When we started building PotenAI, the obvious move seemed like a chatbot — user talks to an AI, AI figures out their career fit through conversation. We actually prototyped this direction early on.
We moved away from it. Here’s why.
Open-ended conversation is hard to keep focused, and it’s even harder to turn into a consistent, comparable output. Two users answering the same underlying questions in a free-form chat can produce wildly different signal quality — one gives you three paragraphs, another gives you “idk, I like people I guess.” That’s not a great foundation for something we want to be reliable enough for a real career decision.
So instead, PotenAI is built around a structured assessment: a defined set of questions designed to extract specific signal about strengths, work style, and motivations. The AI’s job isn’t to conduct a conversation — it’s to analyze the structured response data and generate a personalized roadmap: concrete next steps for education, career direction, or entrepreneurship. Structure in, intelligence in the analysis layer, personalization in the output.
The technical challenge that’s kept us busy: making the analysis feel genuinely personalized rather than templated. It’s easy to build a scoring system that buckets people into 8 categories. It’s harder to build one that produces output specific enough that two different users with similar-but-not-identical answers get meaningfully different roadmaps.
We’re early — two founders, bootstrapped, building in public. Would love thoughts from anyone who’s worked on structured data extraction, recommendation systems, or AI products that need to feel personal without becoming unpredictable.
Following along or curious about early access? → nopick.site
답글 남기기