Building an AI-assisted design workflow - from Ad-hoc tool use to a team-wide practice
Cut rework by ~60% and shortened turnaround time, while keeping output relevant and client-ready
Role
Duration
Tools
Team
Project Overview
AI-Powered Work shows me building the structure — workflow, guidelines, training — so my whole team could grow into AI-assisted design together, rather than a few people experimenting alone.
20+
Assisted Templates
40%
Improved workflow
2+
custom build platform
Problem Statement
The design team was losing significant time to research and UX exploration — and would often hit creative blocks partway through a project. There was no structured way to use AI to unblock this; individual designers experimented on their own with whatever tools were available (Figma AI, Gemini, and other generative AI tools approved for use). Without a shared approach, the team wasn't consistently evaluating whether an AI-generated idea was actually efficient to build, or whether it genuinely worked for the end user — so AI use was inconsistent and its value was hard to prove.
What I Led
Built a structured AI-assisted workflow covering ideation, drafting, research synthesis, and final conclusions
Wrote a set of guidelines defining how the team should use AI at each stage of a project
Trained the team on prompt engineering — how to structure a prompt to get a usable, relevant output rather than a generic one
Set the evaluation standard: every AI-generated idea was checked for build efficiency and end-user relevance before it moved forward, not accepted at face value
Made sure the team's skills grew in step with the market, building the workflow so people could grow into AI-assisted design together rather than adopting it individually and unevenly
What I Designed / Built
Designed a complete checklist and template system, built using AI, that any designer could open to generate initial concept variations and ideas from a starting brief
Applied this directly on client projects — when a client asked for a quick update, used AI to generate multiple sample directions, then compared them against each other and against the brief to select what actually worked
Used AI to speed up research synthesis, turning raw findings into themes faster than manual synthesis alone
Key Decisions & Trade Offs
Where AI stops and judgment starts: used AI for volume and speed in ideation, drafting, and early concept variations — but every output was evaluated by the team for feasibility and end-user relevance before it moved into real design work, keeping final craft and decisions human-led.
Tool restriction as a constraint, not a blocker: built the workflow around only the organization-approved tools (Figma AI, Gemini, specific generative AI access), rather than waiting for broader tool access, so the practice could roll out immediately.
Standardizing through templates over ad-hoc use: invested time building a reusable checklist/template system instead of letting each designer use AI their own way, trading some individual flexibility for consistency and faster onboarding of the practice.
Training on prompt engineering, not just tool access: treated prompting as a skill to teach explicitly, since giving the team access to AI tools without guidance was producing inconsistent, less useful output.
Final Design
Impact
~60% reduction in rework, since AI-assisted evaluation caught infeasible or off-target ideas earlier in the process
Faster turnaround time on both internal ideation and client-facing quick-turn requests
Consistent relevance: output stayed targeted and usable rather than generic, because every AI-generated direction was evaluated against feasibility and end-user fit before being used
Leadership Moment
I made it a priority that my team wasn't left behind as AI changed how design work gets done and not by mandating tool use, but by building an actual structure (workflow, guidelines, training) so the whole team could grow into it together, at the same pace, rather than a few people experimenting alone while others fell behind.
