Context
Blue Pixel’s own site, when it offered small, fixed-scope software projects. The intake was built for that model and archived before launch.
The problem
For a small business, a full requirements-gathering engagement can cost more than the software is worth. Either the client overpays up front, or the scope is guessed at and both sides lose.
Constraints
- Requirements had to come back as clean, structured data, not free text to interpret.
- Visitors range from technical to non-technical, so explanations had to adapt.
- The AI could not submit anything on a visitor’s behalf.
Approach
An AI-guided intake that collects requirements through a chat, but a structured one. Instead of a blank text box, it offers set choices: pick one, pick several, enter a number, confirm a summary, so answers come back as clean data. It gauges how technical the visitor is and adjusts how much it explains, and it cannot submit on its own: the AI proposes a summary, and the visitor’s click confirms it.
Key engineering decisions
Set choices, not free text
Each answer is a defined choice, number, or confirmation, so the output is data a scope can be built from rather than prose to interpret.
The visitor submits
The AI can propose a summary, but only the visitor’s own confirmation submits it.
Explanations that adapt
How much the intake explains depends on how technical the visitor is, so it neither lectures an engineer nor loses a business owner.
What it demonstrates
How an AI conversation can turn a loosely described need into structured requirements the visitor has reviewed and confirmed, adapting to who it is talking to. The confirmed summary gives both sides a clear, agreed starting point for the scope.