Percepta AI

Inspiration

More people now use AI tools to learn about products before talking to sales. These tools summarize products, compare options, and help buyers make decisions. We noticed that many products are misunderstood by AI. This is usually not because the product is bad, but because its positioning, limits, and key differences are unclear or spread across many places online. Product managers try to check this today by manually testing prompts and reading AI answers. This process is slow and based on guesswork. We wanted to build something that shows clearly how a product is understood and helps teams fix problems before they affect users or revenue.

What it does

Percepta AI shows product teams how AI and potential buyers currently understand their product using only public information. It helps teams see what AI gets right, what it gets wrong, and what important details are missing. The product then gives clear suggestions on how to improve product explanations and positioning. Teams can repeat this process over time and see if perception improves.

How we built it

Users start by entering basic information about their product such as their website, competitors, and target users.
Percepta AI collects and analyzes public content and shows users exactly what sources are used and where information is unclear.
The system creates a set of important questions that reflect how buyers compare and evaluate products. Users can edit these questions or add new ones.
We used Keywords AI to manage prompt templates and evaluation logic so results stay consistent across different AI models.
Each AI model answers the questions as if it were a buyer who only knows what is publicly available. Users can see trace logs that show how answers are generated.
Responses are scored using clear criteria such as correctness, clarity, and differentiation. Users can see why each score was given.

Challenges we ran into

Connecting our TRAE code to Keywords AI
Understanding how Prompt Management works

Accomplishments that we're proud of

We turned a manual and unclear product management task into a structured and repeatable process.
We built full transparency so users can see sources, reasoning, and scores.
We created a system that compares how different AI models understand the same product.
We focused on giving clear actions, not just analysis.

What we learned

Small gaps in public information can lead to big misunderstandings.
Clear structure and good prompts matter as much as the AI model itself.
Seeing how AI reasons builds trust and confidence in the results.
Product positioning problems are easier to fix when they are visible.

What's next for Percepta AI

We want to add stronger competitor comparisons and category level analysis.
We plan to track perception changes automatically as public information updates.
We want to support teams working together across product, marketing, and sales.
We also plan to connect product documentation and release notes to improve accuracy over time.