I often brainstorm new ways to monetize my skills outside of my main job. But starting with market research, then planning, then building a prototype… it always feels like a heavy lift.
So recently, I decided to use AI as a sounding board. Honestly, I didn't expect much. It started late one night when I, in a burst of sleep-deprived inspiration, threw out a very vague question: "What's a profitable blue ocean idea?"
The AI's initial answers were predictably generic: "blogging," "affiliate marketing," "selling skills." That wasn't going anywhere, so I gave it a bit more context about my situation:
Adding these conditions made the AI's suggestions more specific. A direction for "one-off digital content" started to emerge. After a few more exchanges, the AI finally proposed a "cooperative mystery game where the organizer just plays a video" as the core idea.
The concept: participants team up to solve a mystery at events like wedding parties or company get-togethers. The selling point is ease of use for the organizer, who just projects a dedicated video. Hmm, that's a one-time purchase, no inventory. Not bad.
"Sounds interesting," I replied. The AI then asked, "Shall we proceed with this plan?" It was late, so I responded almost on autopilot: "Let's do it. I'm going to bed, so do what you can." Then I closed my laptop.
The next morning, I checked the AI's notifications and was genuinely surprised. Overnight, it had built a complete, truly playable Minimum Viable Product.
The generated files included:
Playing the video, it clearly narrated the incident in Japanese, displayed a consultation timer during the puzzle-solving parts, and even included an answer reveal section at the end. Simple background music and captions were present, making it almost usable as-is for an event.
What surprised me even more was that it didn't stop there.
These were also provided. From ideation to prototype development, and even a rough sales strategy – the AI had sprinted through it all in one night. I hadn't expected such concrete output just by vaguely saying "do what you can" before bed.
Of course, what the AI produced isn't perfect. The video is just a demo, and it needs quality improvements to be sold as a product. There are many areas I'd want to refine: BGM selection, caption design, narration nuances, etc.
Also, this is just the first scenario. To develop it into a product, I'd need to produce multiple scenarios. While AI could assist with scenario creation, checking for fun factor and logical consistency ultimately falls to a human.
Setting up the sales page, integrating payment systems, and actual marketing. Even with AI-generated drafts, executing these falls to me. Even if AI handles 80% of the tedious work, the final 20% of refinement and getting it out into the world still requires human touch.
What I learned from this experience is that brainstorming with AI can start with just "throwing out a rough idea." From there, through dialogue, communicating your skills and constraints will significantly improve the accuracy of its suggestions.
If you're looking for new ideas, try asking AI this:
**"Can you create the simplest possible prototype (MVP) for (your idea)?"**
Instead of just idea documents or text, try getting it to create "something dynamic" like a video or a tool, as it did for me. This transforms an idea you've only considered in your head into a concrete form, making your next steps much clearer.
I build and run small Python systems — trading bots, RAG APIs, scheduled automation — and write up whatever breaks along the way.
If a provider-agnostic RAG Q&A API is useful to you, mine is MIT-licensed on GitHub: rag-faq-api. It runs and passes its full test suite with no API key (offline stub LLM + hashing embedder), swaps to Claude / Gemini / OpenAI via one env var, and ships a retrieval-quality harness (Hit@k / MRR / Recall@k) with a chunking sweep.