Planit: An AI-Assisted Design Workflow for Target

Role

Senior UX Designer

Client

Target

Team

3 designers, 2 PMs, 6 engineers

AI tools

Copilot, Figma Make

4 min read

Planit Cycle Segmentation screen showing assortment variations grouped by segment

At a glance

Problem

Three designers had to keep six engineers moving on dense merchandising data that broke static mockups.

Key decision

Combined segments and footages into one drawer, because footage depended on the segment chosen.

Result

An AI-assisted workflow that delivered build-ready screens mid-sprint, and a data-driven prototype the team kept using.

Overview

Planit is Target’s merchandise planning application, where planners build and adjust assortment plans. The hard part wasn’t any single screen. It was the data behind them: dense, deeply nested, and different in every business unit.

My role was support, not direction. But once the lead designer’s early features reached me, making them real was on me: defining every error state, accounting for edge cases, refactoring designs as real data exposed problems, and seeing each feature through build. To keep pace with six engineers on a three-person design team, I built an AI-assisted workflow, and the prototype at its center stayed with the team for the features that followed.

I joined as Senior UX Designer to take the early feature designs and make them ready to build. That meant designing for the messy cases, testing against real data, and getting engineering screens they could build from with confidence. AI tools were central to how I did it, speeding up research synthesis and prototyping so a small design team could keep pace with six engineers.

My role

One of three designers, alongside 2 product managers and 6 engineers; partnered with the lead designer to scope each feature
Defined every error state and edge case, and refactored designs as real data exposed problems
Built the AI-assisted workflow: Copilot research synthesis and a data-driven Figma Make prototype
Delivered annotated, development-ready screens and wrote acceptance criteria for each Jira story

The risk: design becoming the bottleneck

Complex data

Merchandising data is dense and deeply nested, so layouts that looked clean with sample text broke with real values.

No two business units alike

Each unit had its own data shapes, volumes and conventions, so a design that worked for one could fail for another.

Static mockups hid the problems

Hand-typed placeholder data looked tidy and gave engineers a false picture of what they’d actually be building.

The workflow I built

I set up a three-step workflow so research, design and handoff all ran on real data, with AI handling the repetitive work.

Synthesize with Copilot. I used Copilot to work through user feedback transcripts, pulling out recurring themes, pain points and requests. What used to take days of tagging notes became a first-pass synthesis I could review, challenge and refine.
Prototype on real data in Figma Make. I built the data-driven prototype in Figma Make. Instead of hand-building every variation, I could generate working screens and swap in different business units’ datasets, so each design was tested against real data shapes, not placeholder text.
Hand off what engineers will actually build. Engineers got annotated screens filled with realistic data, plus acceptance criteria I wrote for each Jira story once stakeholders signed off. Every story had a definition of done that included error states and edge cases.

Keeping judgment in the loop

AI output was a first draft. I checked every Copilot theme against the source transcripts before it shaped design priorities.
Every generated screen got a review. I reviewed each Figma Make screen against the design system before it went to engineering.
I owned the edge cases. I defined every error state, empty state and edge case the core flows needed, tested them against data from different business units, and refactored designs whenever real data exposed a problem.

Key decision: one drawer for segments and footages

Segments and footages started out as separate screens. But in the data, footage depended on the segment chosen, so splitting them meant planners had to pick a segment on one screen and set its footage on another. I made the call to combine them into a single combo drawer, so planners could handle segments and footages in one place. It’s the kind of call that only surfaces when you design against real data instead of placeholder text.

The finished Cycle Segmentation prototype in Figma Make (opens in a new tab)

Open the prototype in Figma Make (opens in a new tab) ↗ Password: planit26

Outcome

The workflow did what it was built to do: it kept a three-person design team ahead of six engineers instead of becoming the bottleneck.

Faster screen delivery. Screens were delivered mid sprint, with ample time for review and revision.
Fewer surprises in build. Designs were tested against real business-unit data before handoff.
A reusable approach. The data-driven prototype stayed with the team for the features that followed.

Planit launches after my time on the project, so these outcomes are about the design process rather than post-launch metrics.

Next project

Trane Total Design Assist (TDA) →