Walmart

2025-26

AI Task Prioritization tool

Role

Senior UX Designer

Timeline

Aug 2025 to Jan 2026

Team

2 product managers, 7 engineers, a design researcher, 3 design peers, and multiple stakeholders

Deliverables

Discovery interviews · Cross-team workshop · Wireframes and concepts · Task card and workspace framework · Concept testing (contributor)

Overview

Walmart merchants juggle hundreds of tasks and emails a day, and AI agents were about to add more. I was asked to redesign a small task widget that almost no one used. I turned it into a framework for how merchants see, prioritize, and act on all their work suggested by AI.

Impact

One shared language

for tasks, messages, and AI suggestions across teams

Concept testing

across Walmart US, Sam's Club, and Canada, whose feedback shaped the final design

Pilot use case

resolving items stores couldn't scan, with AI recommending a fix and merchants deciding

Adopted as the design framework

for how merchants review and act on AI-generated work

Problem

Merchants had too many tasks everyday and no way to know what mattered, or why

Prioritization was manual

Merchants set priorities with leadership every day and kept their own task lists.

A widget no one used

It showed only task titles and links, so merchants turned to email for context.

More work coming from AI

AI agents were about to send far more tasks, with no consistent way to review them.

Solution

01: AI sets the priority. The merchant makes the call.

Every task shows its impact and why AI ranked it there, all in one focused view. Merchants review, then accept, reject, or decide in bulk.

I built this concept final concept in Claude Code, as I exited before pilot launch

02: Verification, Allowing users to confirm AI decisions

Every task shows its impact and why AI ranked it there, all in one focused view. Merchants review, then accept, reject, or decide in bulk.

AI Principles

Impact first: value

Every task shows what it's worth and how long it takes.

AI is always marked: trust

Merchants know what came from AI.

Every decision teaches AI: learning

Accept or reject, and the system learns.

03: Supplier platform adoption

I adapted the same framework for suppliers, so they could act on their own tasks the same way.

The framework holds beyond one use case, Dynamic panel interface

Research

01: The widget had no real job

Two merchant interviews in week one showed they needed context, not a list of titles. That changed the brief from "make it better" to "make it worth using.

"I need to see the big picture of what needs to be done today and this week, and I want to know about the fires before they happen."

Interview participant

Merchant

02: Every team defined a "task" differently

Ran a workshop with designers from three teams to agree on one shared set of definitions.

Ideation

From a widget to a workspace

I explored several directions. Each round of feedback brought a new insight, and the design kept changing. It stopped being about a better card and became about how merchants feel doing the work they do every day.

This was their bread and butter, and their most frustrating part. So the goal became understanding what mattered most to them, and showing only that.

Each round cut more noise and surfaced what mattered.

The decisions that shaped it

One list, not tabs

Tabs hid AI suggestions behind a click, so I merged all work into one list ranked by impact. Trade-off: the ranking has to be right, every time.

One card, many items.

Merchants deal with volume, so a card can hold one item or fifty, decided in bulk. Trade-off: bulk is faster, but easier to approve without reading.

Only what drives the decision.

Impact and estimated time stayed. Due dates and extra icons went.

AI suggests, the merchant decides.

Every recommendation shows its impact and reason, and the merchant accepts or rejects it. Each response helps the AI improve.

Reflections

The hardest part wasn’t the screens.

New ideas get mixed reactions.

It took trial, error, and many perspectives to find what worked.

Alignment is design work too.

The workshop made no screens, but every team built on it.

People trust AI when they see why

The reason behind a suggestion matters as much as the suggestion.

More case studies

Fewer charges, faster to shelf

On the go tranportation notifcations

Finding evidence made faster by 80% with AI

© 2026 Akanksha Kulkarni Portfolio