We identified new entry points into the market, and several major clients expressed interest in enterprise-level AI workspaces. We pivoted toward the enterprise sector.
The main design challenge lay in introducing role-based access control (RBAC) to what was previously a personal workspace without any roles. Unlike the workspace itself, each AI Agent also has distinct, independent roles, making the flow design particularly complex and challenging.
Enterprise scale isn't a bigger model — it's a permissioning layer that doesn't drown the admin.
- Workspace-level roles and seats
- Per-agent roles, independent from the workspace
- One cohesive account model across surfaces
Log in, sign in, and pick a workspace.
Worked with Auth0 to build a seamless and enterprise-secure login / sign-in flow — account creation, SSO, and workspace selection live on a single continuous surface.
One session, many workspaces.
Users can switch workspaces seamlessly, without needing to log in each time. The switcher preserves session context so work and personal environments feel like rooms in the same house.
Put the primary action first.
Moved “Message to AI” to the top, as user interviews indicated the messaging interface is the most frequently used feature. Secondary actions recede; the main loop becomes a single keystroke away.
Admin controls without the admin overwhelm.
The workspace owner can manage members, assign roles, and add seats as needed. The surface prioritizes the five everyday decisions and hides the long tail behind disclosure so admins don't drown in settings.
Train agents the way teams already talk.
Users can communicate on the platform via DMs and channels — and can train their agents directly through messaging. Training becomes an extension of the conversation, not a separate tool trip.
A workspace the enterprise can actually adopt.
Shipped to production with enterprise customers on multi-million-dollar contracts. The RBAC model — workspace-level and per-agent — now anchors how every new enterprise surface is designed.