[Invention Deck Series] EP 0: AI Service Development Without a Developer
Category
🚗 The Invention Lab Story
Date of issue
Sep 4, 2025
🌫️ It starts with very light words
As we witnessed AI's growing presence across industries, we realized we could no longer just sit back and watch. We began to question, "Couldn't there be problems within our company that could be solved with AI?" So, we began experimenting with automating repetitive tasks within the company, leveraging n8n's AI Agent.
As I explored the possibilities by integrating various tools, I realized that much more could be automated than I had anticipated. Naturally, this led me to expand my thinking to other areas, and I delved deeper into the subject.
But then someone next to me said:
👤: "Can't you put that on the invention deck too?"
Our team just laughed.
👥: "How can we make that without developers..."
But strangely enough, everyone started to think, "Can't we really do this?" Vibecoding is all the rage these days, and we've been experimenting with various tools. If we could just connect the data and ideas we have, along with the tools we're working with, wouldn't we be able to create a truly "working service"?
The idea quickly led to an experiment, and that's how the project began.
🪜 What is an invention deck? And the limitations we faced.
InventionDeck is a card-based ideation tool originally designed to help startups and new business managers at large corporations quickly generate ideas. It presents information on various industries, targets, revenue models, trends, and technologies in card format, allowing users to combine these cards to create new business ideas.
The initial version of Invention Deck was developed through outsourcing in 2020. The idea itself was quite promising, and many people actually used the tool in workshops and training sessions to generate various business ideas and achieve results. However, as time went on, several distinct operational limitations began to emerge.
First, creating and entering new card content had to be done entirely manually. Even if new trends emerged, they couldn't be reflected unless someone manually organized them. Second, even minor feature modifications or updates required constant reliance on developers. This made small-scale experiments and rapid feedback impossible, ultimately diminishing the service's sustainability.
We were convinced that reviving the Invention Deck required determination.
🤖 And then switch to AI
In April 2025, we made a decision.
"Let's completely transform the Invention Deck into an AI-based system."
Let me spoil it for you... We actually succeeded in developing it! We even tested it with users.
As a planner-centric team, our goals were clear.
"Let's create a working service first!"
To do that, we marshaled our resources. Fortunately, we had several weapons at our disposal.
•
New business ideation know-how optimized for startups, in-house ventures, and new business planning.
•
Card content accumulated over the years
•
Experience observing user interactions
•
AI tools available to everyone these days...
We didn't design the service architecture or code the front-end ourselves. Instead, we assembled the necessary features piece by piece using tools accessible even to non-developers. Of course, the process wasn't simple.
•
GPT was good at creating text, but it was difficult to properly interpret the context and control the output.
•
Supabase seemed easy at first, but I ran into some major problems with things like fields within tables and RLS settings.
•
N8n was flexible, but one small exception that humans didn't think of stopped the whole flow.
•
Various Vibe Coding services can create a web service that looks good with just one sentence, but when it comes to actually implementing it, the development process is not easy due to frequent errors and unexpected limitations.
Through all this, we've learned one important thing.
"More than choosing a good tool, "It is more difficult and important to understand and use the tools well."
🎊 So here we go with this series!
We're still running the Invention Deck AI. Of course, it's not perfect. But
•
We provide users with the features we envisioned
•
We are updating card content in real time.
•
We are continuously improving our services through planners alone.
What we wanted to create was not a “perfect system”,
It was a "service that actually works" that even planners could create using AI.
We hope that the results will be useful to startups and planners.
In future series, I plan to include problems I have actually encountered and solved.
I'd like to explain why we decided to rebuild our existing service from scratch, and how we're doing it.
I hope this series will be of some help to those struggling with similar issues.
👉 In EP.1, we will honestly share the trial and error we experienced while using Lovable, Google App Script, Replit, and Firebase Studio, as well as the things we ultimately realized.
Subscribe to 'The Invention Lab Blog'
If you subscribe to the site, you can be the first to receive the latest updates, including new posts, via notifications and email. Join Slashpage and subscribe to 'The Invention Lab Blog'!
