hey, thanks for the video. csn you please tell me how to do a automation on instagram where if someone wants the link befire that they have to follow me than after the automation should send the link. how to do that?
Glad to hear, thanks! Since that video I already learned some new things about handling multi-language assistants. Will make a new video again some time soon :)
Hi, thaks for your video. I am wondering, How do you connect GPT API to ManyChat , I mean, when you have an personal Assitant at GPT API, Is it possible connect it with ManyChat
This doesn’t work for me. Ios 17. Doesn’t matter how I try Siri just shows a definition for the word perplexity. Combed through the settings, there are no enable Siri integration anywhere like the internet suggests.
@@aronrad Then unfortunately I don't know an easy way for you to fix it. Given that the video is soon a year old it's possible that this does not work anymore, since everything with AI moves so quickly these days. When I have the time I'll look into this again and potentially make a new video - altho probably I'll wait until the new version of iOS is out, since that brings a big Siri update in general
¡Thank you very much! This was super helpful. I could implement it without any problem. I have a question. Is there a way to implement a more advance gpt? one trained with files, for example.
Thanks for the kind words. About your question: Yes that is possible and there are different ways to do that. One option is to use the OpenAI Assistants API, which supports a knowledge base An other option is to use VoiceFlow as the backend - in which case you also have other connection options to social media beyond ManyChat as well Or you could use tools like chatrace / chatbotbuilder.ai, which also have a knowledge base and integrate with these services well. I intend to make videos about these topics in the coming months as well :)
Nice video! some feedback: could've done it in 10 mins. I assume people coming on here to see if we can make multi lingual bots are mostly people who already know the steps you spent quite a lot of time explaining. either way thanks for making a clear video to follow nonetheless!
Thanks for the feedback! I do intentionally make these videos suitable for beginners as well. However in the future I will probably try to structure my videos in a way that the beginner part can be skipped more easily, so those who don't need that can more easily skip it. Thanks again for letting me know!
Great video thank you for this! I just wanted to ask if there is a way to cancel an appointment through the chat bot been searching for a while any help would be greatly appreciated thanks!
Unfortunately I have not seen any simple solutions that can achieve cancellations through AI Assistants. If you know the API of the scheduling service you use, you could of course program it yourself, but probably that's more effort than the functionality is worth. Perhaps in the coming months and years we will see easier solutions for this
This video just complicated things for me to understand and kept getting the 422 error and no matter what i do, nothing fixes this. "error": { "type": "INVALID_REQUEST_MISSING_FIELDS", "message": "Could not find field \"fields\" in the request body" },
Haven't tried that yet but I see no reason why it shouldn't work. When I used assistant last time it still had some issues (e.g. how it shows citations) - and given that it is still in Beta I am not sure those have been resolved yet. But if that doesn't bother you give it a try :)
Thank you for the video. I still have a question: How can we transfer information from Calendly or any scheduler into our CRM when we collect name, email, and phone number in a VF. I ask this because it might annoy customers to have to enter their information twice
Yes you need to pay to use any AI model via API (OpenAIs GPT or any of it's competitors), but if you use an affordable model (GPT 3.5) it can be very cheap. This is unrelated to the billing of Manychat.
You are a fantastic teacher. I love your clarity and presentation. I'm wondering the best way to copy/paste into MacOsPages, including your images and links.
@@sketchingbyyash6358 Interesting suggestion, but that is actually quite challenging. The main issue here is user authentication, which I currently don't know any good solutions for inside of chatbots. I'll consider making that video when I see technology that allows for it, but I don't expect this to work easily in 2024.
Thanks for this video! What if we have a script we want chat gpt to follow, like to take the lead through a sequence of messages to triage if they are a quality lead for our offer or not. Is there a way to do this with this manychat setup yet?
There are different ways to approach this Perhaps the simplest way is to do it with questions and buttons in ManyChat, which I also show you in this demo video. Then you can still use ChatGPT selectively at the end of the decision tree. Another way would be to try to use AI already in that stage, but I don't think ManyChat is suitable for that. Then you would probably want to use Voiceflow (probably together with ManyChat). This is slightly more complicated though, but one of my next videos will be exactly about this topic. Hope that helps
@@GregorSchafrothAI Thank you, is there a way to send information from manychat to the AI step? For example if the first response is from manychat and it knows the users name, to then pass that name to the AI step so that it can continue the conversation and it knows their name?
@@Joe_Leech Good question. So the proper way how you would want to do this in the long term is by using the OpenAI Assistants API (platform.openai.com/docs/assistants/overview) or a similar API from competitors that can remember the conversation. Unfortunately there is really no good support for that yet in other tools (even Voiceflow does not really support it at this point) That means for now the only way to do this is by storing the previous answers in a variable and then feed it in at the next prompt again. This way you can sort of build a small conversation with a few steps, even though it is a rather improvised solution. Not sure how soon the 'proper' solution works but we might have to wait until 2025, given the progress with the Assistants API so far.
This one does not have memory. To do it properly with memory you would need the Assistants API, but bot builders don't support that yet. I'll definitely make a video once an easy option for that is available. Meanwhile what you can do is insert all the relevant context in the prompt by using several variables. And you can of course also store AI Answers in variables, so then you could feed it in the next prompt again. I'd say if you need a solution with memory right now that is probably the best workaround.
Did you check if the OpenAI API is working otherwise? You might first need to charge it with a small amount of money to get it to work. If that is not the problem then I'd suggest you try disconnecting and reconnecting OpenAI to ManyChat. Let me know if you tried both of these and it still doesn't work
@@GregorSchafrothAI Could you please make videos on complex stuff like connecting shopify to voiceflow where after customer verification they can cancel the order, return the order or modify and also they can shop directly from bot it self. It would be really helpful for me and people who are running an ai agency
That depends on various factors If you are just starting out you might want to offer them at a low price or even for free first Once you have a good offer, you want to calculate the value that a bot can provide to a company and capture a share of that. Perhaps that's a good topic for a future video.
Hello Gregor im struggling to use the patch request in the chat bot the post request gets successfully triggered but the patch request doesn't work I'm thinking maybe is because of my record id i used. I got one from chat gpt but it seems is wrong i don't know maybe i didn't ask chat well in how to get it if u can please explain how u got yours or perhaps make a video about it then it'll be very helpful coz im sure a lot of ppl will also struggle with that part
You did a great job explaining this but I am getting an error for some reason even after trying many times to get the steps right. This is the error: { "error": { "type": "INVALID_REQUEST_MISSING_FIELDS", "message": "Could not find field \"fields\" in the request body" }, I went though the documentation with Airtable and Voiceflow but still keep getting the error. What could I be doing wrong? All fields are entered as you described...
@@paul7721 I think most likely the issue comes up because you copy exactly my 'raw' text. Your AirTable might look a bit different, so what you need to do is go to your own API documentation in AirTable and copy the code from there. (this documentation is really smart, it adapts according to your table). I mention how to open the API documentation at 10:36 and then at 13:40 the exact place where you need to take the code from. Try that and let me know if it works :)
@@DealingWithAB Sorry to hear that, I tried to make it as simple as a I can in this video. Did you try asking an AI Assistant (ChatGPT/ Claude) for help with your specific issue? I find that this can often be helpful.
That would take a lot of time, but I'll consider it in the future. Let me know if there is any specific topic you are most interested tho, then I could do that sooner :)
Yes you need to set up the OpenAI account and top up some money. To save money just go with GPT 3.5 rather than GPT 4. While not totally free, it is at least very cheap by now :)
I followed your instructions throughout this video. When I say "Ask ChatGPT" to Siri, it says "ChatGPT was not found in contacts". ChatGPT doesn't seem to have a good answer on how to fix this either. Is it as simple as one setting that I don't have turned on?