AI product · MVP
AI Project Management Assistant
From meetings to project documentation, planning and tracking
An application at 57Blocks that uses AI workflows to automate software-project documentation, planning and tracking from meetings. I led its technical planning and initial development, up to an early MVP.
Sources
- Meetings
- Documents
AI workflows
Dify- Work with documents
- Generate content
- Query project information
- Documentation
- Planning
- Tracking
Initial architecture: hexagonal principles
Context
57Blocks, where I worked on projects for international clients, set out to build an application envisioned as a commercial product. I was given its technical planning and initial development.
Problem
Project meetings produce decisions, tasks and knowledge that someone then has to write up, plan and track by hand. The idea was to let AI workflows produce and maintain that documentation, planning and tracking from the meetings themselves.
What I owned
- Technical planning and initial development, from scratch.
- The initial frontend and backend architecture, on hexagonal principles.
- Designing and integrating the Dify workflows to work with documents, generate content and query project information.
- Taking the application to a first MVP version.
- Boundary: defining the backend architecture was part of the job; running a production backend was not. My role stayed frontend and product-oriented.
Decisions
- Hexagonal principles from day one, on both sides, so the application's logic stayed independent of the interface and of external pieces such as the AI workflows.
- Dify to build and test the AI capabilities, so the workflows could be tried and adjusted without rewriting application code each time.
Result
An early MVP of the application, with AI workflows for documents, content generation and project queries, built on the architecture I defined. It was a first version, not a finished product.