Case Study

Pearl: Privacy-First AI Note-Taking for Professionals

Overview

Pearl is a privacy-first AI note-taking app built for professionals handling sensitive information—lawyers, therapists, executives. All transcription and summarization runs locally on-device, eliminating the cloud-privacy risks that make competitors unusable for confidential work. Over a 3-month design sprint in 2024, we addressed a critical gap in the market: most AI note-taking tools are dealbreakers for users bound by data protection regulations.

The Problem

Every major AI note-taking tool uploads recordings to the cloud for processing—a non-starter for professionals under strict data governance. Through user interviews, we identified three interconnected pain points:

Privacy Barrier

Lawyers, therapists, and executives simply cannot upload confidential meeting data outside their organization.

Processing Friction

Existing tools require minutes or hours for transcripts to appear—too slow for real-time note review or meeting follow-up.

Useless Transcripts

Dense walls of text with no structure mean users spend hours hunting for key takeaways and action items.

“This is a quote from someone that i think is very special!”
Person
Role and Title

User research findings: Core frustrations and desires for better meeting notes

Research & Strategy

We started by understanding the real context of professional note-taking. How do people actually capture information during meetings? When do they write, when do they listen, and how much cognitive load can they handle while staying present?

User Interviews & Contextual Inquiry

Observed lawyers, therapists, and executives during meetings to understand note-taking habits, pain points, and regulatory constraints.

Competitive Analysis

Analyzed Granola AI, Otter AI, Fireflies, and Notion AI. Found all relied on cloud-based LLMs—a dealbreaker for privacy-sensitive users.

MoSCoW Prioritization

Focused on must-haves (on-device recording, transcription, summarization), should-haves (downloadable models, speaker tagging), and explicit exclusions (no cloud, no data sharing, no ads).

Design Solution

Pearl's core innovation is simplicity: everything runs locally. No cloud upload. No data export. Users download and manage AI models on their device, then record, transcribe, and summarize—all offline.

On-Device Model Selection

Choose lightweight models for speed (quick summaries) or larger models for deeper reasoning, all stored and processed locally.

Background Recording & Dynamic Island Integration

Record continuously even when switching apps or locking the screen. Control recording (pause/stop/tag) directly from Dynamic Island without interrupting the meeting.

Actionable Transcripts

Tag meeting participants inline, convert discussion points into tasks, and extract high-level summaries instead of dense word-for-word records.

Core user flows: note-taking, Q&A with Pearl, discussion transcription, and transcript review

Feature highlights: note library, recording interface, processing state, and model selection

Onboarding flow: login, verification, and model selection with clear performance/storage tradeoffs

Real-time summarization in action: Pearl processes notes while recording and displays instant summaries

Competitive Landscape

Pearl stands apart in an increasingly crowded market. While Granola AI, Otter AI, Fireflies, and Notion AI each have strengths, none address the privacy-first use case. Here's how we differentiate:

Competitive analysis: Pearl vs. Granola AI, Otter AI, Fireflies, and Notion AI across key dimensions

Impact & Results

of early testers trusted Pearl over competitors due to on-device processing

reported feeling more present in meetings when using Pearl

0months

from concept to functional prototype

Key Takeaways

Privacy as a Feature

For a segment of users, privacy isn't a nice-to-have—it's a requirement. Designing for their constraints from day one unlocks an underserved market.

Simplicity Through Constraints

By saying no to cloud processing, we didn't limit Pearl—we clarified it. Constraints forced elegant, focused design decisions.

Context Drives Features

Understanding how professionals actually work in meetings—multitasking, listening, intermittently writing—shaped every interaction, from background recording to Dynamic Island controls.

Next Project

Vitalacy Virtual Care Dashboard

View project