Foundational UX For Agentic AI
Creating a first-of-its-kind service ticketing system and AI canvas, plus conversational interface experience.
My Role
Contract designer
The Problem
DevOps work, landing on engineers who were never trained for it.
DuploCloud is a startup that automates cloud infrastructure provisioning, management, and compliance, often described as “DevOps-as-a-Service.”
Their customers are small-to-medium size businesses that need to stand up and maintain cloud environments but don’t have dedicated DevOps teams. Internal engineers are left to handle this work themselves, often for the first time. Those engineers need two things at once:
Comprehensive guidance during setup, so installations are complete, secure, and performant
Continuous observability & remediation support post-install, so issues get caught early and fixed fast
DuploCloud, being in the early startup stage, wasn’t designed to support these needs. And needed help to quickly conceptualize solutions to inform the platform buildout and support customer needs.
Platform Suite
Setting the platform up to scale.
DuploCloud already had a platform offering but needed to extend it to support their new AI suite. They needed a new information architecture and navigation system to support the platform’s ability to scale now and into the future.
Information Architecture
Restructured to account for the new suite hierarchy as well as deeper workflows.
Navigation Structure
Navigational tiers and logical hierarchies makes it easy for engineers to find what they're looking for quickly. And deep links enable quick access across logical branches.
Navigation System
A top-level app switcher partitions the platform into multiple discrete feature sets while allowing full access to everything in a couple clicks.
AI Service Desk
Service ticket management for AI agents
Engineers work with DuploCloud’s AI agents to set up and monitor their cloud infrastructure environments. The AI service desk is the management dashboard that enables engineers to create tickets and sessions with these agents, monitor their performance, as well as usage. A first-of-its-kind experience for AI agent interaction and management by end users.
Research
Without any 1-to-1 features such as this I looked at existing help desk applications to gain insight and inspiration into what makes such a system efficient. Analyzing Jira, ClickUp, Zoho, and Zendesk I was able to synthesize some impactful insights, used to influence this brand new system.
Dashboard
A bird's-eye view of how assistants (agents) are performing and ticket management using a kanban interface and data visualizations.
Ticket Detail
In-depth details about the ticket inquiry, assistant (agent) assignment, discussion thread, and ticket resolution.
Agentic AI Canvas
Half conversational AI, half visual canvas
This is where the work gets done. Engineers can converse with agents on the left side, provide instructions, make requests, and implement actions. And on the right side they can see and interact with summoned statistics, environment entities, and observability signals. They can then either work directly with that data in the canvas or have the agents do it for them. A true hybrid experience designed to build user trust in the agents and expedite resolutions.
Canvas
Agent conversation on the left, interactive artifacts on the right. Providing the efficiency engineers want and the trust signals they need.
Outcome
Three concept systems, delivered in six weeks.
DuploCloud went from a platform that wasn’t built for its new AI suite to a complete conceptual blueprint for it:
A scalable information architecture and navigation system
A first-of-its-kind AI service desk for creating and monitoring agent sessions
A hybrid canvas where engineers and agents work side by side
Together they gave DuploCloud’s leadership a clear product direction and their engineering team concrete, high-fidelity designs to build against on an aggressive release timeline.
