Agentic AI
Pattern System
Year
2025 - ongoing
Company
JPMorganChase
Role
Lead Senior UX Designer
I designed an agentic AI pattern system that helps product teams decide when AI belongs in a workflow, where it should appear, and how users remain in control.
© Agentic Pattern System
Overview
for enterprise AI
AI was appearing across employee tools, from drafting and summarizing to autofilling forms and supporting complex tasks. But similar capabilities were being designed through different entry points, placements, and interaction models.
I led the creation of a shared system that helped teams design AI experiences more consistently, without forcing every workflow to look the same.
CHALLENGE
No shared logic for how AI should appear
ROLE
Led patterns, framework and adoption guidance
SCALE
30+ products and 100+ designers
OUTCOME
A consistent foundation for enterprise AI
The turning point
Across employee products, teams were adding AI to help people draft, summarize, complete forms, and make sense of dense workflows. But each team was solving the interaction model in isolation.
What looked like small placement differences created a bigger product problem: users had to relearn what AI meant from screen to screen.
The reframe
Reusable components could standardize the interface, but they could not tell teams whether AI belonged in a workflow or what kind of relationship it should have with the user.
The framework
Instead of treating every AI feature as a chatbot, I created a framework that helped teams choose the right AI surface based on the user’s task, context, and level of control needed.
Each pattern answered a different workflow need: completing structured input, summarizing dense information, editing in place, or supporting multi-step decisions.
AI Autofill
Pattern A
For repetitive, structured input
AI Summary Cards
Pattern B
For understanding dense information
Inline Actions
Pattern C
For focused content creation and editing
Assistant Side Sheet
Pattern D
For contextual and multi-step support
Decision model
To make the framework actionable, I mapped AI patterns across two dimensions: what the user is trying to do, and how much of the interface the AI needs to occupy. This helped teams move beyond “where should we put an AI button?” toward a more structured decision: whether the experience should live at the portal, workflow, section, or field level.
Scale
The final system connected reusable interfaces with guidance for placement, behavior and oversight. Teams could adapt the patterns to their products while preserving a consistent interaction model.
Foundation
trust principles
Decision guidance
AI role, scope and risk
Patterns
autofill, inline, summary, side sheet, and more
Product application
real employee workflows
Impact
The system reduced duplicated AI design exploration, accelerated product delivery and created more predictable interaction models across employee tools.
Adopted across
products
Used by
designers
Designed for
employees
Unified by
AI pattern system
