I wanted to start this video just by telling you how many companies people have uh contacted me after I started making these AI call agent videos, so there must be a market here somewhere. And I haven't tried to take advantage of this uh, but I just think if you are

watching this, you should have an opportunity here, not even kidding, right? Just because of the sheer number of people that have contacted me, and I have not helped anyone, so there must be a market here. But let's just go on to today's video. So in my previous AI agent

video, I mentioned I wanted to take this a bit further, so I kind of wrote a bit more code, and now we can actually do uh take inbound calls. So we can set up our agents to take in appointments and we can handle angry customers. So the customer in this case is of course going

to be me. So I'm going to call our number uh I'm inbound AI agent, and we're going to see how it handles, and we can also extract data from that inbound call. Uh on the outbound side uh, like I talked about in my previous video, we're going

to try to extract kind of rare data that you can't find online, right? So we're going to try to extract some hyper-local travel data by calling these yeah kind of small uh County bureaus, right? We're going to ask a trading card store for

like a specific Pokemon card, and I'm going to try to call a store, could be any store, to collect like this ultra-rare seasonal data. So in my case, I think we're going to try to call like a yeah, some kind of decoration store, and

ask what type of Christmas decoration do you sell most of, so this could be data you can again use right, so uh I think we're just going to go over to the code, but like I said in the beginning this is an opportunity so if you want to dive

deeper into this, I have like a tutorial on my member section and I'm going to put up the code there too, so if you are really interested in looking more into this, just become a member and yeah, take it from there, but now let's head over to

the cursor and take a bit more look at the code and kind of how I set this up and listen to all of these examples here that uh I have done, I had gone through this a few more times on my previous videos about this, but we are using the

OpenAI real-time API here, the voice part right, and since they made this with prompt caching, we are actually saving a much, much more money, so we're not spending that much, that is nice, and they updated the voices so now we have voices that I think act uh, acts a bit more, yeah

realistic, and I've seen that in the data we get more responses now since the previous voice was so robotic, this works much better and everything kind of uh is up to the system message here right, uh so we can start looking at the system message for an inbound agent right that

is going to answer the calls, so the first one I created is uh an inbound agent that is going to answer and take dental appointments for today, so you can see here your name is Julie, uh can we zoom in a bit maybe, yeah I think that's fine, so we have some

knowledge here because we need some knowledge so we have appointments that were available today, a list of treatments offered and the prices, so we have all of these prices here right, that is important. We have an AI Dental money return policy, so AI Dental is kind of the

name of the company that the inbound agent is set up to handle, right? So it has some information about uh return policies and everything. This is just mock data so don't don't look too much into it and we kind of have our setup here like our voice type, voice config, we

have an instruction, so kind of the caller answer, the caller with the questions, use your knowledge to help them. Introduce yourself as Julie, be very positive and the goal of this is of course to have the conversation with the customer and after that we can extract all this data. Uh we're going to look at

that after we have listened to our conversation where I call in to ask for an appointment, right? And yeah, so far the results have been very good so I think we're just going to listen to a conversation or a couple and then we're going to look at the data we collected

and then we kind of reassess before we go into the the outbound agents, right? Okay, so let's play this and then we take a look at the data afterwards, shall we? Hello, hello, this is AI Dental, you are speaking with Julie. How can I help you

today? Hello, Julie. So I had some issues kind of with, kind of, miscoloring on my teeth like from coffee and stuff, you know, that's the way it goes, right? It's so, do you have any treatments for that? Oh, I'm so sorry to hear about the staining.

we definitely have some options to help with that we offer like in-office Teeth Whitening that can really help with coffee stains the treatment usually takes about an hour and you can see a big difference right away would you like to know more about this treatment uh just the price I guess it

sounds pretty good oh amazing yeah for sure so the in-office teeth whitening typically ranges from like $300 to $500 it's a bit of an investment but like the results can be super impressive plus you'll like skip any hassles of do-it-yourself kits if that sounds good

would you um maybe like to book an appointment or if you have other questions I'd love to help I think that sounds pretty good do you have anything for this morning around 10 oh 10:00 a.m. today let me check real quick we've got a 10:45 a.m. slot

available this morning does that work for you if so can I have your name and number to secure the appointment yeah uh 10:45 sounds pretty good my name is Chris so c h r i s and my number is 5592 great Chris so we've got you down

for 10:45 a.m. today for the teeth whitening treatment if you need to make any changes just give us a call thanks so much for choosing AI Dental we look forward to seeing you soon bye for now bye Julie thank you bye

okay so that was not too bad right uh but now let's take a look at the data we collected from this conversation so you can see 65 F4 uh we have the caller name Chris appointment requested that is true that's a Boolean requested service teeth

When we have the preferred time, estimated cost, or income in this case, follow up with Anita. That is false because we don't need to do anything. And additional notes: caller is concerned about teeth staining from coffee. So this is some data that actually the businesses can

use right, because this could be like any like here we can take additional notes that are very important to the dentist that he needs to know just from the conversation. So the dentist kind of knows this when the patient comes in, that is helpful. And we have all this

other information and there's a bunch of much more information we can add that it could be specified to this business. So this is what kind of surprised me about the system, how well this data extraction works, and in this JSON format it's super easy to kind of set this up with

different SQL databases and we can just inject this data straight into the database right. So yeah, I think this is something that we're going to see a lot of in 2025, and I've been super impressed by this. But now I think we want to

listen to kind of where uh I answer a call to handle like the angry customer because that is something that happens right. It could be pretty cool if we can outsource this to these agents that don't get affected emotionally. So we're going to listen to

the conversation where uh I as a customer call in and am very disappointed with something that happened at my dental office. Okay, so let's just play this: hello, hello, this is AI Dental. You're speaking with Julie. How can I help you today? Well, I was there this Friday, so

that was the 10th of November right and my kid is killing me after your appointment so I'm very dissatisfied kind of with what you did for me now because it's hurting like crazy and I've been on painkillers all weekend so you just

need to fix this for me okay this is not good enough Oh I'm so sorry to hear about the discomfort you're experiencing Let's get this sorted out for you Can I schedule a follow-up appointment for you like possibly today so the dentist can take a look and help

relieve your pain Yeah of course I knew that but I'm not willing to pay for it because this was your fault right I completely understand your comfort is our priority and we want to make things right We can definitely look into this during your follow-up and our

staff will review everything to ensure you're taken care of without additional cost for addressing this issue May I book you in for an appointment today We have openings available Yeah dude 8:30 I already gave you my name and phone number All right we'll see you at 8:30

a.m. today There's anything else feel free to call us Thank you for letting us assist you and we look forward to making you feel better Take care Yeah bye bye Okay so yeah I I might cut out some of that I don't know if you heard that but

uh I think it did a pretty good job if you look at the extracted data here you can see we get a caller name Chris Johnson caller request scheduled for follow-up appointment without extra cost caller issue severe pain after dental appointment uh the sentiment is negative right customer

feedback on dissatisfaction with previous service. A followup appointment was scheduled, so again this try to handle customers that are not very happy. But this could, of course, backfire, right? So I think we have had to work a bit more on it, uh, on this, but I think it did a pretty good

job. And remember, we fed the uh return policies for the AI Dental store into our system message, and they kind of used that information to try to figure out how it should navigate these uh these claims, right? And this is something LLMs are very good at, to use the information

they have to make a good response. And I think that worked pretty good, right? So I think these uh inbound agents are working pretty good. Of course, there's room to improve, but so far my experience has been pretty good with this. And yeah, I tried it a few times, and every time I

think it did a pretty good job. So there's definitely some value you could extract from this, right? So I guess I haven't shown you this, but the way we make the call now is just to run this curl command here, and then uh we're going to

be set over to the websocket, and we can actually record the the conversation uh, but the best way to listen to it is actually to hear the recording afterwards. So I'm just going to make this curl request now, and you can see outgoing call in initiated. Uh, let's just

see now if we get a response here. Okay, so we completed the conversation. I think it went pretty good uh, but let's have a listen now, shall we? Thank you for calling the Shiaka County Visitors Bureau. This is Matt. How can I help you? Hi Matt, uh, I'm Julie. Um, is this

Chautauqua County it is yeah awesome So like what do you recommend for a visit in Jamestown in regards to like some attractions to go to or lodging or what? What kind of information are you looking for? Um, attractions? Yeah, totally like what, like fun things are there to do in

Jamestown? Okay, um so a couple of the bigger things, especially if you're going to be in the Jamestown area, um we are the birthplace of Lucille Ball, okay? So that seemed to go pretty good, like we don't have to listen to the all full conversation, we might do

some cool full conversations on my member page sometime, but you kind of get the point. Uh now if you look at the data, so you can see I want to zoom in a bit on this because this is pretty good, right? Uh, recommended activities, so we got visit Lucille Ball attractions as

Jamestown is her birthplace. Uh, recommended restaurants, uh The Chop House, a steakhouse in downtown Jamestown. We got Wicked Warrants, a brew pub offering pizzas and sandwiches. We got Sauce, known for deep dish pizzas, right? So these are kind of data we won't get just

by looking online, I guess we could, but here we can get like more data that is kind of rare to find that we kind of need, uh a conversation with someone, right? Uh so I think it's pretty interesting and uh again super good job by the agent.

I think it asked the correct questions. I think you got the correct data. Now let's move on to the one where we ask about a specific Pokemon card, shall we? So for this, of course, we have to change our system message again, so the task is going to be to ask if they have a gold

star Rayquaza from The X Deo set don't ask me about that I also put in some knowledge about this sort of Pokemon card right so we have some background information about that for our agent we're going to introduce ourselves with this we're going to ask for the name wait

for response ask about the card and say I probably stop by or no problem do you know where I can get one right so we're just going to try to yeah convert this and uh extract the data we want is the name of the store Pokemon card available

yes or no we could have done boolean on this but let's do string and other Poke card locations if they give us some information about this so let me just go grab the name of the store and let's try it so we fill in the name of the store

we fill in the phone number we are ready to actually do our curl command uh so I'm just going to run that and hopefully when we listen to the conversation now we have some information about this Pokemon card we completed the call now let's listen to a bit of it and look at

the extracted data okay thank you so much Julie and you're looking for the the gold star Rayquaza Pokemon cards exactly yes the gold star Rayquaza from the e deoa set do you happen to have one currently there are none of those on

the site so they're not available in store either truly sorry Julie oh no worries um do you like know where I might be able to find one maybe another store or a place I could check truly sorry you're not entirely sure I won't be able to best advise you truly

store Julie okay so that was not too bad she really tried to help but the connection was very weird I don't know it sounded very strange right uh but let's look uh at the data anyway here so we extracted uh we called bler trading and the Pokemon card was not available

so we found out that but we did not get any locations we could follow up on this but again pretty successful some issues with the connection I don't know if that was on their part but we could have just continued this tried a different kind of store or maybe done this in like a

parallel call or something but again uh pretty happy how it performed even though the connection was a bit strange so the final data we're going to try to find using this call agent is actually to call ahead and ask if uh yeah so the knowledge behind this is

you're struggling to find a gift for your sister you're looking for a popular Christmas decoration so what we're going to ask is they can help you with what is the most popular Christmas decoration they usually sell so this could be good for let's say you want to start some

kind of e-commerce store you need some research what kind of decorations you should sell on your store right and then you can call around to different shops ask what they usually sell the most of uh so yeah they're going to ask hello this this there we're going to ask uh is

the most popular decoration mention you're struggling to find a gift say goodbye and thank them and we're going to extract popular Christmas decorations and other information so I'm just going to fill this into our main and let's call up a store or something and see if

we can get this yeah kind of hard to find data right okay so we did actually call the West Elm that's like a decor store in New York I think uh let's listen to the conversation a bit uh let's look at the data we extracted

Chelsea West Elm hi there is this West Elm yeah in Chelsea 112 West 18th Street oh awesome so like um I'm kind of struggling to find a gift for my sister and I wanted to ask what's like the most popular Christmas decoration you guys

usually sell well the most popular thing let me see if they're still on the the most popular thing selling right now is like those you know now the what's trending are those red bows that go on the trees that's popular I mean there's so my God

there's so many popular ornaments that we sell okay so that was pretty good right and now let's take a look at the data we extracted here so you can see popular Christmas decoration we have red bows for trees most popular at West Elm uh the store is located in Chelsea

we got the address popular variety of ornaments though specific ornaments was not mentioned so yeah I would say that was well done we got some information about uh some Christmas decorations that kind of has like expertise knowledge from the guy that works in the store he was super nice by the way and he said

red bows for trees are very popular so if you needed that for your let's say your e-commerce store this could be an item you could feature right so I think there's some valuable data here like I said in my introduction I think there are some opportunities here to reach out

to businesses and actually start using this. And I also said, if you want to dive deeper into this, uh, I have the tutorial you can see on the screen here on my members page, if you want to become a member. Uh, I'm going to put up this code

here that kind of the updated code with the inbound agent on my GitHub, so you if you become a member you can also try this from there. Uh, if you don't want to rewrite the full code. Uh, but like I said, I think uh we did pretty good. If we look

at uh what we wanted to do today, I think we did pretty good on like one and two here, right? Uh, on three, yeah, I was happy with that. Uh, yeah, and the other two were okay, I think, not too bad. Uh, so I was pretty in

uh, so I am kind of, so I'm very optimistic about this, and like I said, there are some opportunities here to actually reach out, be a consultant to help businesses set this up, if you have the knack for that. So yeah, it's going to be super interesting

to follow this space, see how it evolves. We are definitely going to do more of this on this channel, try more challenging tasks than we did today and actually see if this is something that can extract some value finally from using these AI agents. So I think the

next step we're going to do is start introducing some tool-aid to this agent, so maybe the inbound agent can actually look up for me, and I can maybe use it like I can call the agent if I need some information sometimes. Uh, so that is

going to be something for the future. If you want more of this, just give this video a like, leave a comment, follow along, subscribe, I don't know, but thank you for today. Thank you for tuning in and I'll see you again very soon.

This AI AGENT Will Disrupt Industries BIG TIME

Welcome to my channel All About AI =)

Website:
https://www.allabtai.com

How you can start to use Generative AI to help you with creative or other daily tasks.

So I aim to bring Generative AI to everyone.

- AI Engineering
- ChatGPT
- GPT-4
- Midjourney
- Stable Diffusion
- Python
- AI Automation

We will answer all your questions about ai, genereal questions about the future of ai and the ethics of artificial general intelligence.

Follow our channel if you want to get going in this space =)



All About AI

많은 기업이 AI 통화 상담원 동영상 이후 제작자에게 연락을 취해 시장 수요가 있음을 알 수 있었습니다.

이 동영상에서는 인바운드 및 아웃바운드 AI 통화 상담원 기능을 시연합니다.

인바운드 통화는 약속과 성난 고객을 처리합니다.

아웃바운드 통화는 매우 지역적인 여행 정보처럼 찾기 어려운 데이터를 추출합니다.

자세한 내용을 알고 싶은 회원을 위해 튜토리얼을 제공합니다.

비용 절감을 위해 즉각적인 캐싱이 가능한 OpenAI의 실시간 API를 사용합니다.

시스템 메시지에서 치과 예약을 처리할 인바운드 상담원(줄리)을 설정합니다.

상담원은 예약 가능 여부, 치료 정보 및 가격 정보에 액세스할 수 있습니다.

대화를 분석하여 데이터를 추출합니다.

추출된 데이터에는 발신자 이름, 요청된 서비스, 선호하는 시간, 추가 메모 등이 포함됩니다.

인바운드 AI 상담원: 성난 고객 응대하기

불만족한 고객을 처리하는 상담원의 능력을 보여줍니다.

추출된 데이터에는 발신자 정보, 문제, 감정 및 해결 방법이 포함됩니다.

상담원은 시스템 메시지(반품 정책)의 정보를 사용하여 불만 사항을 처리합니다.

아웃바운드 AI 에이전트: 데이터 추출 예시

아웃바운드 통화는 curl 명령을 사용하여 대화를 시작하고 녹음합니다.

예 1: 카운티 관광청에 전화하여 지역 관광 명소 데이터 요청하기.

추출된 데이터에는 추천 액티비티와 레스토랑이 포함됩니다.

예 2: 특정 포켓몬 카드를 확인하기 위해 트레이딩 카드 가게에 전화하기.

추출된 데이터에서 카드를 사용할 수 없음을 확인합니다.

예 3: 가장 인기 있는 크리스마스 장식을 찾기 위해 장식 가게에 전화하기.

추출된 데이터에 따르면 빨간 리본이 가장 인기 있는 크리스마스 장식으로 나타났습니다.

제작자는 AI 통화 상담원의 잠재력에 대해 낙관적인 견해를 밝혔습니다.

제작자가 성공적인 데모를 요약합니다.

크리에이터가 상담원에게 툴 지원 기능을 추가하는 등 향후 계획에 대해 논의합니다.

크리에이터는 시청자가 좋아요, 댓글, 구독을 하도록 권장합니다.

This AI AGENT Will Disrupt Industries BIG TIME

스크립트