, BrightLogicGroup. com. so yeah, you can find me anywhere. Please reach out to me. I'd love to know if this was helpful and useful for folks. And I look forward to hearing feedback.

Isar:

Thank you so much. As I mentioned, absolutely brilliant. I think what you're doing is fantastic and I really appreciate you joining us and sharing all of this with us.

Heidi Araya:

Thank you so much.

People on this episode

Head Shot

Isar Meitis

Host