Angela Yang

 

TL;DR

AtlasNova — UI for AI-assisted document generation

Platform

AtlasNova AI

Enterprise macOS suite

Multiple AI product solutions—this case study covers one feature: document generation.

Feature

Doc generation

AI-assisted PRDs & research docs

Knowledge base, templates, step-by-step section generation, persistent chat.

My role

Intern

UX/UI for this feature

Mockups and end-to-end UI flows—not the full platform or other solutions.

Context

One feature inside a larger platform

AtlasNova AI is an enterprise macOS platform with multiple AI product solutions. This case study is scoped to one of them: AI-assisted document generation for product teams writing PRDs and research docs—connect a knowledge base, configure a template, and generate each section with persistent chat beside the draft.

As a UX/UI design intern, I owned the UI for this feature—mockups and end-to-end flows from whiteboard sketches through hi-fi screens.

Principles

What we designed for

PMs using AtlasNova sit between two failure modes: AI that feels like a black box, and workflows that split the document from the conversation. Each principle below maps directly to a design decision.

  • AI-assisted, not AI-replaced — Generation is step-by-step, but the PM stays in the loop. Chat stays open while the doc updates, so steering a section feels like commenting on the artifact—not switching apps mid-draft.

  • Transparent generation — Each section surfaces where the model is in its pipeline (indexed → reading → generating), so users can tolerate wait time and catch bad source pulls before the whole PRD is wrong.

  • Split-pane layout — Document left, chat right, one macOS window. Users never lose context toggling between generated copy and directing the AI.

  • Context you can see (@mentions) — @mention pills reference knowledge-base artifacts—competitive analyses, prior specs, research summaries—so grounding the model is a visible, repeatable interaction, not hidden prompt engineering.

  • Template settings upfront — Format, page count, detail level, and output type belong in onboarding, not an advanced menu after a bad first draft. Surfacing controls before generation cuts “regenerate everything” loops.

  • Connected platforms — Document generation pulls from scattered team knowledge. Connect flows for Gmail, Google Drive, and other platforms let users import context without leaving AtlasNova—familiar OAuth-style sign-in keeps the handoff trustworthy.

Sketches

Whiteboard from team discussions

Early IA and layout explorations for this feature—before hi-fi mockups.

Whiteboard sketch — IA overview with split-pane doc view and Home → workspace → doc hierarchy
IA — Home, workspaces, PRD / Research doc types.
Whiteboard sketch — workspace navigation, tagging, and AI affordances
Nav, tagging, and AI touchpoints.
Whiteboard sketch — home dashboard with Welcome and Suggested Actions
Home — Welcome, suggested actions, project grid.
Whiteboard sketch — document canvas with project-scoped sidebar feed
Split-pane — document left, project feed right.
Mockups

Hi-fi screens

Hi-fi screens I delivered for the document generation flow—home → import sources → split-pane workspace → platform connect.

AtlasNova home dashboard — recents, expert templates, and workspace sidebar
Home — recents, templates, workspace nav.
AtlasNova file import — drag-and-drop upload with Google Drive and OneDrive
Import — local files, Google Drive, OneDrive.
AtlasNova split-pane workspace — PRD document with chat, @mentions, and project settings
Workspace — document left, chat and @mentions right.
AtlasNova Gmail connect — Google OAuth sign-in modal over split-pane workspace with open document tabs
Gmail connect — choose an account to continue to AtlasNova.
UI flow

New project onboarding

Onboarding UI for new document projects: dashboard → prompt → AI clarifying questions → template pick → file upload. The flow narrows scope before generation—and surfaces template settings upfront, not in a buried menu.

AtlasNova UI flow — new project from dashboard through prompt, AI Q&A, template selection, and file upload
← Back to home