Hooh. An AI assistant translating complex bureaucracy into human language
We shipped Hooh’s MVP on web and iOS: upload, layered reading, assistant chat, and file organization. I designed the core flow and MVP system; Progressive Disclosure became the shared pattern that let people choose how deeply to read without losing context.
- Launch
- Web + iOS shipped
- Validation
- Core flow completed unaided
- Core pattern
- Progressive Disclosure
Problem
People face document anxiety when dealing with contracts, reports, insurance papers, or legal texts. The problem is not just volume, but uncertainty: users don't know where to start, what matters most, or whether they are missing hidden risks, deadlines, or critical clauses. They need a lower-stress way to understand the essence of a document before deciding whether to go deeper.
Task
Design an MVP for web and iOS that helps users understand complex documents faster, with less stress and less manual reading.
Users
The product is designed for people without a legal background who regularly deal with complex paperwork — expats, digital nomads, and everyday users. They use Hooh to quickly decode lease agreements, visa requirements, insurance papers, and banking contracts, and understand what matters without reading every page line by line.
What we launched
We designed and launched the MVP across web and iOS, defining the core experience for document upload, layered reading, conversational exploration, and file organization.
Design process
Started with competitor analysis — Genei, Notion AI, Microsoft Copilot for docs all gave full summary upfront, which removed user agency: you couldn't choose how deep to go. Early prototypes tried single-step summary, but internal testing showed users either felt over-summarised or wanted to dig further. This pushed us to layered reading.
Key UX decision: Progressive Disclosure
Single-step summary risks oversimplification; full document keeps overwhelm. Three layers let users self-calibrate depth: skim with snapshot, dive deeper if signal is meaningful, ask targeted questions if specific concern surfaces. Each layer is a commitment threshold — users don't pay attention cost until they've decided the document is worth it.
1. Snapshot
A one-glance verdict — skim the gist before committing any attention
2. Summary
A structured breakdown — dive deeper when the signal looks meaningful
3. Q&A
A contextual chat — ask targeted questions when a specific concern surfaces
Designed a conversational experience for document exploration
The assistant uses uploaded documents as context, allowing users to ask follow-up questions, explore categories, and retrieve relevant information without manually scanning long text.
Trust by design: the assistant can be wrong
An assistant that reads your contracts and lab results only earns trust if it can show its work. Three guardrails run through the whole experience: every claim links back to the exact source document, users see and control which documents the assistant reasons over, and the interface says plainly that Hooh may be wrong — keeping the original file one tap away.
Sources on every claim
Citation chips in answers open the exact document behind each number
Scoped context
The All-documents selector shows — and limits — what the assistant reads
Honest uncertainty
A permanent "Hooh may be wrong" notice points users back to the original
Outcome
MVP shipped on web and iOS. Public-release analytics weren't part of this engagement — the evidence below is what we validated before launch.
Shipped on two platforms: the MVP went live on web and iOS — upload, layered reading, assistant chat and file organization.
A clean pass on the core flow: in every internal test session, participants took a real document from upload to understanding and completed it without help.
Patterns that stuck: Progressive Disclosure became the core pattern for every document feature — and quick snapshot the most-used way into a file.