Case Study 03
AI Across the Entire Delivery Cycle
Enterprise AI Operating Model · Lowe's
A 30-person UX organization was absorbing a structural failure the company kept reading as a design problem. I rebuilt how work entered the system — and turned volatile delivery into a measurable, durable operation.
Organization
Lowe's Companies, Inc.
Role
Senior UX Manager
Scope
3 disciplines
Adoption
Under 60 days
The Mandate
A mandate without a model is pressure without direction.
Leadership issued a company-wide AI adoption mandate to stay competitive. Tools were provided but no playbook, no use cases, and no implementation direction came with them.
The real exposure wasn’t that the team wouldn’t use AI. It’s that they already were informally, inconsistently, and without any review. Leadership issued the mandate. I built the model.
No playbook
Tools with no direction
Ungoverned
Informal use already started
No standards
Prompt quality varied widely
No gates
Unvalidated outputs in the cycle
Skeptical
Three disciplines, low trust
Exposed
Risk without a measurable return
The Diagnosis
Governance first. Speed second.
I built the governance before I built the workflows. Ungoverned AI at enterprise scale isn’t innovation, it’s risk without a return. One validated path ran across all three disciplines, with a mandatory human review gate no output could skip.
The Validated Path
Two competing explanations surfaced early. Some believed it was design capability; others believed it was insufficient Product engagement. The data showed symptoms of both but neither explained the pattern.
The Intervention
3 structural moves.
Research
From trapped insight to a living repository
Content
A custom GPT trained on brand voice
Design
Sketch-to-wireframe, with a human gate
Before: sketches manually recreated as wireframes. Discovery cycles of 3-5 days. 2-3 alignment rounds before sign-off.
After: a sketch-to-wireframe pipeline. Iteration compressed to ~2 days. Alignment rounds down to one. No final UI without human review.
The Result
More rigorous work, in less time, with less rework.
Research Report turnaround
5–7 days → 2–3 days
Sprint commitment accuracy
3–5 days → ~2 days
2–3→ 1
— → <60 days
Source: research workflow tracking; sprint analytics; alignment session logs · Lowe’s, 2022–2024
Leadership in Action
Building the bridge between speed and integrity.
Built structure before anyone asked for it.
The mandate came with tools and no direction. I designed the governance model first, then built the workflows around it.
Created use cases where none existed.
The team was skeptical because they had no examples. I built the review checkpoints, and repository standards then showed them what responsible AI looked like in practice.
Made adoption safe, so it became easy.
Three resistant disciplines became structured daily adopters within 60 days. The result wasn't compliance, it was capability.