Agentic AI Platform

My Role
Lead Product Designer
Target Audience
Design
YEAR
2025
Overview

SimplifyX needed to create a unified platform that would make AI agent technology accessible to enterprise users. The challenge was to design an interface that could handle complex AI capabilities—from document parsing to voice agents and workflow orchestration—while remaining intuitive enough for non-technical users to create, manage, and deploy AI agents across their organizations.

Problem Statement

- Designing for multiple agent types (Voice, Doc, Robotic, Chat, Email, Web, Calc, API, UI) with distinct capabilities - Creating a scalable information architecture that could grow with new agent types - Balancing powerful enterprise features with ease of use - Establishing clear visual patterns for agent status, versioning, and management - Designing complex configuration interfaces that felt approachable

Process

Research & Discovery
User Research Findings:Through interviews with enterprise users, CTOs, and business analysts, I identified several critical pain points: Fragmented Tools: Users were juggling multiple platforms for different automation tasks
1. Steep Learning Curve: Existing AI tools required significant technical expertise
2. Poor Visibility: Users couldn't easily track agent performance or status across their organization
3. Collaboration Barriers: Teams struggled to share and collaborate on agent configurations
‍

Competitive Analysis: I analyzed platforms like Zapier, Make.com, and custom AI solutions to understand: - How users navigate complex automation workflows
- Common patterns for agent/workflow management
- Information density vs. clarity trade-offs
- Onboarding approaches for technical products

Design Goals

  1. Unified Experience: Create a single platform where all agent types feel cohesive despite their different capabilities
  2. Progressive Disclosure: Show simple options upfront while making advanced features accessible when needed
  3. Clear Mental Models: Help users understand what each agent type does and when to use it
  4. Collaboration-First: Enable teams to share, version, and iterate on agents together
  5. Enterprise-Ready: Include robust permission, audit, and governance features

Design Process

Information Architecture

I restructured the navigation to create clear separation between agent types while maintaining discoverability:

Primary Navigation:

  • Home (Dashboard), Workflows, Agents (expandable menu), Chat Agents, Email Agents, Voice Agents, Web Agents, Robotic Agents, Calc Agents, API Agents, Doc Agents, UI Agents, Tasks, Reporting, Users

This hierarchical structure allowed users to quickly find the agent type they needed while understanding the full breadth of capabilities available.

Agent List Views

For each agent type, I designed consistent list views that displayed:

  • Project names with clear hierarchical sorting
  • Creation dates for temporal context
  • Status indicators (Active, Testing, Deactivated) with color coding
  • Owner information with profile pictures for accountability
  • Version numbers for tracking iterations
  • Quick actions (Archive, Download)

Design Decision: I used a table layout rather than cards to maximize information density—critical for enterprise users managing dozens of agents.

Agent Builder Interface

The builder interface was where complexity really lived. I broke it down into three clear tabs:

1. Builder TabThis is where users configure what the agent actually does. For example, in the Doc Agents:

  • AI Model Selection: Dropdown to choose the LLM (e.g., "Simplify AI")
  • Document Type: Auto-detect functionality with manual override
  • Document Extract Table: Clear display of extracted fields with status indicators (Valid/Invalid) and values
  • Source Document Viewer: Split-screen view showing the original document alongside extracted data

2. Agent Background TabConfiguration for agent behavior and integrations

3. Call Settings Tab (for Voice Agents)Voice-specific configurations like language, persona, and conversation flow

Key Interaction Pattern: I used a left-panel configuration approach with real-time preview on the right. This allowed users to see the impact of their changes immediately.

‍

Branding + Visual Design

To ensure consistency and scalability across the SimplifyX platform, I developed comprehensive design and brand guidelines that served as the foundation for all design decisions. The guidelines emphasized clarity, trustworthiness, and innovation—key attributes for an enterprise AI platform. The visual language balanced technical sophistication with approachability. The primary brand color, a vibrant purple (#7C5CFF), was chosen to convey both creativity and cutting-edge technology while standing apart from traditional enterprise blue palettes. This was complemented by a carefully curated secondary palette: greens for success states, reds for errors or critical actions, and oranges for in-progress or testing states. These colors were tested for accessibility, ensuring WCAG AA compliance across all text and UI elements. Typography followed a clear hierarchy using modern system fonts that ensured readability across different screen sizes and devices. Headers used bold weights to create clear section breaks, while body text maintained comfortable line heights and letter spacing for extended reading sessions—particularly important given the technical nature of agent configurations. The component library I created included reusable patterns for tables, forms, buttons, badges, modals, and navigation elements. Each component had defined states (default, hover, active, disabled, error) and spacing rules using an 8-point grid system. Icons were designed with a consistent stroke weight and corner radius, creating a cohesive visual language throughout the interface.

Outcome

Designing for AI platforms requires balancing power with approachability. The most successful patterns were those that showed users exactly what would happen before they committed to a configuration. Making the invisible visible—whether through conversation previews, document extraction tables, or workflow steps—built user confidence and adoption.

Prev
Next
Scroll
Profile Pic

Say Hello.

shettygahana@gmail.com
© Gahana Shetty.  Powered by Webflow
ChangelogLicense
Scroll