Toolbar workflow
Key features
Contextual AI actions: We integrated AI assistance as an always-there, context-aware editor and critic. Specifically, AI pre-built tools were built for common writing actions, and were prioritized based on user interviews and observation of writing processes.
- Highlight text to access relevant AI-powered tools
- Apply AI actions to specific sections of text
- Lets writers focus on core ideas while AI does the heavy lifting
- Continue writing: This button provided free-flow expansion—if you needed to go into more detail on a particular part of writing.
- WYSIWYG editor: A clean, minimal writing interface.
Design considerations
- Workflow-driven UI: Designed the interface based on real-world writing patterns from users and my own writing workflow.
- Incorporated insights from user interviews and direct observation
- Mapped common pain points in existing writing tools
- Prioritized features that addressed frequent context-switching issues
- Minimal AI friction: Focused on making AI assistance feel like a natural part of the writing process.
- Placed AI tools within existing UI patterns (e.g., right-click menus, formatting bars)
- Used progressive disclosure to avoid overwhelming users with options
- Made sure that AI suggestions maintained the writer's voice and style
- Platform stickiness: Aimed for the bare MVP writing experience due to resourcing, but prioritized a deeper, magical AI experience.
- Analyzed features of popular word processors to match core functionality
- Added specific AI capabilities focused on writers
- Designed for easy publishing and collaboration to increase user commitment
- Contextual AI understanding: One challenge was editing and generating isolated blocks of text without broader context.
- Structured and designed prompts to provide sufficient context for the LLM pipeline
- Implemented a system to capture implicit acceptance/rejection data from users
- Used this feedback to continuously improve AI suggestions and maintain coherence with the overall piece
Takeaways
The editing interface was critical to the overall user experience to keep users in-app and avoid going off platform to another word processor. Providing an AI assistive toolbar that felt like a natural extension of typical editing toolbars helped bridge the usability gap and help writers feel comfortable writing their work on the platform.
Providing actions to apply to specific ‘layers’ of text—a la Photoshop—helps to edit faster and with more granular control, ideally helping writers to write and edit at the speed of thought, and stay in a flow state.
Layering the assistant on a familiar interface (the WYSIWYG editor) was also important in minding the gap, since we could tap on mental models of how a word processor is supposed to function. Anything too unfamiliar would’ve created more frustration.
Addressing the challenge of contextual understanding was crucial for producing coherent AI-assisted edits. By designing prompts that provided broader context and leveraging user feedback data, we were able to significantly improve the relevance and quality of AI suggestions, so they were in sync with the overall piece rather than just isolated text blocks.
Future directions
- Expand the range of AI-powered writing blocks based on user feedback and usage patterns
- Real-time personalized toolbar suggestions based on individual writing styles
- Collaborative features allowing multiple users to work on a document with shared AI assistance