Overview

I took a design team that was using AI only superficially — occasional chats with LLM tools, no real integration — and led its transformation into a deliberate, AI-native practice. Working without formal ownership but with clear initiative, I ran the discovery, built the strategy, proved the concept on a live feature, and produced the adoption plan now guiding the team's direction. The core insight I established: with AI, output quality is capped by context and system legibility, not by the tool — so the real work is foundational.

The Challenge

The team faced pressure from two directions: its own appetite to work in more modern ways, and leadership's push to adopt AI meaningfully and soon. But adoption was shallow and unstructured, and the underlying practice had real friction: technical constraints surfaced late in the design process causing rework; a design system whose components were scattered and whose composability wasn't well understood; slow, iterative ticket-writing and heavy prototyping effort; and no consolidated source of truth for a product actively being built.

The risk: rushing into AI tooling on top of shaky foundations would just automate the mess faster.

My Role

I was the de facto driver of the initiative — setting direction, facilitating, building the artifacts, and navigating stakeholders — while keeping every decision validated with the wider design team rather than imposed. I operated at both altitudes: strategy (the adoption plan, the governance model, the leadership case) and hands-on craft (recreating the design system in the AI tool, running the proof-of-concept).

Approach

1. Diagnose before adopting. I facilitated a full-day team workshop that mapped the real bottlenecks in our process and separated the actual problems from the hype — establishing that foundations, not tools, were the constraint.

2. Prove it on something real. Using our AI design tooling (the Claude ecosystem — chat, design, and code), I recreated our design system inside the tool and ran a proof-of-concept on an already-designed product feature. Even with deliberately minimal context, the AI generated UI matching the content of our shipped designs — a clear signal of the ceiling being context, not capability.

3. Sequence the work honestly. I authored a phased adoption plan built on a "foundations first" principle: secure capacity, make the design system legible to both people and AI, consolidate the product's scattered knowledge into a single source of truth, then layer on skills, workflows, and agents — iteratively, validating continuously, never overbuilding.

4. Protect quality with governance. I established a firm principle that only designers decide what ships to the front end, with non-designer AI output treated as ideation — allowing the team to safely extend AI use to other roles without eroding design quality.

What I Did

Facilitated a full-day cross-team workshop diagnosing process bottlenecks and aligning on direction. Recreated the company's design system inside the AI design tool for hands-on testing. Designed and ran a proof-of-concept translating real product requirements into UI. Authored a phased, validated AI adoption plan spanning capacity, foundations, first skills, and future workflows/agents. Produced leadership-facing materials — a rationale document and a presentation — making the strategic case for investment. Defined a governance model positioning designers as the quality gate for AI-assisted output.

Outcomes

A proof-of-concept that reframed the conversation: a design task that had originally taken about a week of careful design work was reproduced — matching our real designs' content — in minutes from minimal context, with a credible path to a polished, shippable result in roughly an hour once proper context is in place.

A validated strategy, not just enthusiasm: the adoption plan became the team's agreed direction and the basis of the case put to leadership.

A durable governance model that lets AI use expand across roles while keeping design quality protected.

A shift in team posture — from scattered, shallow AI use toward deliberate, foundations-first adoption.

Reflection

The temptation with AI is to chase the flashy end — agents, automation — immediately. My contribution was insisting on the unglamorous truth: the leverage is in making our design knowledge legible, to humans and machines alike. Getting the team and leadership to see foundations as the thing that unlocks speed, rather than as a delay before it, was the real design problem — and the one I'm most proud of solving.

"The ceiling isn't the tool — it's the context we give it. So the real work is foundational."

Capabilities: Design strategy · AI-assisted design · Design systems · Design operations · Workshop facilitation · Change management · Stakeholder alignment · Proof-of-concept & experimentation · Leadership communication · Governance design