Vice Versa

Vice Versa: a studio rebooted, AI-native

Co-founding and repositioning Vice Versa as an AI-native design & engineering studio, where design and engineering shape each other, while agents do the heavy lifting in between.

Problem

Independent studios get squeezed. Discovery is slow, delivery slides into a commodity, and design & engineering too often work past each other across a wall. Vice Versa's founding bet was the opposite: reciprocity, where both crafts improve each other by design, which is where the name comes from. The next bet is a far sharper one: Go from legacy development agency into a fully AI-native one, before the market makes it the baseline.

Outcome

I reset the studio's positioning and rebuilt how it discovers & ships around AI. Four feats: PluginOS (Figma toolchain for AI agents that cuts token cost by roughly 300x), Design-superpowers (a suite of 25 specialized agents around design), VV-sitepack (a solution that allows designers to deploy changes directly from Figma) and Panopticon, an end-to-end AI & n8n platform that runs the whole agency (WIP). These new capabilities allowed the agency to move out of the legacy agency model and spearhead into a highly efficient & impactful AI powerhouse, where top quality meets record-breaking delivery speed.

Snapshot

Client: Vice Versa
Years: 2024-now
PluginOS: ~300x cheaper
Design-superpowers: 25 Agents
Headcount: 2 Dev, 1 Design
Annual delivery: 60+ solutions

Premise

Reciprocity, by design

Vice Versa was built on a plain idea: design and engineering should improve each other, not negotiate across a wall. That reciprocity is the name and the operating principle. Most studios still split the two and pay for it in slow handoffs and work that drifts from intent. Repositioning Vice Versa meant taking that founding idea and making its next version explicit, AI-native from discovery through delivery.

Operating system

An agency that runs on specs, not meetings

AI-native is an operating model, not a tool subscription. At Vice Versa the canonical store is Git and Markdown: if it isn't in a repo, it doesn't exist. Product writes full specs with acceptance criteria; agents build from those specs; humans sit at the bookends, framing the problem and approving the diff. Every repo carries its agent configuration, every PR gets an AI review before a human reads it, and something ships every week. Panopticon is the platform that watches it all run: alerting when a pipeline goes quiet, a daily brief of everything in flight, and HITL gates in Slack, where decisions actually get made.

Tooling

Tools born from friction, not roadmaps

Every tool the studio ships started as a bottleneck in real client work. PluginOS made agentic Figma work affordable (~300× cheaper than the official MCP). Design-superpowers packaged the design lifecycle into commands and sub-agents. VV-sitepack closes the last gap: design tokens, components, and typography sync from Figma into a client's CMS through pull requests and CI. A designer changes a colour in Figma, and the change lands on the live site through the same review pipeline engineers use. Design and engineering stop negotiating across a wall because they literally share one. Read the full tooling story ↗

Leverage

Quality that usually gets cut for time

The point of all this is not novelty, it is leverage. For clients, it means robust, scalable systems delivered faster and at lower cost than a traditional studio can manage. For the people who use what those clients ship, it shows up as the quality that usually gets cut for time. AirHelp's design system now rolls out across its website, dashboards, and mobile apps with dark mode, dynamic scaling, and accessibility in the foundations, so a traveller claiming compensation gets an experience that works on their device, in their conditions.