Enterprise AI
AI Product Design
Knowledge Search
Designing Trust Into Enterprise AI Systems
I designed a knowledge management experience that helps teams understand, monitor, and improve the information behind AI agents.
Platform
Enterprise web platform
Coming soon
01 · Problem & context
AI agents need reliable knowledge
Enterprise AI agents are only as reliable as the knowledge they can access.
Teams need visibility into what information is available to an agent, how it has been processed, and where something needs attention.
The Knowledge Base needed to become more than a place to store information. It needed to provide the foundation for managing the knowledge behind AI agents while remaining usable for both technical and operational teams.
10,000+
knowledge sources feeding agent knowledge
63%
of agent failures traced to stale or conflicting content
0
tools showing teams what their agents could actually see
02 · Approach
Understand the system before designing the interface
I began by understanding how AI agents process enterprise knowledge, including ingestion, connections, retrieval, and evaluation.
That research led to three mental models based on the engineering lifecycle:
Operate
Understand the current state of the knowledge and monitor its health.
Build
Manage connections, ingestion, and the flow of information into the system.
Optimize
Evaluate performance, identify problems, and improve the knowledge behind the agent.
The design strategy became deliberately focused on transparency and density. Users needed access to technical information, but they also needed a clear way to understand what it meant.
03 · Key decisions
01
Simplify through structure
Rather than hiding complexity, I organized it around the questions users were already asking.
02
Make the system observable
Users needed confidence that their data was being processed correctly before they could trust an AI agent to use it.
03
Turn problems into actions
Errors and evaluation results needed to help users understand what had gone wrong and what they could do about it.
04
Design for technical confidence
The audience valued visibility and control. The interface therefore prioritized useful information over unnecessary simplification.
04 · Outcome
Making complex AI understandable
The work received exceptionally positive feedback, particularly around my understanding of AI-agent systems and my ability to translate complex technical concepts into approachable interfaces.
While I don't have visibility into the final implementation, the project demonstrated my ability to quickly understand an unfamiliar AI system and turn that knowledge into a clear, human-centered product experience.