top of page

Designing an Agentic Workflow for Automated Documentation

Role:

AI Content Architect & Project Lead

Challenge:

Wix's knowledge base documentation process was manual and labor-intensive, described internally as being in the 'stone age.' With thousands of articles to maintain across rapidly evolving products, the team could not scale using traditional methods. The company needed a vision for what AI-assisted documentation could look like — and proof that it could actually work.

Contribution:

Designed the end-to-end architecture for a multi-agent system in Figma, mapping specialized roles across the documentation lifecycle: Intake, Task Management, Context-Mapping, Outline/Drafting, and Editing — including a novel 'Loopback Verification' feature where the editing agent re-runs documented steps in a browser to check accuracy. Then built 3 working proof-of-concept agents (ContextMap, OutlineDraft, and Editing) and deployed them for team testing. Hosted a training session to onboard the team, gathered structured feedback, and continuously fine-tuned each agent based on real usage. Now expanding these into production-grade Claude skills — including new specialized skills for testing product flows, writing procedural steps according to KB guidelines, and capturing screenshots along the way.

Outcome:

Moved from concept to working prototypes that the team actively uses and provides feedback on. Recognized internally as a 'massive high impact top priority project.' The architecture provides the blueprint while the POC agents prove the approach works in practice — a critical step toward full automation at scale.

Skills:

AI System Design; Agentic Workflow Architecture; Agent Prototyping; Process Automation; Figma; Claude Skills Development; Technical Leadership

bottom of page