Designing for Trust


Knowledge systems aren't just about information.

They're about trust.

The best systems recognize what users need before they know where to look. They anticipate confusion, reduce friction, and make answers easier to find when they matter most.

Better knowledge doesn't just solve problems.

It helps people move forward with confidence.

As content scales, so does the complexity. Content Management Systems help, but only when supported by clear structure, consistent labeling, and thoughtful version control.

That’s where technical writers come in. We design intuitive content architecture, apply meaningful metadata, and make search actually work.

The goal is simple: help users quickly find what they need, and keep things running smoothly behind the scenes.

Information Architecture

The image shows the centralized, user-friendly “front door” to the IT department, known as the Global Service Desk. Designed to streamline support, it includes:

  • Custom icons for quick access to software, hardware, policies, and procedures

  • Interactive tools like the “Ask pITbot” chatbot and Glean search integration

  • Built-in search for fast access to specific content

  • Icons for connecting with different IT teams

Its intuitive design and visuals reduce friction and make it easier for users to get the help they need.

Global Service Desk

AI Knowledge Systems


As organizations adopt AI-powered support and enterprise search, the quality of answers increasingly depends on the quality of the knowledge behind them.

At Amazon, I helped design the source content, information architecture, and retrieval patterns supporting an AI-enabled workforce planning experience used by HR, finance, and business leaders across more than 160 organizations.

The work extended beyond documentation:

  • Structuring content for machine readability and retrieval

  • Testing how AI systems responded to alternate phrasing and user intent

  • Identifying sourcing and metadata issues affecting answer quality

  • Building evaluation frameworks to measure retrieval accuracy and completeness

  • Creating reusable patterns that could scale across products and teams

When AI gives bad answers, the problem is often not the model…

It's the knowledge.

AI Answer Quality Evaluation

To better understand how users interact with AI systems, I developed an evaluation framework containing:

  • User questions

  • Alternate phrasings

  • Implied intent

  • Source mapping

  • Expected answer components

  • Retrieval risks

  • Failure analysis

The framework helped identify gaps between how users ask questions and how knowledge systems interpret them, improving both answer quality and long-term scalability.

When multiple contributors are involved in maintaining a company's wiki, it's essential to prioritize searchability.

The format of this document demonstrates methods that enhance searchability, not only through the internal search function but also for an integrated chatbot.

The following document serves as an illustration of an Atlassian Confluence article, and features:

  • Labeled sections for quick scanning

  • Highlighted notes to capture the user's attention

  • Clear step-by-step instructions

  • Information on next steps for further understanding and assistance.

Click to Enlarge →

Retrieval-Aware
Documentation

Self-Service Experiences

Landing Pages

Designed Confluence landing pages for technical and non-technical audiences that cut the clutter and allowed employees at Cruise to actually find what they need, fast.

Click the images to Enlarge