Self-initiated concept · Internal tooling · Design-org leverage
Frame - an AI workspace that gives design teams leverage
A concept for the shared foundation a design org actually needs as it scales: shared agents and skills, a research engine, and a place where each squad's context composes into something coherent - instead of scattering into parallel, drifting efforts.
A note on this project: Frame is a concept I designed for myself, not shipped product. I built it because it's the clearest way to show how I think about the problem I most want to work on - turning AI into durable leverage for a whole design organization. Every part of it is grounded in workflows I already use - Claude Code, Cursor, and MCP; none of it is speculative interface for its own sake.
Overview
As a design org grows past a handful of squads, the bottleneck stops being individual craft and becomes coherence. Research gets re-run because nobody could find the last study. The same pattern gets re-solved three ways. Design decisions and their rationale evaporate the moment the project ships. Frame is a concept for the internal workspace that fixes this - not a knowledge base you have to feed, but a shared, AI-powered layer that turns the org's scattered context into shared, reusable leverage: agents and skills any designer can invoke, a living research engine, and a record of decisions that composes across squads.
Problem & context
At scale, design drift is an information problem before it's a taste problem. Each squad accumulates real context - user research, decisions, patterns, rationale - but that context stays trapped inside the squad. The org ends up with parallel efforts that quietly contradict each other, and the most experienced designers spend their time as human routers, ferrying context between teams and re-explaining decisions that were already made.
AI tooling has mostly been pointed at the individual designer - generate a mock, summarize a doc. The bigger opportunity is org-level: use AI to make the connective tissue between squads real, so that moving fast and staying coherent stop being in tension.
Why it mattered
This is precisely the work that raises the ceiling for everyone and carries no single team's name on it. Get it right and every designer starts each project further ahead, leadership can see across squads without a status meeting, and the org's judgment accumulates instead of resetting. It's the difference between a design team that scales linearly with headcount and one that compounds.
My role & approach
I designed Frame end to end as a systems concept: the problem framing, the architecture, the core surfaces, and the point of view about where AI belongs. I deliberately kept it grounded in workflows I already run day to day - building with agents and skills, prototyping in code, doing workflow-walkthrough research - and asked a single question throughout: what would make this reusable for a whole team, not just faster for me?
Key decisions
Agents and skills are the unit, not documents.
Instead of another repository people forget to search, Frame exposes shared skills - "synthesize these interviews against our existing findings," "check this flow against our patterns," "draft the decision record." Encoding judgment into invokable skills is what turns one designer's method into org-wide leverage.
A research engine with memory, not a folder of decks.
Findings are stored so they surface at the moment of relevance - when someone starts adjacent work, the prior research comes to them. The point is to stop the org from paying for the same insight twice.
Reconcile patterns across squads, surface the seams.
Frame actively flags where two squads are solving the same problem differently, so a human can decide whether that divergence is intentional or drift. It makes the invisible connective-tissue work visible and actionable.
Draw a hard line between AI's job and human judgment.
AI does retrieval, synthesis, drafting, and reconciliation - the high-volume, low-taste work. Humans decide what's true, what matters, and what to ship. Frame is designed so that line is always legible, because a leverage tool that blurs it stops being trustworthy.
Tradeoffs & constraints
The failure mode of a tool like this is becoming a maintenance tax - one more system to feed. So the core constraint I designed against was zero-effort capture: Frame has to learn from work people are already doing, not demand new upkeep. The second risk is over-automation eroding trust; I chose to keep humans firmly in the judgment seat even where AI could plausibly decide, because in a leverage tool, trust is the whole product. And a concept has real limits: the honest next step is a narrow pilot - one team, one skill - measured on adoption, not applause.
What this concept demonstrates
- Principal-level systems thinking: designing the foundations, not the surface
- AI used for genuine org leverage - shared agents, skills, and research memory
- A clear, opinionated line between where AI adds value and where human judgment must win
- Cross-squad alignment treated as a first-class design problem
- Comfort operating at the level of standards and infrastructure, not one-off artifacts