1
About the project
Mobius needed a scalable way to create interfaces for many AI agents across tenants, industries, and use cases. I designed MetaUX, a graph-driven experience pattern that translates agent knowledge graphs into reusable product interfaces for both operators and builders.
Impact
Reduced baseline agent UI setup from roughly 1 month to 1 day for journeys covered by the pattern.
My role
Agent Experience Engineering Manager
2
What was to be built
Mobius builds AI-powered digital transformation products through agentic marketplaces. Enterprises can discover, purchase, configure, and deploy agent products. Each agent has capabilities, business logic, knowledge graphs, and governance rules.
CHALLENGE
The challenge was that every agent deployment still needed an interface. If every interface had to be manually designed, the product could not scale.
3
Problem Statement
The marketplace could identify and deploy new agent products, but the interface layer was still manual. Each new agent or tenant variation risked becoming a custom design effort. This created a bottleneck in scale, consistency, and future automation.
STRATEGY PROBLEM
If the marketplace could generate agent products, but interfaces still required manual design, the agentic loop was incomplete.
PRODUCT PROBLEM
Mobius needed a reusable product architecture for agent interfaces across different tenants, personas, and agent capabilities.
UX PROBLEM
Designing each agent UI manually created inconsistent journeys, repeated patterns, and slow delivery.
4
My Role & Responsibilities
I led the UX architecture and pattern definition for MetaUX, translating graph-based agent structures into reusable interface patterns.
5
Core Insight
The breakthrough was realizing that the interface did not need to be reinvented for every agent. Each agent already had a structured graph of intents, entities, actions, results, laws, processes, and traces. Instead of manually inventing screens for every agent, the UI can be composed from graph node types and relationships.
6
The MetaUX Pattern
1
Each Business Intent becomes a dashboard tab. Users switch between dashboards based on what they are trying to achieve.
Examples
Smart home automation agent
Optimize Energy & Cost
Run Office Routines
Manage Devices
DevSec Ops automation agent
Govern Policies
2
Examples
Smart home automation agent
Devices
Routines
DevSec Ops automation agent
Policies
Infrastructure
Pipelines
3
Examples
Smart home automation agent
Energy Usage Pattern vs Baseline
Peak Load Hotspots
DevSec Ops automation agent
Builds Trend
Policy Violation Trend
4
Events becomes action
Event nodes become contextual actions specific to the entity.
Examples
Smart home automation agent
Troubleshoot device
Create routine
DevSec Ops automation agent
Request approval
Run scan
Edit workflow
5
Business Traces become activity logs, audit trails, and action history.
6
7
Operator Experience
For operators, MetaUX translated graph complexity into task-first interfaces. Users could move from intent to object to action without needing to understand the underlying graph model.
Operators do not need to understand the graph
They see dashboards, objects, actions, charts, and traces
The pattern reduces cognitive load
The flow stays consistent across agents
8
Builder Experience
Builders needed a different experience from operators. While operators use the agent, builders shape the agent. The builder experience exposes the graph structure, so agent owners could manage intents, laws, results, processes, workflows, schemas, and executable artifacts.
They shape the agent’s capabilities, governance, laws, and executable artifacts. The graph becomes a product surface. Governance is part of the product model.
9
Product System: How MetaUX Scales
MetaUX turned repeated design work into a product architecture. Instead of designing each agent UI from scratch, teams could apply the same graph-to-interface grammar and let each agent’s structure determine the experience.
Product design contributions
10
Strategic Value
Strategically, MetaUX acts as a bridge between today’s human-designed interfaces and tomorrow’s agent-generated experiences. Because the pattern is graph-based, it creates a repeatable structure that can eventually be learned and automated by a UX agent.
Strategy points
Reduces delivery bottlenecks
Supports tenant-specific variation
Makes interface generation more repeatable
Creates training structure for a future UX agent
Treats UX as product infrastructure
11
Impact
For baseline journeys covered by the MetaUX pattern, agent UI setup time dropped from roughly one month to one day. More importantly, the interface layer became a reusable product capability instead of a repeated custom design effort.
12
What I Would Improve Next
Tulika Pandey ⏤ 2026