AI Revisited - Part 2

Jason Fried explains how AI has become his personal editor, using Claude and ChatGPT to scan Basecamp's customer testimonials and rewrite copy in the exact language users employ. By feeding the models real customer phrasing, he ensures headlines and feature descriptions resonate, while still preserving his own tone. This AI‑assisted writing cuts revision cycles and guarantees that product messaging matches the mental models of the audience, a critical advantage for any business seeking clear, customer‑centric communication.

Beyond copy, Fried leverages AI for rapid prototyping and data generation. He asked Claude to build a temporary sidebar UI that retains chat avatars, a feature he could test in minutes rather than waiting for a developer. He also prompted the model to create three weeks of multi‑user chat history, complete with attachments and emojis, instantly populating a local database for design mock‑ups. These AI‑generated test datasets eliminate the tedious manual setup of realistic scenarios, accelerating design validation and reducing reliance on teammates.

Strategically, the conversation reveals a shift from early AI skepticism to enthusiastic adoption within 37signals. While past experiments like automated bug duplicate detection and weekly summaries fell short of user interest, the team now envisions a hybrid future for Basecamp 5: native AI tools complemented by user‑brought agents that act as co‑workers. Plans include command‑line interfaces to simplify agent integration and careful timing to avoid building features that could be superseded by emerging platforms. This balanced approach underscores how businesses can responsibly embed AI, leveraging immediate productivity gains while positioning for long‑term innovation.