AI-driven Project Management Tool

Led the design of a digital hub empowering Atlanta's LGBTQ+ community, now live on the official Atlanta Pride site

Project Details

Company: Neudesic (Now IBM) | Team: Cross-functional Product & Engineering

Role: UX Intern | Timeline: Aug 2023 - Nov 2023

Tools: Figma, Azure DevOps, JIRA, openAI

The Challenge

Neudesic was evolving its internal project management accelerator to integrate AI-driven work-item generation within Azure DevOps.


However, the platform needed:

  • Greater visual and structural consistency across modules

  • Clearer role-based access logic

  • A predictable framework for introducing AI automation

  • Scalable UI components aligned with Microsoft Fluent


The challenge wasn’t just adding AI, it was ensuring the system could support automation without sacrificing clarity, control or usability.

How might we introduce AI-driven efficiency into an enterprise tool while preserving structure, trust, and scalability?

The Value I Delivered

Led the design of role-based AI workflows and a 20+ component system that improved clarity for users and reduced engineering effort by 40%.

Scalable, permission-aware AI workflows for long-term growth

For the system

20+ component design system → 40% reduction in dev effort

For Neudesic

AI-assisted templating + 5-role access flows for consistent, controlled work generation

For the users

The Solution

Rather than tackling one isolated feature, I contributed across three structural pillars.

Building a unified icon and component framework

Built Hi-Fi prototypes to generate resuable templates for users to generate work items to maintain consistency

Designing structured access flows based on roles within the team

Framing the Challenge

I audited the product’s UI, permissions, and workflows to understand how structure needed to support AI-driven automation.

When I joined, I conducted an audit of:

  • Existing modules and UI inconsistencies

  • Duplicate icon usage

  • Permission-based visibility logic

  • Points of friction in work-item creation


What became clear was the following:

Inconsistent visual language reduced clarity

Role-based access rules created complexity in flows

AI automation required strong structural guardrails

User Research & Discovery

Closely collaborated with product owners and engineers to ensure AI integration aligned with real DevOps workflows and user expectations.

Because this was an internal enterprise tool, research relied on:

  • Stakeholder interviews with product owners

  • Working sessions with engineers

  • Observational walkthroughs of DevOps workflows

  • Internal feedback on early AI templating concepts

The team had the two following main findings:

Permission restrictions must feel logical, not arbitrary

Developers value speed, but not at the expense of control

Designing the Solution

Building Structural Consistency

(Icon System + Components)

I led the creation of a unified icon library:

  • Standardized grid and stroke rules

  • Eliminated duplicates

  • Aligned with the tool's visual language

I also contributed to foundational Azure DevOps extension components:

  • Buttons

  • Dialogs

  • Dropdowns

  • Tables

All aligned with Microsoft Fluent to ensure ecosystem compatibility which created a predictable design baseline for future feature growth.

Designing AI-Assisted Templating

for Work Items

The product aimed to use OpenAI to auto-generate project work items.

The key risk: automation feeling opaque or irreversible.

My focus areas:

  • Clear “Review” states before final submission

  • Editable AI-generated fields

  • Transparent labeling of AI-generated content

  • Reversible actions (approve, modify, discard)

Rather than presenting AI as autonomous, the design positioned it as a structured assistant.

This maintained user confidence while increasing efficiency.

Structuring Role-Based Workflows

Different roles required different levels of visibility and control.

I:

  • Mapped permission-based interface states

  • Sketched multiple conceptual access models

  • Validated flows with engineers for feasibility

The final solution balanced the following:

Administrative oversight

Developer efficiency

Reduced unnecessary UI exposure

Reflection

This project taught me that introducing AI into enterprise systems is less about innovation — and more about discipline.

  • AI requires strong design systems

  • Automation demands explainability

  • Role clarity reduces cognitive overload

  • Consistency builds trust

Wearing both the Systems Builder and AI Tinkerer hats allowed me to approach the problem holistically — not as isolated features, but as an evolving product ecosystem.