How transparency-first design transformed 15,000 scattered documents into conversational intelligence, saving 5,000+ hours
Due to confidentiality, respect of privacy policies, and enterprise compliance, some visuals are recreated or simplified for portfolio purposes. A detailed version is available upon request.
It's 2 AM on an offshore platform. An engineer faces a critical decision that could prevent a catastrophic incident but the answer is somewhere in 15,000 documents spanning multiple time zones and business units.
CONTEXT
Role & Impact
Product Design & Strategy | 8 Months, Oct 2024 - Present | Houston, TX
The Stakes: When 30% of process safety incidents stem from not following proper procedures, missing critical information doesn't just cost time—it risks lives and environmental disasters.
The Transformation:
2,200+ engineers
now access institutional knowledge in seconds instead of hours,
$8M
THE PROBLEM
When Critical Safety Data is Buried, Every Second Counts
Why this matters: 30% of process safety incidents happen because people didn't follow procedures. When engineers can't find the right information fast, people get hurt.
15,000 knowledge assets locked across 14 global businesses. Multiple time zones creating knowledge gaps. Retiring engineers taking decades of knowledge with them.
THE CHALLENGE
The Organization Was Not Ready for This Technology
Cultural: Oil & gas is a traditional industry and cautious to adopt new tech. Seasoned engineers needed to trust AI and integrate it with their established workflow.
Resource: Small team (1 product lead, 1 tech lead, 1 design lead, 2 designers, 7 engineers) solving an enterprise-scale problem.
Data: Each query would have to search through various databases before synthesizing insights across different business contexts and maintaining regulatory compliance. Data wasn't guaranteed to be clean.
Trust: Engineers tested AI with questions they already knew answers to before trusting it. They needed to see the reasoning, not just results, since these answers would inform multimillion-dollar decisions that have large impacts on people, machines, and the environment.
THE SOLUTION
I Led UX Design for Two Features That Changed How Engineers Work
Problem 1: Lost Project Context
The situation: "I can't find my conversation from last week, and now I'm starting over."
Engineers were recreating safety analyses, leading to inconsistent risk assessments.
The solution: Workspaces
Project-based organization mirroring how teams actually work
Smart filtering by business unit, document type, asset
Collaborative workspaces where experts curate knowledge
Context preservation so nothing gets lost
Impact: Reduced setup time, eliminated duplicate analysis
Problem 2: Manual Workflows Taking Hours
The situation: "Creating risk reports takes me 6 hours across multiple systems."
Time-intensive reporting meant less frequent safety reviews and delayed responses.
The solution: Agentic Tools
Conversational inputs that feel natural but capture precise requirements
Chain of thought transparency showing how AI processes requests
Automated report generation connecting insights across data sources
Preview-first workflow maintaining conversation flow
Impact: 6-hour process → 1-hour workflow, enabling more frequent safety reviews
Preview-first report generation
Chain-of-thought transparency
DESIGN STRATEGY
3 AI-Specific UX Principles That Built Trust
User testing showed engineers interact with AI differently than consumers. They validate before trusting. This shaped my approach:
RESEARCH
Understanding How Engineers Actually Use AI
I ran comprehensive usability testing with oil & gas engineers. Detailed methodology and findings are available in my dedicated research case study, which revealed 3 distinct user types with different needs:
Truth Seekers
Need to see sources and validate reasoning
Design response: Chat
In-text citations, expandable chain of thought
Delegators
Want to hand off tasks but review outputs
Design Response: Agents
Editable previews, structured handoffs between deliverables
Orchestrators
Coordinate team knowledge and context.
Design Response: Workspaces
Collaborative spaces with group filtering
Key insight: Engineers validated AI responses against their expertise before trusting new information. Transparency about limitations actually increased trust.
Competitive research: Studied Perplexity & ChatGPT (conversation organization and history management), Harvey (structured professional workflows), Grok (chain-of-thought transparency), Microsoft Teams & Figma (workspace navigation and collaborative filtering). Adapted patterns for safety-critical engineering needs.
SCALING DESIGN
Built Components and Navigation to Handle Increasing Complexity
Component Library
Moved from Sketch/Zeplin to Figma as strategic product thinking. Built responsive components handling both simple chat and complex workflows.
Impact: Faster iteration from idea to prototype to development. Rapid testing of new AI capabilities. Consistency across complexity.
Information Architecture
Platform risked becoming overwhelming as we added features. Redesigned navigation to handle complexity while keeping the prompt bar central as a stable anchor.
Principle: Deceptive simplicity, complex capabilities through intuitive interfaces.
V1 - simple but difficult to separate workspaces vs. chats
V2 - separated but having 2 "new" buttons disrupts hierarchy
V3 - good hierarchy but with large history, too busy and difficult to navigate
V4 - filterable to sort different activities and large history but adds cognitive load
V5 - filterable and straightforward labeling to reduce cognitive load
IMPACT
Results That Changed Company Strategy
Strategic recognition: Featured at Capital Markets Day 2025 as flagship AI capability. Influenced roadmap to prioritize agentic workflows. Contributed to press releases as one of four apps accelerating energy production.
What Users Said
Engineer:
"This saved me a day's worth of work—and I did it in an hour. I just ran the agent on a document and it cleaned up my contract by spotting inconsistencies I would have missed."
Subject Matter Expert:
"For the first time, junior engineers can access the same institutional knowledge that took me 20 years to accumulate."
LEARNINGS
I Learned That Trust Comes From Visibility & Domain Understanding Shapes Interface Logic
Trust comes from visibility, not perfection
Engineers cared more about verifying how answers were derived than speed. Transparency built confidence.
Making complexity feel effortless
Conversational workflows balanced natural dialogue with technical precision. The goal was clarity—navigating complexity without friction.
Design systems accelerate velocity
Modular components enabled rapid iteration and experimentation, not just consistency.
Domain expertise shapes interface logic
My engineering background helped me speak users' language. I understood their workflows and data relationships, designing interfaces matching how they process information.
Enterprise AI users are unlike typical AI consumers
Designing for a traditional industry required balancing innovation with rigor, safety, and accountability. Success came from earning trust through clarity.
WHAT'S NEXT
Scaling Intelligence Across Engineering Workflows
I'm continuing to collaborate with engineers and experts to translate additional specialized workflows into AI-powered experiences. The goal remains the same: don't lose clarity, trust, or control while making critical knowledge instantly accessible.