Back to work
Case Study

Building an AI-Fluent
Design Org from
the Inside Out

Took a 150+ person design organization from AI-curious to 100% daily AI adoption in under a year. Without mandates. Inside an enterprise with real security, legal, and compliance constraints.

🏆 AI Excellence Award Red Hat · UX Manager Q4 2024 – Q1 2026 AI Strategy & Org Change
ai-adoption-tracker — UXD org
UXD Org 150+ designers
Phase 1 · Educate Q4 '24 – Q1 '25
Monthly sessions · Local LLM guides · MCP intro
Phase 2 · Experiment Q2 '25
Slack channel (113 members) · Survey n=62 · 3 tiger teams
Phase 3 · Operationalize Q3 '25
VP org goal · Tools inventory · 🏆 AI Excellence Award
Phase 4 · Scale Q3 '25 – Q1 '26
Figma AI pilot → company-wide rollout
100% daily AI adoption · Q1 2026
15 months · From zero AI tooling to 100% daily adoption across the full 150+ person UXD org
100%
Daily AI adoption across the org
150+
Designers enabled
1st
To use Red Hat's streamlined AI pilot process
Q1 '26
Figma AI live for all of Red Hat
🏆
AI Excellence Award, Product Engineering
The Challenge

AI-curious but stuck

In late 2024, Red Hat's UXD org was in the same place as most enterprise design teams: AI was everywhere in the conversation, nowhere in the workflow. Designers worried about being replaced. Enterprise security meant no unapproved tools. Leadership wanted progress but had no playbook. I took ownership and built one: a phased strategy spanning 15 months.

01
Fear vs. curiosity
Designers worried AI would replace them, but were curious about what it could do
02
Enterprise security vs. speed
Red Hat's security posture meant no unapproved tools, no "just try it and see"
03
Enthusiasm gap
A handful of early adopters were sprinting while most of the org hadn't started yet
Phase 1 · Q4 2024 – Q1 2025

Educate

Monthly org-wide sessions

Make AI feel accessible,
not threatening

Before anyone touched a tool, the org needed honest context. I led monthly sessions demystifying AI for designers: what it actually is, what it can and can't do, and what it means for UX work specifically. Use cases mapped to daily reality: getting up to speed on dense engineering concepts, generating first-draft personas, synthesizing research data, and drafting the stakeholder communications that quietly eat designer time.

Presentation slide exploring local LLM options with the team
Exploring local LLM options: approved tools the team could use before enterprise licensing was in place
Time to experiment slide from org-wide session
Monthly sessions kept experimentation visible and social across the full org
Local LLM getting-started guides

Hands-on before enterprise
tools were available

Red Hat's enterprise AI tools weren't available yet, so I wrote hands-on getting-started guides for local LLMs like Cursor and LM Studio: step-by-step docs built for non-engineers who wanted to start now, not wait for procurement. When MCP (Model Context Protocol) started getting traction, I introduced it to the org early so the team understood what was coming before it arrived.

LM Studio getting started guide, step by step for non-engineers
LM Studio getting-started guide: from install to first RAG query, written for designers not engineers
Phase 2 · Q2 2025

Experiment

#uxd-ai-experiments · 113 members

Make it social.
Make it safe to try.

I launched a dedicated quarter of structured experimentation, created the #uxd-ai-experiments Slack channel, and formed three tiger teams (Tooling, Prototyping, Research), directly leading Tooling. I helped shape the VP's formal Q2 2025 org goal: experiment with AI. The infrastructure from Phase 1 meant the org had a running start. By end of Q2, 113 members were actively sharing experiments across research, prototyping, writing, and workflow automation.

#uxd-ai-experiments Slack channel with 113 members and active discussion
#uxd-ai-experiments: 113 members, real experiments shared openly including what didn't work
Org-wide baseline survey · n=62

Real data, not assumptions

I partnered with UX Research to design and run an org-wide baseline survey (n=62) measuring adoption rates, comfort levels, and barriers. The data showed most designers were in the middle: neither uncomfortable nor confident, with the lowest scores around synthesizing UX opportunities from research. That gap shaped Phase 3's priorities.

Baseline survey: general comfort and confidence with AI, n=62
General comfort & confidence · baseline n=62
Task-specific AI comfort levels broken down by use case
Task-specific comfort by use case
NotebookLM knowledge base

Turn Slack noise into
institutional knowledge

Working with our Ops team, I had designers formally submit their AI experiments and use cases. We used that data to build a NotebookLM knowledge base: an AI-powered resource where anyone could ask questions about AI workflows and get answers grounded in the team's real experiments. Scattered Slack conversations became a searchable, queryable knowledge base.

UXD AI Methodology Advisor in NotebookLM, 13 sources from team experiments
UXD AI Methodology Advisor: 13 sources from team experiments, queryable by anyone in the org
Phase 3 · Q3 2025

Operationalize

Q3 2025 org goal · all 150+ people

From experimenting to
using AI every day

The VP set the Q3 2025 goal: use AI day-to-day across the full 150+ person org. I helped shape that goal directly. The phased work from the prior two quarters became the blueprint for how leadership framed the org-wide direction.

Q3 AI goal in Workday: actively contribute to AI journey by integrating AI into daily workflows
Q3 org goal in Workday: "integrate AI into daily workflows" set as a formal performance objective
AI tools inventory · single source of truth

Work within enterprise security,
not around it

Getting AI tools approved at Red Hat requires coordination across the Design Program Office, Legal, Compliance, and Security. I became the first person at Red Hat to use a newly streamlined AI pilot process, defining the path rather than following one. Out of that work came the AI tools inventory: a single Confluence page listing every approved tool (Claude Code, Cursor, Figma, Gemini, Local LLMs, NotebookLM, Miro AI), what wasn't approved yet, and how to get started. No ambiguity. No shadow IT.

AI tools inventory in Confluence: approved vs not yet approved, with getting started guides
AI tools inventory: approved vs. pending approval, with getting-started links for each approved tool
🏆
AI Excellence Award · Product Engineering · Q3 2025
Recognized across all of Product Engineering (not just UXD) for sustained work getting AI tools approved, adopted, and actually used, including the governance work, the training systems, and measurable org-wide adoption.
AI Excellence Award nomination from Red Hat Recognition Center
Phase 4 · Q3 2025 – Q1 2026

Scale

Figma AI pilot · end-to-end

From pilot proof point
to company-wide access

I ran the Figma AI pilot end-to-end: built executive alignment with a Distinguished Engineer, my VP, and Chief of Staff; negotiated a dedicated sandbox environment directly with Figma; and curated a pilot group of the org's most AI-forward designers. The strongest proof point came from the Ansible Automation Platform team, who used Figma AI on a live, high-priority project and turned around detailed, high-fidelity prototypes of complex automation workflows in a fraction of the usual time, giving engineering a much cleaner handoff.

Ansible Automation Builder prototype built with Figma AI, showing AI Analysis Agent workflow nodes
Ansible Automation Builder: Figma AI prototype of complex multi-agent automation workflows, used on a live production project
VP recommendation document · Q4 2025

A business case built
on our own evidence

I synthesized the pilot findings into a formal recommendation document for my VP: concrete evidence that Figma AI solves a real problem at Red Hat, grounded in our own designers' work. That document shifted the conversation from "should we explore this?" to "our team has already proven this works." I partnered with another UX team outside of Product Engineering to get final sign-off across the company.

VP recommendation document to adopt Figma AI, with Ansible proof point and required actions
VP recommendation: "Adopt Figma AI" with Ansible proof point, pilot findings, and required approval actions
Figma AI documentation · company-wide

A tool rollout is only as good
as what comes with it

Once Figma AI was live, I wrote comprehensive Confluence documentation covering what it can do, how to use it responsibly, how credits work, and how to think about AI-generated outputs within Red Hat's design workflows. The goal was to answer every question a designer might have before they had to ask it.

Figma AI documentation page in UXD Confluence covering responsible use, capabilities, and AI credit usage
Company-wide Figma AI documentation: capabilities, guidelines, responsible use, and credit management
Result
Figma AI turned on for all of Red Hat · Q1 2026
UXD didn't just get access. We drove the process that gave the entire company access.
Reflection

Why it worked

🤝
Empathy over mandates
People adopt what they trust. The experimentation quarter was deliberately low-pressure: try things, share what you learn, no judgment. Participation grew because it felt safe, not required.
⚙️
Systems over moments
Slack channels, tiger teams, surveys, guides, knowledge bases. An operating system for adoption, not a one-off push. That's why it held at 100% rather than spiking and fading.
🛡️
Enterprise navigation as a feature
Instead of fighting security constraints, I did the work to get tools formally approved. Partnering with DPO, Legal, Compliance, and Security created a reusable path that other Red Hat teams now follow.
What Colleagues Say
"

Your leadership in driving the AI initiatives for our team has been nothing short of phenomenal.

Shiran Hirshberg
Senior Interaction Designer · Red Hat
"

Tremendous partner in our efforts on AI enablement this year. This has taken a lot of energy and stamina...

Jeremy Perry
Sr. UX Manager · Red Hat
Next Case Study
Designers Shipping Code →
From design issue to merge request in under 5 minutes, zero code by hand
Hiring for AI, product, or strategy leadership?
I'd rather have a conversation
than send a deck.
Let's talk about what you're building and where I might fit.
📍 Raleigh, NC · Remote friendly
All Work Let's Talk →