Defining the UX of
Agentic AI Capabilities
At AWS Bedrock, my work sits where multimodal enterprise data meets increasingly autonomous AI behavior. Rather than building standalone agents, I focus on designing agentic capabilities within RAG systems, such as multi-step retrieval, reasoning, and self-correction across documents, structured data, and other enterprise sources. My role is to translate these capabilities into clear experiences that help people understand what the model is doing, step in when needed, and confidently use AI with sensitive information.
20 months
Timeline
AWS AI (Bedrock)
Platform
Lead designer
Role
My scope and responsibilities
I work at the intersection of UX, Product, Engineering, and Applied Science. My responsibilities include:
Leading UX strategy and design for Bedrock Knowledge Bases
Translating early AI capabilities into production-ready user experiences
Partnering with science teams to shape how model behavior is exposed to users
Influencing product direction, not just interface design
Designing for trust and traceability
As retrieval and reasoning becomes more multi-step and agentic, trust becomes a core UX challenge. I focus on:
Making model reasoning visible instead of a black box
Showing what data is retrieved and why it is used
Designing clear points where users can review, intervene, or correct outcomes
Multimodal RAG Evolution
I led the design strategy to expand Bedrock Knowledge Bases beyond text-only retrieval. This work enabled customers to:
Use images, diagrams, and diverse data formats alongside documents
Reason across multiple data types within a single AI response
Unlock insights from enterprise data that was previously hard to query
How I Work
In a fast-moving environment, I establish ways for design to work in parallel with model and feature development. This includes:
Prototyping AI workflows and model behaviors alongside data scientists experiments
Using design artifacts to explore system behavior, not just screens
Helping shape the functional roadmap early, before decisions are locked in
Impact
Elevate Bedrock Knowledge Bases from basic retrieval to agentic, multimodal RAG
Improve enterprise trust through transparency and traceability
Position UX as a strategic partner in AI product development
Strategic Reflections
Designing at the Intersection of Science and UX
Working closely with data scientists was new territory for me at first, but it quickly became one of the most important partnerships in this work. What I learned:
In generative AI, user experience is often shaped by how the model reasons internally
Design decisions cannot be separated from model behavior, retrieval logic, or system constraints
Trust and usability depend on making invisible AI processes visible to users
My role was to bridge design and science by:
Translating model capabilities into clear user mental models
Helping surface complex retrieval and reasoning in ways that felt intuitive and reliable
Ensuring technical depth did not come at the cost of usability
Introducing New AI Paradigms into a Mature Platform
Building agentic RAG capabilities inside an established platform like AWS Bedrock came with unique challenges. The tension I navigated:
Pushing forward new, non-linear AI workflows
While maintaining the stability, predictability, and compliance enterprise customers expect
My focus was on:
Integrating agentic reasoning into a structured, API-driven ecosystem
Designing experiences that support multi-step retrieval and reasoning without overwhelming users
Preserving consistent performance and clear boundaries as system complexity increased
Designing Mental Models for Frontier Technology
As AI systems become more capable, clarity becomes more important, not less. I focused on:
Defining mental models that help users understand what the system is doing and why
Making advanced AI behavior feel like a natural extension of an existing platform
Ensuring new capabilities felt robust, dependable, and enterprise-ready
This is the space where I do my best work: translating frontier technology into experiences that feel thoughtful, trustworthy, and seamlessly integrated into world-class ecosystems.
Hear from our experts
Watch demo by our Solutions Architect