From demo to beta
The app is currently only available in demo mode. Not for lead-generating purposes, but simply because it lacks a few features to make it ready for public:
- Account/session management
- Data privacy & encryption features
These are both essential features for going public, not just for GDPR-compliance, but also because personas - no matter their quality - are internal strategic documents. Further features include:
- Implementing a feedback loop for qualitative and error feedback
- A low-barrier paywall (on a tip-level) as every evaluation includes circa 120 requests to the OpenAI API, as well as session-based cloud hosting costs
Bugfixing & prompt engineering
- In ca 10% of cases, the persona ensemble JSON generated by OpenAI vision API doesnt follow the JSON schema I need for follow-up tasks. I need to identify different prompt approaches that further reduce hallucinations
- If the persona document is very information-dense it (for a lack of better words) clogs the app. Solving this would open "peek" for usage beyond just ad hoc-personas.
- Even though all prompts are written in German, some responses are in English. More strict prompts and using the DeepL API for language detection of responses could solve that.
Product expansion
- Switch from the Streamlit framework to a standalone app environment
- Explanations of identified segments for deeper understanding
- Including the mental strategic model GenAI identified that is the foundation for the personas created incl the visualization
- Generating questions for the user to answer in a next step before adapting the personas company-wide
- Including reference information for diversity traits to represent minority interests
- Extension to include international social and segmentation models, as the app currently only contains German models
- Extension to include psychographic segmentation models.
Beyond the R&D Fellowship, I'm looking for funding & sponsorships to implement these features.