Build Your Own AI Receptionist! | Using Bland AI & ChatGPT for Phone Scheduling

eventually AI is coming for all of our jobs but I think after this video the front desk receptionists might be on the chopping block that's because chat GPT and Bland AI can now handle scheduling appointments bookings or reservations all over the phone making sure that your appointments don't overlap or conflict this AI receptionist Works 24 hours a day 7 days a week without breaks lunches or even a day off and handles every call that comes in for your business with perfect professionalism oh and your new AI employee only wants $15 a month in salary instead of $5,000 a month in today's video I'm going to show you how to build an AI scheduling bot that can automatically check your calendar for availability schedule the caller for their desired appointment and log the person's information in each calendar event and you don't need to know how to code or have any previous knowledge to get this set up and it only takes about 10 minutes I have everything you need in the video description and I'll walk you through step by step so you can get it running for yourself now this video came hotly requested for many of you in the comments of my previous videos so if you have other suggestions for AI videos please drop a comment down below I try to respond to every comment and I get a little excited every time I see someone commented on a video if you like this sort of content consider liking the video and subscribing to my channel I try to break down how to set up and actually use some of the new AI technology coming out so if you want to keep up with which new technologies are actually useful for you in your business this channel is a place for you but let's get right into it so to give you a brief overview of exactly how this works so you understand each step after this we're going to use Bland AI to set a phone number number that customers or clients can call to schedule appointments bookings or reservations then we'll use make logic scenarios to check our existing calendar events and send them to our AI agent who's talking to the customer on the phone once the call ends the AI will send the transcript to a second make logic scenario that will get the details from the call and create a new Google Calendar event for you it's pretty easy once you see it in action and I have all the links and files in the video description so first this video is going to assume that you already have a Google account so that you can access Google calendar to manage appointments if you're not sure go to calendar. google.com and set one up the next thing you're going to need is a make.com account and I have a link in the video description so click that and it'll take you right to the sign up page once you set up your account I have what they call blueprints in the video description in my Google Drive folder I've created the flow and modules that you're going to need to get this working so you can load them directly into your own make.com account and then make some tweaks so that it works for your calendar there should be two files with a Json extension to download once you have those on your computer create a new scenario on make by clicking this button here next you're going to import a blueprint and the first one we'll set up is the event Checker so to do that we'll click the three dots here and go to import blueprint and then choose file and find the blueprint that you just downloaded and the first one we're going to do is event Checker so open that save and you see that it's loaded my modules and everything that I saved in the Json file all of these objects you see are what they call mod modules each module performs a specific task that you define and then passes the result onto the next module the first module here is what's called a web hook and it will sit and wait for a site to Ping or trigger it h and then it takes the data from that ping and flows it through the chain that we have here the web hook that's going to trigger this is actually Bland AI the AI phone agent program so let's set that up next if you go to the video description there's a link to sign up for Bland AI if you need help setting that up I have a video where I walk through setting up a bland AI account so check that out if you get stuck once you get it set up you're going to need an inbound phone number that your customers can call that's linked to Bland AI you can buy multiple phone numbers but they're $15 a month and you can cancel them whenever you want once you get your Bland AI account set up from the homepage you just need to go to phone numbers and then make sure you've got inbounds selected you can also purchase an outbound number if you want to set the phone that your phone is calling from I set the country code so they have the US and Canada um I'm going to do us and then you can actually set the area code and then select one of the numbers that you want to be your inbound phone number so this is the number that your customers are going to call so for this example I'm in Los Angeles I'm going to do 213 we'll do the first result here and then hit purchase it's going to ask if you want to confirm the subscription hit confirm once you have your own inbound number we'll set up how the AI agent handles any calls coming into that number first you can pick uh which model you want to use I prefer enhanced it seems to work better on the phone so select that after that you can pick the voice uh they have Bland curated clone voices or public voices in the public voices I've used Blandy 8 because it sounds like the character from her that's kind of cool you'll get used to it after the voice you can actually set the first sentence that the AI will answer the call with so in this example I've said thank you for calling dream Golf Course how can I help you today so for example today I'm doing this as uh a fictional Golf Course called dream Golf Course helping people that call the golf course schedule a tea time this next part here is the prompt for the Ai and I have this in the video description so feel free to copy mine for this example and then you can make tweaks so it works for whatever your use case is I've formatted this based on how bland AI sort of recommends to do their prompting so that it works best with their AI agents but the gist of it is that it's going to follow this script that I have at the bottom and it will use what's called Dynamic data and I'll get into what that is here in a second for now you can kind of read through this and get the gist of it pathway ID um you're going to actually leave this blank for this example um this has to do with the conversational Pathways I opted to use a different approach for this example but you could use that as well but for this example leave pathway ID blank and then down here ranges I'm not going to touch and then actually skip down to the very bottom where it says Dynamic data so in this section you're going to hit the plus API request this is where we're going to link the make.com sequence that we just imported to the Bland AI agent so they can talk back and forth for the URL we're going to actually go back to the make sequence that we imported earlier and click the web hook module and you can say create a web hook uh it's going to say what do you want to name it click save and then you're going to see this URL here copy that address to your clipboard and then paste it into the web hook URL in Bland AI I'm going to redo this one with our example here for the method put it as post um and then for headers make sure it says content type uh application Json now if you scroll down you'll see where it says variable one down here now variable one is defining the response that we get from make so we can use it in our conversation in this example Bland AI is going to Ping our make sequence to get the existing Events off the Google Calendar and then make will send back all the events off the calendar for the next 3 months then the AI can find a time slot that doesn't conflict with any of the existing events from the calendar so to define the busy times that we get sent back from make under Json path to data put dollar sign. dat and this is defining the path in the response to the information we care about now we're going to name the variable and I have it as Gore busy for Google busy time we need to give the AI some context of what this variable is so I have events on the calendar already scheduled for the next 3 months R and then double curly braces Gore busy double curly braces so if you do these double curly braces it'll actually put the value for the AI to read if I had a list of busy events anywhere I have the variable in double curly braces the AI will actually get the full output of that variable value now for this example we only need to pull the busy times one so I have the save response toggle here turned on if you leave this off each time the person speaks it will actually ping the web hook uh this can be useful if you want the AI to have the latest bitcoin price or active order status for example but for this example our busy times are not going to change while the person's on the call so I'm going to leave it as a saved response so it only does the ping one time great so now I'm going to save it now let's finish setting up the make sequence we imported earlier so go back to make and click on the Google Calendar module now you're going to log into your Google calendar so that it links your Google calendar to this sequence once you've done that we need the specific calendar ID for the calendar that we want this tool to look at now mine loaded actually my calendar ID so uh delete that and we're going to I'm going to show you how to find your actual calendar ID you go back to your Google calendar you'll see my calendars here and the three dots if you click that and then go to settings and sharing if you scroll down to where it says inte a great calendar there's a calendar ID here so you can copy that and then paste it into that calendar ID box and make and click okay so now I'm going to save it now we're going to test if this is working correctly or not hit this run once play button and then go back to Bland Ai and click test requests once you do that you'll see the information sent from Bland AI start moving through the modules in make until it gets to the end if it errors that means you need to adjust something and it should give you information on why it failed this little bubble here is the value of what was passed through that module and the green check means that it worked so the fact that you have these little green markers means that it got all the way to the end again if we go back to Bland AI we can see the response we got down here under the data hence the dollar sign. dat path from before you can see here it says accepted the status is 269 that's actually I set that in the end web hook you can see data uh is actually empty uh for the test hook there was no data that that make sent back to Bland AI so in this example the data here is blank but when we're checking our calendars you'll see in our example we're going to get dates of existing events and this will be populated with those with those dates and times great so the event Checker is working correctly so we can save that and then now we need to import the second blueprint from the video description this one's going to be the event creation blueprint so go to scenarios and create new scenario and just like before we're going to go to import blueprint and find the second blueprint here so event creation click save and now you're going to see the sequence here and the first thing I want to do is link my open AI chat GPT account to this make sequence so if you click the open AI icon here and click create a connection it's going to pull up this box here where you need to input your API key and your organization ID now if you need help getting your API key set up with open AI I have a video walking through exactly how to do that step by step and I'll try to link it here open a changed the website a little bit since I made that video but I'll quickly show you how to find your API key and organization ID here so go to open ai.com log in and then go to the API section go to API Keys create an API key going to call this make demo and then copy this API key and this is what you're going to plug into the make module paste that there now I need the organization ID so go back to open Ai and then click on settings and you'll see organization here and then you'll see your organization ID so copy that go back to the make window and paste that in there and then hit save and that should link your open AI account to make and it's going to populate with some information here that I filled in you can change this stuff if you'd like there's certain sequences where I felt chat gp4 was a better model to use when there was more Nuance to what I needed it to handle or a more complex task so this first one here the prompt I'm giving chat GPT is to pull in the transcript from the call that you have with the person calling the phone number so this is the entire call transcript once it's done it gets sent to this web hook the web Hook is going to remove just the transcript and then it's passing it into the open AI prompt here so just like before with the double curly braces it passes in the actual value so I'm saying from the entire transcript summarize the reservation or booking information and assume the year is now so it's going to be today's date and then uh the year in the format 2024 in this example unless otherwise specified uh that's because I realized that gp4 was trained until 2023 so sometimes if it just got you know the first part of a date it would think that it was actually 2023 so if they said like May 20th it would say May 20th 2023 and then it's creating an event a year before uh so I wanted to make sure that it's creating events in 2024 in this example or whatever the current year is so um unless otherwise specified then give the uh date in the standard format of month month day day year year year um and then it's going to give the reservation start time and give the caller's name and the caller's phone number as part of the prompt the AI agent is going to say when do you want this reservation for uh once the person confirms it it's going to say okay great reservation on this date at this time can I get name for the reservation and can I get a phone number for the reservation and this is basically just pulling that detail into a single block so this is just like the important details that we're trying to pull out if you click this next one we're going to go ahead and Link you're going to see it has your open AI account as a drop down now that you've logged in once click that and then you can see for this one it's a little less complex so I'm saying GPT 3.5 pars out the reservation date from the result of the previous module and include the day month and year in the format of this only the date must be a future date on or after today again this is to prevent it from thinking that it's 2023 uh and then I just put this sort of disclaimer that the only text result should be the reservation date in the format uh so it's not going to say thanks for asking here's the date like it's only going to give the the result as the as the date and again this is the reservation date that the caller wants so now start time again like your open the ey account this is going to read this and then get the start date that the caller wanted uh so now what we what we're basically doing is creating variables that we can then set as the details for the calendar event so now we've got the date that the person wants we've got the time defined as a variable here I'm going to get the caller's name in this one and then the last one here is getting their phone number out of the transcript so now we can go to the Google Calendar event and Link the Google Calendar that I logged into from the other sequence so you want to make sure you've got your calendar ID selected here for some reason it's defaulting to mine so make sure that you link yours there uh but then in the event name uh so this is in the Google Calendar event that I'm creating this is going to be the name of the event so when you're looking at the calendar this is what the title of that event is going to be so I'm saying the person's names tea time on the date at this time and then I'm giving the contact phone number so that you can contact the person and then I'm listing when this reservation was created so now you can imagine looking at your Google calendar you're going to see now in these where it says no title in the title is going to say Brad's reservation for May 20th at 9:00 a.m. and then his phone number and when it was created so now you just have some like information about the reservations that you're making in the calendar now I'm going to set the actual start time so this is how it knows where to put it in your calendar um and so it's going to be the date plus the time that we got from module 5 end dat you could put that in here you you'll add that module uh for this one it assumes that the event is 1 hour because how that's how I have it set up down here so you can set the duration you don't need an end time then you could add an open AI module that also finds the end time and put that in here if you want but for Te times I'm just spacing everything out as one hour all right so click okay and now we need to actually link the web hook to Bland AI if you click this it'll let you uh create a web hook you'll see this link here again go ahead and copy this address to your clipboard and go back to Bland AI so if you scroll all the way up here to where it says web hook right below ranges paste the event creation web hook in this section here so what this web Hook is is basically at the end of the call when the call ends the person's hung up and the AI agent hung up it's going to send all the call information to this address all the call information is going to have the transcript right basically what this module is doing is pulling in the transcript this array aggregator is pulling out just the transcript piece of the call information that we're getting and then pulling out only the transcript that it's passing then into our first open AI module and if you want to test it you can hit run once go back to Bland Ai and if you go to test web hook here then go back to here you should see it run so you can see it's going through um and then it's going to error out at the end here because with the test it's not actually sending like a proper call transcript it doesn't have all the information that we need for it to run successfully like the date the start time the person's name their phone number and everything so you can actually see what it sends here so this is under the transcript it shows you when it was created hey it's Bob's Donuts um how can I help you so this is the information from Bland AI being sent to our make module and then this is trying to parse out the transcript and then the final tool here defines the transcript as a variable that thing gets passed into the open Ai and unfortunately with the test web hook it doesn't work right these run successfully but basically they're just saying like hey you didn't give me anything hey you didn't give me anything and then it finally gets to the end it's like I don't have anything to create a Google Calendar event but when you're testing the web hook it actually this is this is a good sign when it gets to this point that means it's actually set up correctly save that so the next thing we want to do is go to our make profile and go to time zone options and make sure that you have this set to the time zone that you're in or that your calendar is in uh I had messed this up initially and I kept getting calendar events like exactly 3 hours after you know it was saved and I realized I didn't have my time zone set little trick there to uh help you out cu it stumped me for a little bit so once you have that set you can go back to the scenarios and then toggle these on and so now these triggers are waiting for a bland Ai call for the fine busy time time so now it's just waiting for an inbound call and it'll fire this and pull the information from our calendar uh and then once the call ends it's going to fire this one and it's going to take the transcript feed it into this uh get all the important information out of the call transcript and then create a Google Calendar event that's going to update on the Google Calendar that we linked so let's test this out and see it in action thank you for calling dream Golf Course how can I help you today uh yeah I'd like to book a tea time please great I can help with that what day or date were you thinking um let's do May 21st great let me check our calendar for available times on May 21st one moment please I found a few available times for you we have slots at 9:00 a.m. 12:30 p.m. and 400 p.m. which one works best for you uh let's do 9:00 a.m. you've selected the 9:00 a.m. time slot may I have a name for the reservation uh yeah it's James and a phone number in case we need to contact you yeah it's 123 555 1 2 3 4 great your tea time is reserved for 9:00 a.m. on May 21st under the name James we're looking forward to seeing you at dream Golf Course have a wonderful day thanks bye and then once the Callins you see our event pops up on the calendar at the date and time that we requested with the title containing all the information including when it was created the contact phone number uh and the name on the reservation all on the title but that's all there is to it and now you'll have an AI receptionist who can schedule reservations so that you can focus on other parts of your business or save yourself some money by not having to hire uh receptionist and assistant uh reminder again all the links and files are in the video description so make sure you get everything there if you like the video please consider hitting the like button it helps me know what type of content to focus on make sure you subscribe so you don't miss out on the next video and ensures that you're up to speed with the latest technology to improve you or your business that's all for now until next time thanks

Channel: Tech-At-Work

Views: 9,268

Keywords: AI, ChatGPT, GPT, GPT-3.5, GPT-4, GPT3.5, GPT4, LLM, Openai, ai data, artificial intelligence, chat gpt, gpt 4, gpt3, how to, open ai, tutorial, bland ai, blandai, aiphone, beginner, make.com, inbound number

Id: bwhBlqhbUrY

Length: 19min 28sec (1168 seconds)

Published: Thu Apr 18 2024