AI-Powered Consumer Decision Platform — MVP Development

Hi Ndasi,

For the first release, do you want product recommendations to be generated from a normalised catalogue maintained within the platform, or should the system query merchant/product APIs live for each user request?

This affects data freshness, speed, API limits, resilience, caching and how easily future data sources can be added.

We are ready to sign the NDA and are awaiting your response.

Hi Ndasi,

I’ve reviewed the brief and the clarification points already raised, so I won’t repeat the questions around merchant APIs, product-data sources or recommendation logic.

One area I’d like to clarify is the founder dashboard: for the MVP, which metrics would be most important for you to act on — for example recommendation-to-merchant click-through, returning users, saved recommendations, AI/API cost per journey, affiliate conversions/revenue, or abandonment during the clarification process?

We’ve already submitted our proposal and included relevant experience across AI/NLP evaluation, conversational AI, procurement workflows and commercial quotation systems, with project demos available upon request.

Happy to sign the NDA and review the detailed specification if shortlisted.

Best,
VConn Pvt Ltd

Hi Ndasi,

Reviewing the requirements, the core technical challenge is balancing conversational flexibility with low-latency, deterministic product matching. Two quick questions regarding data pipeline and orchestration:

1. Data Ingestion vs. Live Querying: Are the approved product sources accessible via pre-synced, structured catalogs (allowing us to index and run sub-second hybrid search locally), or will the MVP rely on real-time API lookups per user query?

2. Clarification Boundaries: For user retention, do you envision a strict turn cap (e.g., maximum 2–3 clarification questions before forcing recommendations), or should the LLM autonomously decide when requirement confidence is high enough to output results?

Happy to sign the NDA to review the full logic and refine the milestone plan.

- Which product data providers, merchant APIs, or affiliate networks are approved for the initial MVP?
- Should recommendation ranking be entirely rule and data driven, or should it combine deterministic scoring with AI-assisted interpretation?

One key point I’d like to clarify: for the MVP, do you want the recommendation engine to be primarily deterministic (requirements → weighted scoring/rules → ranked products), with the LLM handling the conversation and requirement extraction, or do you expect the LLM itself to influence the final ranking?

I ask because I’d architect those two layers separately so recommendations remain explainable, reproducible and grounded entirely in your approved product data rather than AI-generated assumptions.

Hi Ndasi, how many product data sources are you planning to integrate at MVP stage, and do you already have access or agreements in place with them, or is that still being sourced?

-/ Which product categories will the MVP support initially, and do you already have access to the first merchant/product APIs or affiliate feeds?

-/ Should recommendations compare products across multiple merchants and account for live price/availability?

-/ How many clarification questions should the AI typically ask, and should these questions dynamically change based on previous answers?

-/ Will your proprietary recommendation logic be rule/score based initially, or should user behaviour and feedback also influence ranking?

Before submitting my milestone proposal, I have 3 quick questions:

1. Do you already have the approved product/merchant data sources or affiliate APIs selected, or should I also research and integrate the initial data providers?

2. For the MVP, approximately how many product/merchant integrations do you expect to be included in the first version?

3. Is the complete technical specification, recommendation logic and user flow available after signing the NDA, so I can provide an accurate milestone breakdown, timeline and fixed price?

Thanks.

- How do you envision the platform deciding which products are the best recommendations once the user's requirements have been extracted? Should recommendations primarily be based on explicit user requirements such as budget, brand, features, and use case, or should the system also consider implicit preferences, historical interactions, saved products, and previous recommendations?

- How much control would you like over the AI's clarification questions, for example, should the system have predefined question rules, or should the LLM dynamically decide what information is missing?

- Which affiliate networks or merchant tracking systems do you plan to use for the initial launch, and do you already have the required affiliate accounts and tracking parameters?