Creation Assistant — AI‑Guided Page Creation
(02)
Overview
A guided human-centered AI workflow that turns natural language into a structured page draft (content, tags, metadata, visuals) while keeping users in control through review steps, manual edits, and strict role-based permissions.
Year
2025
For
innosabi
Project overview
As part of my Toolbox leadership, I led the Creation Assistant: an AI-powered stepper that helps admins, area managers, and content managers create Toolbox pages across innosabi’s Idea, Community, and Partner products.
The assistant uses natural-language input to draft structured content, recommend tags and metadata, and source visuals—packaged into a multi-step vertical scroll modal that reduces complexity without removing user control.
Key requirements:
• Generate structured page drafts from natural language prompts.
• Suggest titles, descriptions, and meta-tags to improve discoverability (SEO support).
• Provide AI-assisted image sourcing (e.g., Unsplash integration) plus manual upload/edit options.
• Preserve strict access/role management aligned with existing permission structures.
• Allow admins to choose preferred AI providers while maintaining privacy/compliance requirements.
Design elements:
• Modular UI components for a clean multi-step flow (prompt → draft → review → publish).
• Prototypes spanning low to high fidelity to validate interaction details early.
• Interaction patterns that build trust: clear AI states, editable outputs, and transparent choices.
• Configurable AI provider selection for admins, aligned with privacy/compliance needs.
Execution highlights:
• Researched interaction patterns for website builders and AI-assisted content creation.
• Used customer insights to define MVP scope (automation with human oversight).
• Led feasibility workshops to map integration points, dependencies, and constraints.
• Ran iterative user tests across multiple prototype directions (stepper vs wizard vs chatbot), landing on a vertical scroll modal as the most intuitive pattern.
• Developed prompt engineering capability and partnered with internal AI specialists to generate useful titles and meta-tags (behaving like an SEO assistant, not a generic chatbot).
What I learned:
• Trust is the feature: AI only helps when users can review, edit, and override at every step.
• Guardrails aren’t optional: permissions and visibility rules must remain intact in AI workflows.
• The interface is the strategy: testing different interaction models early prevented shipping the wrong mental model.
My role:
I led the Creation Assistant end-to-end across product and design: setting the vision, driving research, defining the MVP, building and testing prototypes, and aligning delivery with engineering, customer success, sales, and customer stakeholders. I also invested early in prompt engineering and partnered with AI specialists to ensure the assistant produced high-quality, editable outputs that fit platform governance and user workflows.