Design documentation is one of the first things cut when project budgets are tight - but it is also one of the most valuable artifacts for construction support, peer review, and future reference. I built an AI-agent tool that captures the complete design trail automatically as a byproduct of the design process itself, requiring zero additional time from the engineer.
My Role
Sole developer. Self-initiated project - identified the documentation gap and built the solution end-to-end.
What I Built
AI agent for automatic design documentation:
- The tool runs alongside the design process and captures every design change, consideration, assumption, decision, and rationale as the engineer works
- Produces a single comprehensive design document covering the full design trail - from initial assumptions through final decisions
- No hallucination - the tool records only what happened during the design session
- General-purpose - works across project types and disciplines, not tied to a specific software or workflow
Zero-overhead integration:
- The tool requires no additional design time or documentation time from the engineer
- It captures the design narrative as a natural byproduct of the work being done
- The output is a structured, readable document that can be included directly in the project deliverables
Impact
- Eliminates manual documentation effort entirely - design trail is captured automatically
- First deployed on a UV treatment facility detailed design project where the documentation budget was limited
- The project manager was impressed by the completeness and quality of the documentation package delivered alongside the drawings
- Applicable across all projects - particularly valuable on budget-constrained work where traditional documentation time is not available
Technical Approach
A Python agent that passes the design-session context to a large language model and writes each decision into the running design document. It records only what happened in the session and adds nothing of its own.