Which Is The Best AI Phone Caller?

AI phone callers have recently exploded in popularity enabling realistic AI phone calls that we can use to fully automate making or answering phone calls with this rapid growth of the industry there are now so many platforms that enable us to build these applications but choosing the right platform can be pretty overwhelming so in this video I will compare the top AI phone caller platforms so you can choose which one is best for you my name is Brendan and I'm from Australia and I run inflate an AI automation agency where I help businesses integrate AI into their operations these AI voice agents have two main use cases one is inbound calling and one is outbound calling so we can configure a number that when it receives a call the AI will start to speak with them and Converse with them and go through the flow that we've set up with outbound callers this is where we get a lead list and we actually send the calls with the AI to those lead lists to do a variety of things some examples of these inbound systems could be a dentist where you want to book an appointment and it simply just asks them for their name and then it is able to understand when they want to book and it can run an automation to book it in another example could be restant or takeaway food where you call up to order some pizza and then it goes ahead and takes that order and then sends it into their system an example of an outbound caller could be something like a salesperson that we configure and prompt the bot to become this you know really good salesperson that goes ahead and calls a lead list and then is trained to ask the right questions and then really just a book calls from those calls another example for the outbound system could be actually going through a lead list of existing previous clients calling them up with an outbound system and then asking them if they need any additional services so there's a lot of opportunity in this space and it's definitely something that you should be looking into many different platforms enable us to build these voice agents so in this video I'm going to be going through the top three most popular platforms that I've been seeing going around in the communities and then I'm going to take these three platforms and compare them to five different categories so number one is going to be realism that's how good it actually sounds does it actually sound like a human and can it convince someone that it is not a bot number two will be pricing so how much it actually costs at scale so that we can see whether or not it's actually affordable to actually use this in a business number three will be speed so seeing how quick it actually responds to our questions number four will be Integrations and ease of use so the Integrations are how easily we can connect it to something like zapia or make.com or just a variety of automations platforms so that we can run a whole heap of different requests from our voice agents and then lastly number five I'm going to go through the country compatibility of these systems so we're going to be seeing which ones are available within the US and outside of the us as well many different platforms enable us to build these voice agents so in this video I've gone ahead and got three of the most popular ones that I've seen and I'm going to be comparing ing these together the first platform is going to be vy. a the second platform is going to be synf flow and the third platform is going to be Bland AI so I'm going to get started on which one of these platforms sounds the best so if you're familiar with some of the AI voice stuff that's been going around for a while now you'll probably know that 11 Labs is one of the best sounding sort of realistic AI voice platforms currently on the market so 11 Labs has Integrations for both vapy and sylow where we can actually take out 11 Labs API keys and use them on these platforms and then Bland actually uses 11 labs in the background so it doesn't explicitly state that they're using 11 Labs but they are using 11 labs in the back end now since bland is using 11 labs in the back end this means we can't actually upload the 11 Labs custom voices and one of the big benefits of using 11 Labs custom voices is that they have sort of a marketplace where people can upload different types of voices that they've created themselves uh for other people to use now the big problem here without being able to upload custom voices is that for someone like me I'm from Australia so the Australian accents and using that to sell to Australian businesses is a pretty important factor and the fact that we can't take some of these realistic sounding Australian accents and get that onto blend uh is a pretty big shame so I'm going to start off by testing synf flow with a custom 11 Labs voice so the voice that I'm going to be using is Jacob De from 11 Labs so you can go ahead and copy that one if you want to use it as well gone ahead and connected this to synf flow as you can see here you can pretty much just put in your API key and get access to all of the custom voices from 11 Labs I'm going to go ahead and call it hey could I just get some garlic bread sh what size would you like for your garlic bread small medium or large uh just small if that's right got it will this be for delivery or pickup so that sounds pretty good for what it needs to do the connection with 11 Labs definitely helps out a lot to get that realistic sounding voice I'll now test vy. a which is probably going to sound pretty similar as it's using the same 11 Labs connection John's Pizza what can I get you today hey could I just get Margarita Pizza what size would you like for your margarita pizza just a small thanks will this be for delivery or pickup and could I also have your name please so once again they sound pretty much the same they're both using that 11 Labs assistant uh that really helps it to give that really realistic sounding voice however one thing I did notice is that vappy actually has the option to go ahead and turn on a background sound so this would enable us to actually have something like an office sound in the background so if I clicked on the office and then went ahead and published this hey welcome to John's Pizza what can I get you today hey could I get a cheese pizza sure what size would you like for your cheese pizza so hopefully you were able to hear that background sound there but that really helps to make it sound a whole lot more realistic now obviously when you're calling a pizza place it's not going to sound or have a background noise of an office but I did contact VY and they said that they are working on a feature where you can actually upload custom background noises so you can actually record you know the sound of a restaurant go ahead and upload that file um and have a really realistic sounding voice agent now I'm going to try Bland AI so this is Bland here and I've just gone ahead and selected the uh Sophie Australian voice but like I said earlier you only get the voices that they give you so they've got Public public voices as well as these Bland curated voices but you don't get to upload your own 11 Labs keys to get access to those custom voices so this is one of the restrictions of bland hello thanks for calling John's Pizza how can I help you with your order today hey yeah could I just get some garlic bread okay got it for your order you'd like an order of garlic bread is that correct yeah that's it great I've got you order for one order of garlic bread that will be $4 99 can I get your name and address to place the order so there you have it definitely not a terrible sounding Voice Assistant but one of the big issues like I highlighted I selected an Australian voice and that didn't really sound Australian that sounded more of a British accent and so that's going to be one of the restrictions of using Bland if maybe you're in a country where your accent is a little bit different so you're not going to be to get access to those customer systems where something like vapy and cflow gives us access to a whole range of uh different accents but obviously if you are in the US you can pretty much go with anything uh in terms of the accents so now in terms of which one was the best I would say that vapy having that background sound really helps out a lot I did contact sylow and they did say that they were working on getting a background noise as well uh so I think that's really a key thing when calling up these assistants to really just add to the whole experience to make it sound realistic and then on top of the fact that obviously we're able to use our custom 11 Labs keys to get you know hyper realistic accents and sounds coming through that's a really big benefit so the first point for realism is going to go to vy. so now on to the pricing of these systems so pricing is a pretty important factor we're building this for a business that gets an incredible amount of volume if we've got a really expensive system these charges can rack up really quickly and become quite an expensive thing over time so you should really understand which system should be used for which use case maybe the lower volume systems could be using more realistic voices whereas the higher volume needs to really tone down on the realism to save on prices so I'm going to start with Bland so Bland costs 12 cents per minute so this is 12 USD uh per minute and so this is a flat rate so it's 12 cents a minute uh to the exact second uh and that's just their flat rate it doesn't go up or down I believe it is 12 cents a minute and that's just blend flat rate cost so now on to vapy so vy's pricing can vary quite a lot because vapy actually enables us to change the AI model the transcriber so it transcribes voice into text to be sent into models it's got the ability to change obviously our voice providers and so because we can change all of these different variables the cost will change depending on that so at the moment I've got it set up to use GPD 3 and A2 so the open AI GPT model we're also using 11 Labs uh turbo V2 and so these systems put together comes up to 22 cents per minute but then obviously if I go ahead and change the provider to something like grck so it's not using open AI it will go down a little bit so now we're down to 21 cents but the biggest cost here as you can see in this bit of bar here if we hover over this 11 labs is150 cents uh per minute that that's being charged so obviously that's a pretty heavy cost uh if we go over to the voice function we can turn it off and go to play HT um we can see here it's now down to 16 cents I can go down to open Ai and we can actually get it down to 8 cents so I do like that we can change all these models and voices to really configure this to be more efficient for different use cases and we're doing something like inbound calls where people are calling up to get information out of the phone number obviously a more realistic sounding voice isn't really required because they're calling you for information but if we're doing something like an outbound call where you're being called you probably want the most realistic sounding thing that's pleas to hear uh to actually stay on the phone because nobody likes cold calling now moving on to sylow so sylow charges 58 cents per minute uh as their minimum plan so that's at the start of plan 58 cents per minute so very expensive uh to come in and start using sylow now with sylow you also have to pay $30 a month this does give you 50 minutes uh for that 30 bucks but then it's 58 cents per minute so 58 cents per minute very expensive probably not going to be the best thing to go out and run a scalable system on the starter plan they do have their Pro Plan which is $450 a month where then it comes down to 15 cents a minute so unless you have the sort of capital um and the volume already there to be requiring the $450 a month plan to get that cost down to 15 cents a minute and then onto the sort of 8 to 13 cents a minute plans with their growth and agency plans unless you've already got the volume to do that since flow is going to be a pretty expensive platform to go ahead and start using so for pricing I'd definitely say that bland AI has the advant AG here coming in at 12 cents a minute is definitely a very competitive price they've also got obviously access to that 11 labs in the background so you are getting access to 11 Labs but once again you're obviously not using those custom Voices All Through Bland AI but with vapy obviously being able to get your prices down to as low as 8 cents cheaper than any of the other platforms that is pretty valuable to have that sort of flexibility if you're working with a multitude of clients or you have a multitude of use cases beinging able to bring that price down to 8 cents a minute and then deploy that for different use cases will be a lot more efficient in the long run when building out all these different systems so I'm going to go ahead and actually give a point to both vapy and Bland eye because the two different use cases and flexibilities that they both provide now moving on to speed so how fast it can actually respond to the questions we give it I'm going to go ahead and actually Swap all the models for their fastest one so I haven't been using the fastest on so far I've been using the more sort of realistic voices I'm going to go ahead and so this is syn flow I'm going to go ahead and click on the low patience level so on synth flow we get the ability to adjust response speed so uh that's pretty convenient feature you can go ahead and click low I'm going to go ahead and put the time on how long it actually takes to get an answer back from these calls for each one and then you're going to be able to see which one is the fastest hopefully I'm going to be able to tell just by talking to it which one's the fastest but if not you're going to be able to see just on the screen so for vapy I've gone through all of the providers and found that open AI provides the best latency 950 milliseconds so we'll see if that's actually accurate but if we also have a look at their actually latency sort of chart here we can see that 11 Labs contributes the most to the latency at 500 milliseconds so once again we could just remove 11 labs and then reduce that latency like by 500 milliseconds so quite a lot there um but one again that's one of the benefits of vapy being able to be this flexible in removing 11 labs and then just using something else that's then going to speed it up a lot obviously when we remove 11 Labs the quality is then going to go down and bit be a bit more of a robotic voice for blend AI I've gone ahead and chose the turbo model which should be their fastest model so let's start off with blend hey could I just get some garlic bread sure how many slices of garlic bread would you like to order uh just two slices excellent do you want anything else to add to your order we have a wide selection of other delicious pizza options toppings and that's all right hey welcome to John's Pizza what can I get you today hey yeah could I just get some garlic bread of course what size would you like for your garlic bread we have small medium or large sizes available yeah can I get a large as well as a medium certainly will this be for delivery or pickup add for pickup got it to confirm you've ordered one large and one medium garlic bread for pick up hey welcome to John's Pizza what can I get you today can I get some uh Pizza of course what size would you like we have small medium and large just a medium cheese pizza sure the medium cheese pizza would you like any additional toppings on that no that's all good all right a medium cheesee Pizza it is would you like anything else with that maybe some drinks or sides so they have the speed test from just listening from that I do think that bland was actually really quite quick uh in generating those answers and I think vapy probably comes in about second or maybe even tied with Bland um there in terms of just listening to it I'd say there really isn't too big of a difference I did think that sylow was quite a bit slower um in generating those responses now number four which is Integrations and ease of use so Integrations is really connecting these systems to talk with external data uh as well as sending out automation requests so if somebody wants to order a pizza we can go ahead and then send that pizza to the kitchen uh and just stuff like that automations uh where we're connecting it to the system so for blend AI if we go down to the bottom of our sort of assistant configurator this isn't obviously going to be a full tutorial but what they do give us is a request data function so the request data says when you want to says when you want your AI to know a specific specific fact like the caller's name or other relevant context the AI agent will be aware of both the key names as well as their corresponding value so pretty much you just give it a key like name and then you give it sort of an example uh and it will just save if you want to capture an email you can type email uh and to really you're just capturing static values if you want to get back the order of a pizza so you want to know what they ordered how many they ordered you want to know if they've got a drink these are the types of things that bland is going to natively be able to capture so in addition to this Bland AI actually has their own zepia integration connection as well so if we go to zapia and then look up Bland AI we can see here we can connect Bland to obviously all of the 6000 applications that zpr has um and then we can obviously run a whole heap of different automations with that so if we wanted to schedule daily blend AI phone calls we can schedule it with schedule by zapia that's just a tool that they offer we can send phone calls in blend for new actions that happen in Glide that's just like a application Builder we can send phone calls for updated rows and Google Sheets so these are sort of the external automations for calling Bland to then go ahead and make sort of outbound calls so there's obviously a good amount of Integrations that can come from sort of a native zapia integration now moving on to vapy so vapy has something called functions now functions operate quite similar to how bland AI sort of data capture system worked but one of the key things here is that vapy can actually pull data from their functions back into the conversation so that we can make a call pull that data back and then continue the conversation so rather than just capturing some data throughout the call and then enable abling us to access that once the call is over we can get a real time function run during the call to pull back data from a whole heap of endpoints to make a much more advanced system so I'm just going to quickly demo something that I've been working on what I've gone ahead and built is an e-commerce uh sort of assistant chatbot that is connected to a table of products uh and what we're going to do is ask for a product we going to ask it to recommend us a type of product and I've actually connected it to this spreadsheet here where it's then going to go ahead and look up through this entire spreadsheet it's then going to pull back back products that it finds relevant to the query that I said in the conversation and it's going to give us the price it's going to give us the weight it's going to give us all this data that is stored in this database that we didn't store in the prompt benefit of doing this is that we can have this set up to connect with some maybe like Shopify where this is live updating with it and then we don't have to change the prompt with all the products uh and we're going to have a much more structured way of accessing this data hey welcome to eare what can we help you with today hey I was just looking to see if you had any red sling bags give me a moment yes we have a red sling bag available it's called the E wear sling bag natural hemp red the price is $ 23342 and it weighs 0.58 lbs it's a small size bag would you like me to send you the product link via email for purchase so as you can see there that is an incredibly powerful system where we're able to accurately pull back these rows of data through a function that pulls the data back during the conversation so in that call it was able to see that I was talking about the red bag here the red sling bag and it gave me the accurate price and other information from that spreadsheet of data let me know in the comments below if you want me to do a full build of an e-commerce a voice assistant chatbot with something like product recommendations like I just showed you uh canceling orders and emails and just stuff like that if you want to see that let me know and I can build a video for that and obviously in addition to running these functions where it pulls back the data live during the conversation we can have that data just sent off uh as a onetime trigger similar to all the other applications as well now moving on to synf flow so synf flow actually has quite a few options uh for sort of running actions configurations and just a bunch of different connections so go on synth flow and we're go to our prompt section these are several actions that are sort of pre-built for us to run during the conversation so one of their pre-built actions is sending an SMS so they can a customer might have a call and then we can trigger it to say send an SMS to this person uh after the call is finished so that's an automatic action they've already got we can have real time booking so that we can converse with it and then actually book A Time if I click on this it will show us that we can use cal.com or go high level to actually book a call during the meeting so that's a preset action we've got a call transfer so that's an action that allows us to transfer the call to a live Human uh and that would be just depending on a trigger if somebody says uh I don't want to speak to anymore I want to speak to the human it'll go ahead and Trigger that action and then it's also got the basic information extractor which is uh conversing with it understanding that you want a margarita pizza and then going ahead and just storing that in some sort of value then next to this prompt section we do have deployment and deployment gives us a whole another range of Integrations that are sort of native into the application one of them being go high level where you can actually install the sylow app uh into go high level like it says um and then there's just a bunch of different automations that we can run sort of natively without doing any setup yourself uh with go high level as well as zapia so they got zapia um you can put in your API Keys If I just scroll down a bit you can see that they've got popular with sylow AI user so sflow has their own zapia block just like Bland as well and you can configure this um natively within the actual website if you want to send emails and Gmail when the phone calls are completed in sflow we can just click this turn it on um and have that working so a whole heap of functions that are pre-built for us so in terms of the winner of this I'd say that they really all have different use cases uh for each other vapy is really great for the ability to pull data back during the conversation I definitely would say that for my use cases that's the most important thing the fact that we can pull in live data during the conversation really enables us to build much more advanced Bots but then equally as important Bland Ai and simflow both have their native zapia Integrations which is a really powerful tool to really allow us to build these sort of complex automations really easily so once again it really does depend on the use case uh for which one you're going to choose me personally I will probably be using vapy in terms of these sort of calling functions where we're able to get data live now lastly I did just want to mention the country support of these applications so vapy and cflow do support international calling and so that's a really powerful thing obviously since I'm an Australia not being able to call uh numbers that are Australian or call from an Australian number uh is a pretty bad thing we obviously want to use that if we're going to be selling this to businesses unfortunately Bland AI is a us only thing at the moment so if you want to go ahead and obviously call from different numbers and call two different numbers um using Bland AI that isn't possible at the moment I wouldn't be surprised if they rolled it out in the future uh but just for now you aren't able to access this um if you're not in the US so definitely something to consider if you are outside of the US if you want to get access to the international calling so I hope that was a helpful video in understanding the differences between these top applications and understanding their use cases some of their strengths some of their weaknesses and then figuring out how we can configure these systems with the speed and the pricing and the realism of the sound to be most efficient for our particular use case if you would like to see more sort of specific and complex builds of these assistants definitely drop a comment below of the sort of Industry that you'd like me to build bu one for maybe I'll go ahead and release a video on that as well if you're a business owner and you would like a custom system for your business you can go ahead and book a call with me using my calendar Link in the description if you found the video helpful definitely go ahead and subscribe to the channel turn on the notification Bell as well as like the video

Channel: Brendan Jowett

Views: 22,046

Keywords: ai automation agency, ai automation agency for beginners, how to create ai chatbots, how to make ai chatbots, ai voice agents, ai phone call, ai voice calls, chatgpt, gpts, openai, chatbots, ai chatbots, how to sell ai chatbot, synthflow, bland ai, vapi ai, vapi voice ai, voice ai, voice ai tutorial, automatic phone calls, liam ottley

Id: KloYd6cANkM

Length: 21min 13sec (1273 seconds)

Published: Wed Mar 27 2024