Project Context
Editing support inside the document
Writing Assistant brings AI editing support directly into the document, so users can work on selected text without switching to a separate chat or losing the surrounding context.
CASE STUDY
Writing Assistant
A selection-based document editing workflow where users highlight text and apply AI actions without leaving the editor.
The UX Challenge
Keep AI editing clear and connected
- Make AI editing feel connected to the document
- Make it clear exactly what text is being edited
- Support multiple actions without overloading the interface
- Keep original content and AI output visually distinct
My Role
I designed how in-document editing and AI assistance work together: select content, trigger an action, review the result, and keep writing in the same place.
Core Workflow
From selection to reviewed output
The interaction rule is simple: selection communicates intent. Users highlight text, choose an AI action, review the generated result, and continue editing in the same workspace.
- Add or open a document
- Highlight the text to edit
- Choose an AI action
- Review the generated output
- Continue editing in the same workspace
AI Action States
Three compact states show how the same selected text can be transformed in different ways.
Translate
Converts selected text into another language while keeping the original document visible.
Summarize
Condenses longer text into a shorter version for faster review.
Bulletize
Transforms dense text into structured points for easier scanning.
Process
I designed the experience as a repeatable loop: select, act, review, continue. That keeps the interaction easy to understand while leaving room for additional AI actions over time.
Interaction Strategy
The assistant works as a contextual editing layer rather than a separate destination. It supports the document by making common AI transformations available at the moment they are needed.
Selection-Based Input
- Highlighted text becomes the active input
- No copy-paste is required
- The user always knows what content is being edited
Action-Based Editing
- Common AI transformations are explicit actions
- Users do not need to write a new prompt for every task
- The action set remains easy to scan and extend
Review Before Applying
- Generated output appears separately from the source
- Original content stays visible
- The user remains in control of the final text
Final Solution
AI editing that stays inside the document
Editing and AI assistance stay inside the document. Selected text is the input, actions stay explicit, and generated content is reviewed before anything is applied.
That keeps the loop short: select → choose action → review → continue editing, instead of sending people into a separate chat.
Outcome
Assistance that stays inside the draft
Users can revise, compare, and keep writing without leaving the document.
New editing actions can be added later without changing how the user works: select text, choose an action, review the result.
See something you’d challenge?
I like talking through design decisions. If you would have approached something differently, I’d like to hear it.