Activity
Mon
Wed
Fri
Sun
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
What is this?
Less
More
20 contributions to AI Automation Society
My WhatsApp sales agent works in testing but falls apart with real customers. How do I fix it?
I built a WhatsApp AI agent for a local clothing shop in Quetta. Customers text it in Roman Urdu like they would a shopkeeper, and it handles the whole flow: shows products, negotiates the price, collects name, size, quantity and address, saves the order to a sheet, and looks up order status later. What it can do - Replies in Roman Urdu, answers questions about products, shop location and orders - Shows products 3 messages at a time (to save cost), then the next batch if they ask for more - Understands photos/screenshots of a coat the customer found elsewhere and checks if we have the same or a similar one - Handles voice notes (transcribed, then answered) Stack: n8n webhook, code node to drop old/duplicate events, photo/voice to text, AI agent with 3 tools (catalog search, save order, order lookup), a code check on the reply, then the reply is split into text and real images and sent. The problem It works perfectly when I test it, because I know how to order properly. With real customers it breaks. People type casually, don’t know waistcoat terms, and the agent misreads what they want. It sometimes sends wrong products, and after a customer picks one item, it still sends photos of the other products it showed earlier. What I’m unsure about - Is this a system prompt problem, a catalog search problem, or an architecture problem? - Should I add a step that rewrites the customer’s message into a clean query before the agent sees it? - How do you stop the agent from “forgetting” which product the customer picked? I can share the system prompt and a failed chat if that helps. The template is "Free" to grab. Go check it out and let me know what could be changed or improved...
My WhatsApp sales agent works in testing but falls apart with real customers. How do I fix it?
0 likes • 2d
@Dylan Murray more on the roman urdu part bcz for negotiation i have added 5 step price for it, it doesnt go below that...
1 like • 2d
@Leon H Thank you for spending your time to check it, really appreciate this deep analysis of the workflow. you have found out mant major and tiny problems also which i missed. this will definetly help a lot.❤️
Clients Want to Manually Msgs the Customer...
i built a Whatsapp AI Agent for a local traditional outwear cloth shop which sells Waistcoats and coats. The Agent answers product questions and sends real photos through Cloudinary,Helps customers,Negotiates like a real shopkeeper with a floor that protects your margin, Collects full, clean orders every time Item, color, size, quantity, name, delivery address, and the final agreed price in Google sheets, Built in N8N. now there is one problem,he liked it, but the client wants an ai agent which he could manually turn off/n and could reply to customers own his own when he wants. In my current setup, once the number gets registered on whatsapp cloud Api, only the Agent could answer the incoming questions and the whatsapp stops working like it normally do. is there any way to add that in this ai agent or any other platform for such thing👀
newsletter automation
first my english is weak prompt problem and the api keys where i will find this is it paid ?
1 like • Sep 16
@Zakir Hussain for the weak English, you could explain the AI model in your Language and it will understand it. just tell it to give output in English if you want to although, it is better if you learn it for future as well. For the API just search any Model and add Api key in the end and follow through...
1 like • Sep 16
@Zakir Hussain initially the API keys are free for small usage. Consider it as a trial. After that you have to add atleast 10-15 Dollar to keep them running. Its burning depends on your Usage...
Help Regarding Whatsapp AI Agents
So, I got by first customer. Its a clothing business of traditional outwear waistcoat and coat. Its a local business. Now the business wants a whatsapp ai agent with free-text bot, which could also upload photos to show customers type and colour of waistcoats and i think ideally book the order. Now im having a problem on how to make it. Im using N8N for it. Anyone got any templates or suggestions 🥲?
4 likes • Sep 16
@Mofedul Alam Joy appreciate the advice, i have stored a column with the description of the product such as its style,colour etc and i have stored images in cloudinary. is it okay or should i do as you suggested ?👀
2 likes • Sep 16
@Igor Ganapolsky im using Whatsapp cloud API.. and te ordering is free-text👀
Deep Dive: Claude Opus 4.6 vs. GPT 5.3 Codex
Here are some of my key Takeaways after researching and trying out both Claude opus 4.6 and Gpt 5.3 codex.👇 1. Core Architectural Innovations Claude Opus 4.6:- 1M Token Context Window: This isn't just about "memory"; it’s about the ability to ingest an entire multi-repo architecture in one prompt, eliminating the need for complex RAG pipelines for many projects. Autonomous Agent Teams: Perhaps the biggest shift, Claude instances can now spawn sub-agents to handle different parts of a project (e.g: one for Backend, one for Frontend) and communicate internally to resolve merge conflicts. GPT 5.3 Codex :- Dynamic Steering: Unlike traditional LLMs where you wait for a full response, GPT 5.3 allows you to nudge the logic while it's streaming, drastically reducing the feedback loop. Granular Reasoning Control: You can now toggle the "Reasoning Effort" (Low to Max). This prevents the model from "over-thinking" a simple script, saving both time and compute costs. 2. Benchmarks & Logic Stress Tests Terminal Bench 2.0: GPT 5.3 Codex holds a dominant lead at 77.3% (vs. Claude’s 65.4%). This suggests GPT remains superior for system administration, complex CLI tools, and raw algorithmic logic. SWE-Bench: In terms of resolving GitHub style software engineering issues, the models are essentially neck-and-neck, proving that the gap in high level problem solving is narrowing. 3. Comparative Build Analysis Game Development (Fighting Game): Claude prioritized the User Experience, creating a visually impressive game with better character sprites. GPT focused on the "Engine," providing cleaner code structure but failing on the physics a bug caused players to fly off-screen, showing a slight disconnect between logic and visual intent. Application Design (Travel App & Portfolio): Claude followed a "Professional/Safe" design language great for corporate environments. GPT demonstrated "Cross-Project Consistency," mimicking the design language from previous sessions to maintain brand identity across different apps, a huge win for product continuity.
Deep Dive: Claude Opus 4.6 vs. GPT 5.3 Codex
1-10 of 20
Muhammad Ahmed
5
330 points to level up
@muhammad-ahmed-8802
AI Agents & Workflow Automations | Helping Creators & Businesses | My AI Automation Journey

Active 6h ago
Joined Sep 15, 2025
Powered by