AI Navigation Assistant
Agentic AI app design and developments to simplify, accelerate, and improve accuracy in manual data collection
Developing an agentic AI assistant for Wolters Kluwer Enablon to replace traditional enterprise navigation with a conversational interface, transforming how users interact with complex EHS software.
Challenges
The identified critical navigation issues affecting our users’ daily platform usage:
Outdated interface: Users described the current platform as "clunky" with a outdated aesthetic
Complex navigation: Difficulty finding and accessing key functions through menu structures
Cognitive load: Users struggled to remember where specific features were across the platform
User Conflict: Field operators needed rapid, mobile-friendly access, while EHS managers required deep, accurate reporting.
Opportunity
Rather than incremental redesign, we saw an opportunity to leverage emerging AI capabilities to fundamentally transform user interaction.
Key Insight: What if users could simply tell the system what they want to do, rather than hunting through menus?
Design Vision
From Static to Conversational
I transforming the traditional point-and-click interface into an intelligent conversational experience.
Natural Queries: "How can I report an incident?" or "where to find the audits?"
Contextual Response: The system identifies intent and provides direct, actionable links.
Exploring ways to display links in AI response
User flow for orchestrator to connect with another agent
Using AI multimodal prompting alongside imported Figma design frames and design system components, I iterated between the prompt tool and visual editor to rapidly prototype solutions and explore concepts
New product feature: user onboarding by AI agent
RAG (Retrieval-Augmented Generation) for hybrid AI answer and search
Design Research & Strategy
Understanding AI UX in 2026
To design effectively for this new paradigm, I researched:
Current state of AI interfaces: How users interact with conversational AI and their expectations.
AI capabilities and limitations: Understanding what happens under the hood to design realistic experiences
Key Design Considerations
Detecting User Intent
Developing techniques to help users articulate their needs clearly.
Building Trust
Fostering user confidence in AI-generated responses by setting guardrails against inaccuracy and hallucination.
Dynamic vs. Static Interaction
Designing flexible, proactive experiences that adapt to user needs rather than rigid, deterministic flows.
Individual Personalization
Tailoring AI responses to align with user preferences and frequently-accessed features.
Solution Architecture
I designed a natural language navigation system that:
Interprets user intent from conversational queries ("go to," "create," "search")
Maps requests to specific Enablon URLs.
Uses hierarchical paths, e.g. breadcrumbs, URL
Respects user permissions and maintains context for follow-up questions
Design Patterns & Innovations
Novel Interaction Patterns
Proactive accommodation: System anticipates user needs based on context and history. Examples include prompt suggestions.
Information Architecture
Designed hierarchical navigation responses that provide:
Clear breadcrumb paths showing location context
Multiple options when queries are ambiguous
Impact & Next Steps
This project shifted enterprise interaction from mechanical menus to dynamic conversation.
Roadmap: Conducting user testing with satisfaction and time-saving metrics.
Scalability: Establishing a design framework for future "augmented search" and hybrid AI services.
Future Vision: An Intention Detection Layer will act as an orchestrator, directing users to specialized agents like a Report Writer or Safety Adviser.
Reflections
Proactive Paradigms: Moved from rigid, deterministic flows into engaging, dynamic, proactive experiences
Designing for Uncertainty: Since AI output is unpredictable, I focused on a holistic research approach—analyzing existing models to design robust fallback mechanisms and error states.
Managing user expectations for natural conversation under the technical constraints.
