Loft AI: Defining What to Build and Why

I joined as the first PM hire. The product leader had already shaped a vision: an AI-powered bookmarking tool that didn't just save content, it helped you get it back when it mattered. Web and mobile.

Back in April 2025, the major players (Pocket, Raindrop) were still built on keyword search. Natural language search was just starting to emerge in consumer products. None of them had it. Save an article about “managing remote teams,” search later for “how do I run standups better?” and actually find it.

With that in hand, I wanted to get to the core question first: what problem are we actually solving, and who are we building for?

I ran interviews anchored in recent behavior: “Tell me about the last time you saved something.”

The PRD had already identified retrieval as the core problem. Research confirmed it, but also surfaced what made retrieval fail upstream: content scattered across a dozen apps, and the cognitive burden of organizing and remembering where things lived.

Three pain points surfaced consistently:

• Scattered saves: content spread across Instagram, Twitter, LinkedIn, tabs, notes apps. No single home.
• No native organization: the platform puts the full burden of creating folders and remembering which bookmark went into which folder on the user.
• Retrieval failure: after 2-3 months, people couldn't find what they'd saved.

On segmentation, demographics didn't predict behavior. It was motivation: why do people save? Three segments emerged:

• Productivity Savers: save to use later, retrieve frequently
• Lifestyle Curators: save for inspiration, retrieve when needed
• FOMO Savers: save compulsively, almost never retrieve

Productivity Savers and Lifestyle Curators became the focus for early build decisions.

The initial vision specified cross-platform: web, browser extension, and mobile app. But research revealed most people discovered content on mobile. Most save moments happened on mobile, when people were on the go, scrolling through feeds, saving things forwarded by team and friends. That changed where to invest first.

I weighed the call against value, usability, and feasibility:

• Value: Every interview surfaced it. People already saved on mobile, in the moment.
• Usability: Saving from mobile mapped to a familiar pattern (sharing between apps).
• Feasibility: Engineering confirmed the architecture was buildable. Vector database and RAG on the backend, mobile as the client surface.

Mobile-first was the call.

The PRD laid out 5 functionalities (bookmarking, contextual resurfacing, AI summarization, AI chat, daily discovery cards) across 3 platforms (web, browser extension, mobile app).

MVP narrowed to 3 capabilities, mobile only:

• Bookmarking: the save action across apps
• AI categorization and tagging: new from research, addressing the upstream organization burden
• AI chat with summarization: natural language retrieval, the validated differentiator

The product vision gave a direction. The research gave a destination.

Discovery defined what to build and why. The next case study is how we shipped it: AI evals, mid-sprint crises, and getting to App Store in 12 weeks. Read it →