How To Build Useful AI Voice Bots In Minutes With Make.com & Bland

what's going on guys it's Nick and I got a video for today on conversational AI agents and make.com if you guys have been sleeping on conversational AI I decided I would do a live demo or I'm getting a phone call from my appointment scheduler let's see how it goes hello hello this is Carl from left click I'm calling to confirm your appointment with us tomorrow at 9:00 is that still a good time for you yeah 9:00 works thanks great I've got you down for tomorrow at 9:00 is there anything else I can help you with today no that should be okay thank you wonderful I'm glad we were able to confirm your appointment have a great rest of your day Nick you tooe byebye okay so as you can see I uh set up a phone call in make.com to give me a ring and confirm an appointment for 9:00 a.m. I'll also show you how to jump in pull data out of that phone call confirm whether or not let's say an appointment needs to be rescheduled if it is resched what the time uh that the person you know suggested for the reschedule is and generally just do a bunch of cool with voice and make.com on the internet if that sounds like something you're interested in then stay tuned and let's get into it okay so the success of this hinges on a wonderful platform that has recently gained a ton of notoriety called bland. a and essentially they have an AI agent here which they are referring to as Blandy which I personally think they are anything but bland these guys are hilarious um and then you can just send or receive phone calls using their service it's a oneclick sort of API integration where you just grab the API key dump it into make.com a couple parameters that we might have screw around with if you're unfamiliar with apis that's cool I'll walk you through it and at the end you can have like a conversational AI agent that like goes out asks for things collects information and then you can actually just analyze that information and use it to like update a spreadsheet or something too so super cool uh we are currently sitting on the cusp of like an Revolution and so I think gaining knowledge and skill in this area is probably one of the most important impactful things you can do uh and yeah I just figured it was like right place right time so let's dive into how to actually build out a solution using this service and what I'm going to do is I'm just going to pretend that like I have a big Google sheet with phone numbers we're then just going to call the phone number and then maybe we'll have instructions for each phone call as well and then based off of the information that phone call maybe we'll do some else something else some update kind of depends how I feel and how long the video is all right all right so without further Ado first thing you got to do is jump on a bland. a um if you've never used this before obviously just going to start for free It'll ask you to put into much information including your first name last name email phone number all that fun stuff so when you do it they'll give you a couple dollars in free credits just so that you can start sending and receiving and sort of be wowed by how amazing the technology is um after that what you're going to want to do is head over to the home you'll see a homepage kind of like this and then all you need to do is just go down to um your little profile then go to API Keys then go to API key here and then you have your secret key and this secret key is just all you need in order to interact with the Bland API so super easy uh you can copy it and then you know use it in the flow that I'm about to show you but yeah that's about that obviously you know this is really exciting technology I don't have the time to cover every single one of these tabs but uh there's a lot that you can do here you can like test out flows you can set Pathways you can control the flow of a conversation a lot more if you don't want it just to be like a random call like I just did um there there's a lot you can do so definitely go in and check it out but what we're going to be doing is we're going to be using this Google sheet here which has it's just going to be uh two columns one with phone numbers and then one with instructions and then uh we're just going to go in and then for every Row in this sheet we're going to pull the data and then have Bland give that phone number a phone call with the instructions so you can imagine how you know you might be able to build out a much more complex flow like this where you have some CRM and the CRM is responsible for doing something I don't logging call information or seeing where a person is at a specific step in the pipeline what you could theoretically do is you could take that CRM data like the whole update log like called Peter on such and such day he said this you know Peter seems a little ambivalent going to follow up with him on Friday you could feed all this information into gbt 4 and then have it just like summarize um into a list of next step instructions what you can do then is you could have that next step instruction go right here on this Google sheet maybe in like an instructions column uh and then you could use that to like give the the prospect a call and maybe follow up or you could do it much simpler you could just have you know an integration like I set up just now that calls people to confirm um times that they're available or or whatever I guess the point I'm making is that the potential here is quite Limitless completely honestly uh which is pretty sweet so anyway uh the instruction that I'm going to post here is you're an appointment schedule for left click ask how the client is doing and confirm his appointment tomorrow at 9:30 a.m. if he's notable to make it reschedule and then I also provided just some some availabilities over here then wish them a nice day pretty straightforward pretty simple the cool part about this is you're just kind of coding in natural language now we don't need to be like a natural language proficient you know AI engineer with 50 years of experience here uh we can literally just write what we want it to do and it'll probably do a pretty reasonable job so that's pretty sick now we're going to take this bland. a Google sheet and I'm going to use it as a data source to actually go and then build out the rest of this integration so all I need to do here is I have a scenario set up that's titled number one launch bland. call then I have a Google sheet um search rows module and what I'm going to do is oh I actually don't know if I set this up right let me just share it here and then send this to my email address first there we go I'll jump back over here and I'm just going to set this up so let me close this down uh we'll just pretend that this is my drive and then I think bland. I don't actually know I might have to refresh this let me do some searching Bland at a example there we go so I'm finding the specific spreadsheet what I'm going to do is grab the sheet name the table does contain headers because we have the phone number and we have the instructions then for now I'm just going to return one row and then I'm just going to run this module in isolation just to confirm that it works okay great so I just ran it I received a single bundle which is good I received a phone number with my personal phone number please do not hit me up uh then instructions saying you're an appointment schedule for left click ask how the client's doing okay great we have everything that we need now to do this h TP integration now in order to you know get the Syntax for this what we're going to do is we're going to go over to their API documentation which you can find at doc stop land. api-1 poost uh and then there's just a bunch of different things that you could do here um what I'm what I'm going to do is I'm going to go down to this V1 API reference then scroll down and just take a look at all of these API endpoints you could send a call Simple you can send a call using Pathways Pathways is like their way to Route calls depending on people's answer which is quite brilliant honestly um then there are some longer API endpoints like send a call you'll see it's not simple so we actually have a bunch of things we can do we can select a specific model we could change the voice we could do all this fun stuff and I'll actually leave that for you guys to try because I want this just to be an introductory video I don't want this to be like 2 hours but you can literally like change how quickly people how quickly the model responds the faster it responds typically the Dumber it gets but you can like go really far and really Nuance with us like you could do a lot here um treat it just like you would the open AI API um there are that many different features here way you can then analyze the call with API uh with AI and this is a really cool feature um basically what you can do is you can feed in a call transcript back into another AI module and you can ask it information so you can do stuff like um hey uh you know in my in my reschedule example you could say Hey you know did the customer want to reschedule and then if the answer to that is yes then you can also go in and then analyze and pull out okay what time do they want to reschedule for and then maybe it's like 11:00 a.m. you could then use that in make to pull that data and then use that to like update a freaking calendar right so here are basically Limitless I'll just show you how to use the send call Simple endpoint for now I'll kind of take it from there on the right hand side here we see like the curl um details and basically what we need to do is we need to call this endpoint https backb api. bland. A/V one/ calls and so this is going to be the AI API that we're going to call so I'm just going to copy this in and then I'm going to go over here and make an HTTP request module just pump that in and then I'm going to go down and make a request and then connect it this is the URL and then we have a bunch of settings that we need to change here if we want this to work so I'm going to go over here and just make this you know put this right next to me so it's a little bit faster first thing we need to do is we need to add a couple of headers and so the headers here are content type application Json so I'm going to go back here and I'm go to headers you can also uh just change this by using using this but I'm just going to try and be as like accurate as possible so just going to copy and paste these headers in and then there's also an authorization string that authorization is going to be our API key which I have up here I'm just going to copy this in and then I'm going to paste this and now we have everything that we we should need to interact with the API oh sorry we actually need two more we need the data um oh and we also need to be post so we're going to have to change that so um we're going to have to change this method from get to post there we go and then we're also going to have to add in some data in Json format with phone numberers Task string which is great so what I'm going to do is I'll go to body typee raw content type Json application Json and then I'm going to paste in this and you'll see here that we put a string in the phone number and we also put a string in the task well if you remember we have that data from um our sheet and so what I'm going to do here is I'm going to use the phone number that I got here then I'm also going to use the task that I have under instructions and then um I think that's it yeah I believe that's all we need it may bug out because I've written content type application Jason up here and then I also have to select that down here so we'll give that a try we'll see how it goes um but yeah let's run once and let's see what happens let me just get my phone out just in case just in case I get a call from our AI overlords uh okay great it says uh status call successfully cubed we got the call ID here we got everything and I just received a call from Washington let's just pick that up and just verify that this prompt worked oh this is left click appointment scheduling how are you doing today hey man I'm doing really good thanks so much what's this about great to hear you're doing well I'm calling to confirm your appointment with us tomorrow at 9:30 in the morning does that time still work for you yeah I can actually make 9:30 could we do 12:45 no scratch that 1:00 okay no problem let's go ahead and reschedule your appointment to 1:00 tomorrow afternoon does that work better for you yeah sounds good book me in book me in okay I've gone ahead and booked your appointment for 1:00 tomorrow afternoon is there anything else I can help you with today NOP that's okay thanks so much I really appreciate it great I'm glad I could help have a wonderful rest of your day man the way that it says has have a wonderful rest of day it's like it's like secretly plotting to murder me and everybody that I know have a wonderful rest of the day anyway obviously that worked um keep in mind because we're now interfacing with the real world the testing Loop is a little bit longer so we actually have to go in deal with AI come back to this it's not as quick as just like looking at an output right uh but that that looks pretty good to me I'm pretty happy with how that went um so yeah that's uh that's pretty cool and I basically just already queued up the queued up the thing um I think just given the length of this video why don't I actually go in and then do the analyze call as well and just see how that goes so um now that I have the reschedule in why don't I figure out what time I rescheduled for the way that you do it is let's call this endpoint here with the analyze call with AI endpoint anyway so I'm just going to actually copy this paste this in this is the quickest way to do this sort of work um just copy and paste the uh the previous modules because odds are most of the other API stuff is going to be the same so content types the same authorization going to be the same we'll do goal uh and then we'll do questions so I'm a little bit 5050 on what the goal would be I don't really know the overall purpose of the call provides context for the analysis to guide out the questions transcripts or interpret okay yeah that's pretty simple so let's go down here U the goal is just going to be um identify if the call was rescheduled and if so what time it was rescheduled for write the time in military 24hour format so 13:00 IE 13:00 this should help um and then we have a couple of questions to be analyzed for the call so we'll do who answered the call they'll be kind of human or voicemail so oh I guess you can kind of Define like one or the other positive feedback of the product negative feedback of product this is like a customer survey example um but I'm just going to go questions let me just check this out questions this is going to be a an array looks like and then inside of the array the first question is going to be like um um was the appointment oh sorry I should actually write was the appointment rescheduled was the appointment rescheduled and then we'll have another item and that'll be like sorry let me just make sure I I got this format down true or false and then what time was the appointment rescheduled to and then uh it should just be a string okay I don't know if my formatting is right here so I'm going to go to Just Json formatter uh which is just a service that you pump in your Json and then if you click format beautify and it and if it goes through then you know that it's formatted correctly so it's just a quick and easy way that I space it out and make it look sexy um so okay we got the goal identify the appointments you're scheduled and if so what time is your scheduled for write the time in military 24 hour format IE 13:00 questions was the appointment rescheduled true or false if so what was the appointment rescheduled two string I don't actually know if this is going to work but I think it will the last thing we need to do is we need to grab the call ID um if you look back on the output of this then you'll see that one of the parameters was the call ID and so if we want to get this sorry we need to parse the previous module so I'm going to add a parse Jon module here and then I'm going to grab this data and then um I'm just going to do all this in one jump here uh there will be a variable called callcore ID and that's what we're looking for so this call ID will be um curly brace curly brace 6. callor ID curly brace curly brace oh there's some issue here with the formatting for some reason so let me just paste this up here 6. callor ID okay cool should be fine sometimes that happens with make.com um any who let's do that and then what I want to do here is I actually just want to test this first and I can test this because I have the old call ID of this run so I'm going to copy it in and then I'm going to run this and then my call ID is just going to be this big long string and so now it's actually going over the transcript and then it's essentially identifying whether or not uh this worked so okay status um success message successfully analyze the call answers are that the call was rescheduled true and that um it was rescheduled for 13 o' which is awesome so you can imagine how now if we were to put a parse Json module after this so copy this puppy paste that in now we'd be able to get the um the answer here which is true or 13:00 and then what we could do is we could conditionally route this and say you know or filter it and we could say hey if true and only if true then go into our CRM and update the meeting right so pretty exciting stuff guys you can do a lot with this and it's freaking awesome um the last thing that I would do is if you wanted to run this in practice you'd probably need a sleep module in between here um so maybe a sleep module and then you might want to run the I don't know you don't know how long this thing's going to be talking to the a ad I'll probably be talking about you call some people they'll be like my wife left me and it'll be like oh it's okay like what do you want to how can I help you right so maybe we want to like have this be 300 seconds or maybe we even want to do longer maybe you want to like call some other service wait for like 20 minutes or something and then you know do that extra request um I'll uh get the blueprint for you I'm not going to include my own API key obviously and I'm going to like um uh deprecate mine so you know you guys can't do that but let me just rename all these so that you guys just have all the information you need to do this so this will be call customer and then this one will be analyze call contents and then why don't I also parse this Json just so it's a lot easier for you guys to take let's do that okay hope you guys enjoyed that video uh whatever your personal ethical interpretation of the fact that you can now have ai call people on your behalf whatever that is let's just all agree that whoever controls this sort of thing and whoever knows how to use this sort of thing will grow undoubtedly very economically powerful in the future so uh it is probably in your best interest to pick up this AI licky split um hope you guys enjoyed this if you have any questions about maybe other ways that we could use AI uh feel free to drop them down below as a comment I love sourcing ideas from you guys I actually Source bland. a from somebody in my comments like quite a while ago and I just totally forgot about it so yeah you know if you guys want to see a video at some point in time I'll just add it to my queue and and get to it eventually otherwise please like subscribe and hope you guys have a lovely rest of the day cheers

Channel: Nick Saraev

Views: 6,547

Keywords: automation, make.com, content creation, ai content, google sheets, chatgpt, wordpress, openai, gpt-3.5, blogging, integromat, make, automating, automate, gpt-4, gpt, openai api, indie hacking, small business, $20K/mo, cold email, make money online, make.com for people who want to make real money, make.com money, make.com entrepreneur, make.com guide, make.com tutorial, make.com money guide, make.com course, ai voice bot, ai caller, ai call center, ai phone agent, ai voice, ai call

Id: pGAqubcGA9A

Length: 18min 56sec (1136 seconds)

Published: Wed Mar 20 2024