Robin
Building an AI assistant for the firm's productivity experts.
The opportunity
A tool built for experts
The challenge
Executive assistants are masters at productivity, typically managing 15-20 delegated accounts at once. Each with its own inbox, calendar, and set of sensitivities. This meant crafting a tool to augment their skills with real control over what an assistant touched, and integrations deep enough to actually fit how they worked. Additionally there was an emotional layer, with some executive assitants were genuinely worried AI would replace their role.
The opportunity
We treated that complexity and hesitation as a reason to slow down and co-create, running research and testing with the EAs, rather than designing for them from a distance. That trust building sat alongside a bigger bet: Robin was built in house, for the flexibility and integration depth a generic assistant couldn't offer. When Copilot arrived anyway six months later, one of several AI tools the firm was racing to land on, the case for Robin had to be proven on real depth, not just cost.
From hands on delivery to leading the team
For six months I was the most hands on designer on Robin, setting the core interaction model and designing how agents handled scheduling, email, to do lists, and handovers. As the program scaled I stepped into a lead role, overseeing 7 designers across 3 different AI tools.
year
2025-2026
timeframe
1 year
role
UX/UI Program Lead
tools
Figma
category
UI/UX
The work
Robin scaled its use case as our integrations grew
Key decisions
Decision 1
Control is key
For high stakes tasks like calendar changes, client emails and detailed handovers, EAs needed to see and control what was happening, not just trust a black box. This approach tested really well with users, giving them visibility & control was part of what made the tool usable in practice.
Decision 2
Adapt & scale
I chose to create a design system adapted from Microsoft's Fluent UI, providing us with agility in the begining of the program and a familiaryt for users already embedded in the Microsoft ecosystem. As the usecases took shape we built custom compoents to scale the system.
The impact
Robin is early, still proving adoption at 500 users. The system built for it isn't. It's already the shared design language for the firm's wider AI tooling, an outcome that's arguably bigger than Robin itself.
01
02
