How I Built a Multi-Branch AI Automation System in Make Using Routers, JSON Parsing & Aggregation

Knowledge Hub
, , , , ,
1

AI Email Classification & Routing System in Make.com

make 2
make 2997×434 85.4 KB

Hello everyone! :face_blowing_a_kiss:
I recently designed a multi-branch AI automation system in Make that handles dynamic user requests through a Telegram bot trigger and routes them intelligently using layered routers. Instead of building a linear workflow, I structured the architecture to classify incoming messages using chained OpenAI modules, convert responses into structured JSON, and dynamically route each request based on intent. Each branch processes a different logic path, allowing the system to handle multiple scenarios simultaneously without breaking flow. This approach keeps the automation scalable, clean, and easy to maintain as complexity grows.

To ensure reliability, I implemented JSON parsing, iterators, array aggregation, and structured logging into Google Sheets before sending contextual responses back through Telegram. The key focus was maintaining predictable data flow between AI outputs and Make modules, avoiding messy mapping errors, and designing the routers in a way that prevents scenario chaos. This kind of structured AI architecture allows businesses to build intelligent assistants, automated support systems, and scalable workflow engines without constant manual intervention. Curious how others are structuring multi-branch AI systems in Make? DM is open :hugs:

#Makecom #AutomationArchitecture #AIAutomation #WorkflowDesign #NoCode #SystemDesign #OpenAI #BusinessAutomation

5 Likes
3

This looks like a very smart way to structure the AI inside of Make. I really love the use of Routers to make the workflow seem more dynamic than a simple AI prompt. Something that can really process information and make decisions, all while returning clean information back to the sender.

We’ve been building similar AI-driven workflows at our company with a strong focus on no code platforms. Keeping the data flow predictable between AI outputs and Make modules is key to avoiding chaos in a scenario. Love seeing a design like this in the community.

2 Likes
4

You don’t need parse json after a chatgpt response. There is an option to automatically do that inside the create a competition module.

1 Like
5

Thank you, Anthony. I’d love to connect with you.

6

Thank you for the observation, Stoyan.