Activity
Mon
Wed
Fri
Sun
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
What is this?
Less
More

Memberships

AI Automations Unlimited

32 members • Free

AI Automation Agency Ninjas

18.7k members • Free

2 contributions to AI Automations Unlimited
🚀 New Demo Tool: What I've been building in the background
I used to do demos by actually getting people into WhatsApp conversations but I thought that was a bit clunky... And I never liked the OpenAI Playground approach because it wasn't something that was "on brand" So I started building something.. A demo tool to make life easier when on coffee dates and zoom calls. Here's how it works: - Give the bot a name - Set the company name of your prospect - Choose the bot type to demo (Database Reactivation or Reputation & Feedback) - Have the chat conversation - When finished you just click to "End" the conversation and it delete the contact and resets When you start the conversation it will create a new contact, set the details and add the contact to the related workflow. The workflows are the ones I'm using for my business and with clients. 🤖 So then you can just have a chat with the bot based on some default prompts for the types of bots. What I'm still working to add: - Ecommerce prompt and conversation (I've already got the web scaper working) - Speed to Lead for new enquiries - More options to customise the prompts for each coffee date/zoom call Once this is built out I'll also be able to set prompts for different industries/niches and re-use the existing workflows And the crazy thing is....I've been using AI and Vibe Coding to build this whole demo tool 😱 I'll share more details once I've added some extra bits and done more testing.
🚀 New Demo Tool: What I've been building in the background
0 likes • Oct 29
@Philip Wride okay you should check out codex IDE. It has access to your full project so it always has the full context, and can make changes for you
1 like • Oct 29
@Philip Wride yes you can disable full agent mode, then you would approve each change, or you can just do chat mode and it still has the full context. I do full agent mode but just ensure i have comitted to git before it makes changes
2 Key Learnings from a DBR Campaign....
[VIDEO NOW INCLUDED!] Last week I finished a small DBR campaign for a MedSpa in the US and wanted to share some learnings in case they help you with your campaigns. I'll also look to create a video for YT where I'll go into more detail about what I built.. But first.... Learning 1) EMOJI RESPONSES!! 🤣😱 I hadn't even thought about this but when going through testing one of the test users sent "Liked" the SMS from the AI.. And that meant in the conversation history in GHL the user sent back the same message the AI sent.. In the format "Liked {{message from AI}}" You can probably guess what happened?! The AI didn't know how to respond and so repeated the same question or message it sent before Here's what I did to resolve it.... - Created a part in a workflow where I'm recording the responses from the user - Using an IF/ELSE step to detect if the user message started with "Liked" or "Loved" - If it did, then I updated the custom field on the contact record with a manual update like "Yes sounds good" rather than the full contents of the message from the user (which was a repeat of the message from the AI) - If the message from the user didn't contain "Liked" or "Loved" the other branch in the IF/ELSE just updated the conversation as normal Learning 2) AI doesn't like the end of conversations Ever had a situation where you've added into the prompt the message for the AI to send at the end of a conversation and then it seems to get into a loop and repeat itself whenever a user responds? Here's what I did to solve it... - I thought about conversations "states" - like a story, a conversation has a start, middle and end - I created a workflow that uses a ChatGPT step to determine the "intent" of the conversation so far - It outputs a single word - "continue", "closing" or "stop" - That then feeds into an IF/ELSE step - if chatgpt thinks the conversation is mid-flow and outputs "continue" the user goes down a path that adds them to a workflow to create the next AI message in the conversation but if it think's the conversation is "closing" it will send a static message like "hope to see you soon" and then not do anything else
0 likes • Oct 27
Interesting. I also have pretty good experience in just specifiying what it means when someone "likes" the message in the system prompt as well, works mostly ok. Otherwise smart to have chat gpt analyse the state of the conversation, however in my flows i just tell it to allways include a keyword when its supposed to end the converstation, and never in any other context - also works ok.
1-2 of 2
Jesper Dovland
1
3points to level up
@jesper-dovlandcom-3392
Optimization consultant

Active 6h ago
Joined Oct 27, 2025