Build A Really Good ChatGPT AI Lead Conversational Bot In GoHighLevel In Under 20 Mins
Vložit
- čas přidán 24. 09. 2023
- In this video, I'll guide you through creating an incredibly effective appointment scheduling bot with message queuing using ChatGPT.
Want a more robust white-labeled chatbot that includes AI voice and image capabilities with 0 effort?
→ agencentric.com
Not using GoHighLevel? Don't be silly, get 14 days free:
→ www.gohighlevel.com/?fp_ref=a... - Zábava
Needed this. I’m going to play around and see how far I can get. Thank you 🙏🏾
You got this!
If you have any questions or want to walk through more in depth, feel free to reach out-
GREAT video! Very well thought out flow!
Thank you so much for the kind words!
This is such a great video, thank you so much for sharing. Did I miss what the checker flags are for and what should be inside that goal?
Great video, thank you. I have one question though, the main trigger is replied to; AI-Bot and the channel it came through. Where and when was the tag added? Did this come from another workflow? I'm new to the set up so I just needed that info. Thank you. Also, do you have the Conversation AI turned on to auto or something else?
Really needed this, is it possible to attach a custom SMS provider? API calls?
incredible video
Hi! Great video and well explained. Question - How can I create this same thing but initiate the outbound text instead of wait for an inbound trigger? Just set the trigger as a tag added like "aibot" and then have the first step be send Conversational AI SMS? // Or add Chat GPT step to drop into SMS ?
Thanks Jorden, I really enjoyed the explanations and prompting layout. Quick question, can every conversation be fed into a growing knowledge base to make questions from customers and the responses even smarter and smoother over time?
Not unless you're doing a fine tuned model. And at that, fine-tuned LLMs become expensive overtime because they have to pull from data that you give it. Best choice for keeping your costs down and mitigating these risk is through trial and error with your prompting. LLMs are context-based on the prompt so you can throw in key points to the prompt to eliminate some of that. Best thing is to use a cheap model like 3.5 (not as good as 4/4turbo depending on your OpenAI acct) but as long as you feed it enough context, itll do the job
Jorden I created the bot same as yours.. however the coversation stop when the leads got in the Send Message Workflow
Great video! First time someone addresses double texting. Do you offer this snapshot to your affiliates or any other bonuses? I’m interested
I dont do the whole affiliate thing - I think I have a link but I dont actively pursue it
I run a couple softwares that do these things, this one would need an update etc
Thanks for this. I get how it works but what i don't get is from the start, the initial messaging channel and how it gets tagged with the AI BOT tag before it even starts in the workflow. Something has to happen. "if they fill a paticular form then they receive the AI BOT tag", then it works" am i correct with that? then what about chat widget, does it work with that too?
Hey Chris, that is correct! Once leads comes in, send your follow up- add it to the conversation (for the AI to have context), and tag them.
I'm going to do a follow up video expanding on this with initialize bot and start bot workflows for situations just like this, ill send it to you as soon as I get it out
I created a form to do just that and it still doesn't work not getting responses from the bot.
@@jayd7974 I got it to work, it’s fine and all still putting my own tweaks on it
Hey thanks for the video! How does the actual booking work though?
If you want booking functionality, if you have access to highlevel's native booking bot- you can put in a trigger and listener workflow/ if/else for a keyword switch
Depends on how granular or if you have the cash to invest into something. 3rd party solutions are good for native conversational booking without being on the 497 plan but highlevel's is good enough... if you have access to it
otherwise, prompt the bot to send your calendar link after appt intent
Awesome can this be used to reply to cold email outreach from Instantly AI?
Not familiar... as long as you can webhook/api into highlevel, I dont see why not- just trigger a WF based on new contact tagged 'instantly AI' and when that contact is added, create a custom field where you feed that conversation logs for it to reply
Hey Jorden! What's the difference between the "AI Appointment booking bot" and "Conversation AI" workflow modules? We're having some issues with the appointment booking bot. Specifically,
1) One lead said "yes" to a specific time that the bot offered, and yet the bot proceeded to restart the conversation with additional times. An hour has passed in the text conversation. Our bot is set to timeout at 24 hours.
2) Why doesn't the bot first check if a call was already scheduled on that calendar? For example, if the user already scheduled a time before the bot was able to reach out, the bot should check if that person is scheduled on the calendar, and if they are, it shouldn't reach out.
3) How can we get the bot to stop chatting with the lead if a user takes over the conversation manually?
Good questions. HighLevel is good at what they do - but for stuff like this they are very high level (no pun intended). Now, I will preface, I've never used their native conversational AI (because we own a 3rd party white labeled ai plugin for highlevel) but it all about the data being fed to it.
I need a tad more context for these questions - hit me up here and lets chat: api.leadconnectorhq.com/widget/booking/CknL5ZwFA6lAhRHeTTpB
^^ thats my personal link
@@jordendubs thanks man! I just saw your other video from your Agencentric account. Subbed btw. That was great. I appreciate the offer. Will make a note to book when I’m on my computer after the holidays. Merry Christmas!
Merry Christmas my friend! Talk soon
Great video, is it possible for you to share your prompt for ai as an example to go by?
I can include an example - im probably going to film a video this afternoon going over everything including the workflow's and complexity of our whitelabeled highlevel AI
@@jordendubs Okay that would be great. But in the mean time could you put a link to your prompt. That is really the only issue I am having at the moment
@@jordendubs I am have the problem that it will send the whole text thread back to me and I can't seem to get it to stop
@@user-ig6qz7og1dwhat model are you using? And how are you structuring it?
@@jordendubs So I built the workflows the exact same way you have. I am just using the ChatGPT that is native to GHL. I didn't think you could pick which model you use. And I will have a conversation going with my bot and it will repeat the first message it ever sent then it will also include the my message. then have the response the last message. So my guess is that it is a prompt issue that I cannot figure out. It is also kinda hard to explain everything that is going wrong with it when I can't show pictures of it.
Does it do follow up well when the lead goes silent?
To do a follow up campaign, you can do a looping follow up workflow- so when the message is sent out to the lead, put the lead in a follow up workflow and when the lead replies- take them out of the follow up campaign. This creates a looping follow up campaign and make sure they are AI generated so they don't receive the same follow up if they ghost multiple times
Hi bro, I have an AI Whatsapp Bot on GHL with Zapier's integration to open AI's assistant API.. However, I need help building a workflow that for queuing double or triple texts. Is there a way for us to have a paid zoom call where you show me how to do it?
We actually have our system currently doing that with the assistant infrastructure - 0 workflows and assistants API
I'm happy to go over some best practices
Hey I built your bot but the queueing doesn't work. The moment I send another message it just loops up and keeps sending the same message over and over again... I have it set exactly like you.... What do you think it could be? Could it be the prompt structure or that gohighlevel workflow is not updating the custom fields fast enough?
Thank for reply! :)
did you manage to sort?
hi Jorden Can I copy your Prompt? Please
pipeline movement, I want that
Actually pretty easy tbh. Requires a little tweaking to this set up but basically what you do is you have a conversation analyzer that looks at msglogs and, with pre determined pipeline stages, analyzes the conversation and categorizes it. Once it categorizes it, put that in a custom field and run an if / else saying if custom field = intro, then change pipeline stage to intro- so on and so forth.
@@jordendubs wow thanks bro!
hey, great video, is there a chance you can share a snapshot of this? bw
Yeah I can try to get something working with that
Me please! 🙂↕️