General Counsel AI

When I joined, GC AI was a single chat box bolted onto a model — functional, but visually cold and clinical. As the product evolved, the cracks started showing: sidebars grew disorganized, navigation became confusing, and the architecture couldn't scale with new features.

The 2025 rebrand by Monopo brought a warmer palette and updated styles. I used that opening to rethink the platform's visual & UX design from the ground up.

Before redesigning anything, I met with GC AI customers, in-house counsel, to observe how they used the product and probe for friction points.

Their feedback mirrored what I'd already flagged in my own review:

• Chat organization was hard to navigate
• Uploading large document sets was cumbersome
• Prompts were difficult to surface and reuse
• Split sidebars created visual noise

With the direction set, I first conducted an audit of existing components like buttons and inputs. My main findings were that:

• There were too many variants for each component
• The actual Figma library wasn't built as efficiently and lacked new features like slots, advanced variables, etc.
• The color palette of the previous design system used cooler tones, while the new app needed to look more like the marketing page (with the revamped, warmed color palette)

The heart of GC AI is the workspace, where a user poses a question and the assistant works through it. I designed the default state to feel calm and focused, with plenty of room for the response and its citations to breathe.

A three-layer surface hierarchy

The interface is built on three distinct tonal layers that create depth without noise.

The shell — a deep navy #003047 — wraps the entire application and contains the sidebar navigation. It doubles as a brand carrier: navy is GC AI's primary color, so its presence at the outermost layer means every screen opens with a consistent identity signal.

The main section uses two tints of beige — a warmer #FAF3EC for the sidebar and a near-white #FFFCF9 for the content — to quietly separate navigation from work without resorting to hard borders or heavy contrast.

Finally, interactive elements like text fields, buttons, and the chat composer use pure white #FFFFFF to lift off the beige canvas.

Legal work is iterative, so drafts don't always survive first contact. I designed an editing flow that fits directly inside the messaging interface. It lets users refine their input, track changes as they go, and step back through previous versions when needed.

To make the workspace feel like a real product and not just a tool, I also designed the surrounding surfaces too.

Company profiles

Company profiles give the AI persistent context about an entity — its structure, risk tolerance, and more. In-house counsel often juggle multiple subsidiaries or affiliated companies at once, so being able to switch contexts at the top of every chat keeps the assistant grounded in the right company before any work begins.

Account menu

The account menu is where the user manages their profile, settings, and organization. It also doubles as a natural home for new features. The Personalization toggle, for instance, is surfaced here with a "New!" tooltip, making it a low-friction way to introduce capabilities.

When users work through long, document-heavy chats, their context window – the amount of information the AI can hold in memory at once — starts to fill up. GC AI surfaces this transparently through a usage indicator in the chat composer, showing a live percentage and a breakdown.

The indicator tracks two context tiers: normal (up to 200K tokens) and extended (up to 1M). It's displayed as a segmented bar where green #74A75A and purple #B721B7 fill in as the chat grows. When things get critical, the bar shifts to amber #DE8703 and red #D16268, and the status label updates to match. Color carries the signal, but for accessibility, I added icons and explicit text labels to back it up so the indicator works for everyone.

Because legal research can take time, I paid close attention to the in-between moments.

Queueing messages

A queue lets users line up work and keep moving instead of waiting on the AI between sends. Queued messages stack above the composer, and each one can be edited, saved as a prompt, or jumped to immediately by interrupting the current response.

The chat composer was getting crowded. To fix this, I redesigned how users add items — files, company profiles, skills, and more — replacing deeply nested dropdowns with a slide-out panel. Selecting a category opens a beige detail panel on the side, so users can browse and pick without losing their place.

Being the first designer at GC AI meant wearing every hat: researcher, product designer, and sometimes front-end developer. This was particulary useful when I was prototyping complex interactions like the pill loading states.

The biggest lesson for me was that in the legal world, polish isn't decoration, it's credibility. Every pixel of precision earned us a little more trust, and watching the product grow from a single chat box into a real workspace was one of the most rewarding things I've worked on.