Yapz.app - AI-driven interviews that capture nuanced insights : datarockets

Key Features

AI-first UX
Personalized Conversational AI
Automated AI Analysis and Insights

Project Scope

R&D, AI integration, Web Development

Yapz is our first enterprise AI product, debuting on Product Hunt and co-created with collabs. It reimagines how organizations gather and interpret feedback by moving beyond rigid multiple-choice polls and into open-ended, AI-driven interviews that capture richer, more nuanced insights

Challenge

Traditional polling solutions like Typeform have dominated for years, but they force users into predefined answer categories, restricting the depth of potential feedback. Yapz.app’s goal was to solve this limitation by providing:

Real-time, adaptive AI-driven conversations that ask relevant follow-up questions.
Automatic analysis of large volumes of unstructured conversational data.
An intuitive interactive dashboard that delivers both individual transcripts and high-level insights in one place.

Approach

Proof of concept in 4 weeks

We aimed to release a working prototype of Yapz.app within weeks. By integrating Supabase and Vercel for backend and hosting, we kept our infrastructure lean. Vapi handled voice communication with minimal fuss, and Langsmith allowed us to monitor and debug AI calls, ensuring prompt accuracy early on. This streamlined setup enabled us to validate our core concept within the first month.

Parallelizing Prompt Engineering

While core development proceeded, we focused on refining AI prompts in parallel. Debugging each prompt required running full conversation sequences across different surveys. To manage this complexity, we built a specialized internal tool to test and iterate on summary prompts. We conducted manual checks to confirm results, ensuring we captured genuinely useful insights.

Mastering Evolving Summaries

The final hurdle was creating summaries that update in real time for over 200 conversations, each running 7–10 minutes. We wanted precise results aligned with creator inquiries yet flexible enough to capture organic insights. After exploring multiple methods, we settled on a refinement strategy, striking the right balance between detailed accuracy and broad, evolving insights.

Results

Yapz has officially launched in January 2025, quickly attracting thousands of users who have already created hundreds of their own projects. Currently free to use, Yapz is in an exciting phase of exploration—our team is actively evaluating potential pivots and the next steps for this innovative AI-driven feedback platform.

Tech stack

Supabase
Next.js
React
TypeScript
Tailwind CSS

Vapi
ElevenLabs
LangSmith
LangChain
OpenAI

Launching Yapz.app AI agent on Product Hunt

Key Features

Project Scope

Yapz is our first enterprise AI product, debuting on Product Hunt and co-created with collabs. It reimagines how organizations gather and interpret feedback by moving beyond rigid multiple-choice polls and into open-ended, AI-driven interviews that capture richer, more nuanced insights

Challenge