Designing Jessica — an AI teammate bringing speed, clarity, and control to content management
Done
In-progress
ROLE
Product Designer
MY DIRECT CONTRIBUTION
Research, UX strategy, prototyping (Cursor + Lovable), component library design, engineering handoff.
TIMELINE
1 month (end-to-end)
TOOLS
OVERVIEW
CommerceIQ is an e-commerce automation platform that helps brands manage operations and digital shelves across major retailers. We work closely with our users - observing workflows and testing quick prototypes to understand what truly helps them.
During one such cycle, we uncovered a major gap in Newell’s content workflow.
Newell manages 10,000+ SKUs across Amazon, Walmart, and Target. Their content team must keep:
PIM content complete
Retailer PDPs aligned
Titles, bullets, and keywords compliant
High-impact SKUs prioritized
Updates reviewed and published accurately
But their process relied on manual checks, multiple tools, and spreadsheets - slow, error-prone, and hard to scale.
Understanding the Content Manager’s Workflow
Through shadowing and interviews, I mapped the workflow of content managers
Okay, then what’s the problem ???
manual content compliance checks are time-consuming & error-prone
no clear overview of “overall/ brand-level content health”
manual content-creation is a repetitive, inefficient & boring
constant context switch between tools slows down progress
as a content manager I want to...
see content health across all SKUs in one place
know exactly what’s wrong without digging
know which SKUs to fix first
track if updates got pushed to retailer
get AI-powered content suggestions
update PIM/PDPs without switching tools
Understanding the Content Manager’s Workflow
Through shadowing and interviews, I mapped the workflow of content managers
Key research findings
that shaped the entire product direction:
Humans want control, not auto-publish
Content managers resisted the idea of a fully autonomous co-pilot.
Explanations matter
Users needed clarity on: what’s wrong , why it’s wrong, how to fix it
Issue surfacing must be fast
Users wanted a prioritized list of issues before diving into SKU-level details.
Workflows must scale
Any flow must work for 10,000 SKUs, not 10.
Design Principles
I translated insights into 5 guiding principles:
Human-in-control
AI suggests; humans approve.
Explain everything
never ask users to “just trust the AI.”
Reusability
components and interaction patterns must serve other agents.
One home for content health
Eliminate tab-hopping.
Speed > complexity
show issues upfront; drill into details only when needed.
Design Constraints
Designing Jessica wasn’t a blank canvas. I had to work within:
1/ Existing Framework
We leveraged the existing AgentSpace framework, a workspace built to host multiple AI agents for E-commerce Brands - my design had to feel native, consistent, and reusable.
3/ AI behaviour boundaries
We needed to design for transparency, explainability, and human approval.
These constraints became design opportunities, pushing the system toward scalability instead of one-off UI.
Early Exploration with Lovable
I used quick Lovable prototypes to test the core workflow and understand what level of automation users were comfortable with.
To validate this, we designed two prototype paths:
Assisted Mode — Jessica flags issues and suggests fixes
Co-pilot Mode — Jessica auto-fixes and syncs content
Even when co-pilot mode was made seamless, users consistently chose the assisted path. Because content impacts revenue and SEO, they wanted speed with control-not silent automation.
🌻
key design decision:
Content agent won’t auto-fix content. prioritise - while users stay fully in control.
Building Within the AgentSpace Framework
We leveraged the existing AgentSpace framework—a workspace built to host multiple AI agents for E-commerce Brands.
AgentSpace already had the core interaction model & architecture defined.
👩🏻💻 My Role
Shaped the Content Agent within AgentSpace - defining workflows, key adaptations, and system behaviour while driving stakeholder alignment.
🤓 Focus
Ensure the content agent fits naturally into the content team’s day-to-day workflow.”
Agents → Skills
Agents → Skills → Agent’s Tasks
Agents → Skills → Agent’s Tasks → Output
Left Pane
Agent & Task Details
Center Pane
Reports → Widgets
Right Pane
Chat with agent
Agents → Skills → Agent’s Tasks → Output → My Task
Left Pane
Agent & Task Details
Center Pane
Reports → Widgets
Right Pane
Chat with agent
10 skus has title issues
12 skus has image issues
.....
review and take action
Final Solution- Meet Jessica
Jessica is an AI agent that validates content, identifies mismatches, prioritises issues, and suggests fixes - all while keeping the human in charge.
SKU
Yankee Candle... & 2
Action Status
To-Do & 1
Overall Validation Score
>=
70%
SKU
Yankee Candle... & 2
Search Value
3 Selected
Select all
Clear all
BL-002 Ocean Mist Soy Candle
BL-003 Amber Nights Aromatherapy Candle
BL-004 Forest Whisper Cedarwood Candle
BL-010 Frosted Pine Winter Candle
BL-001 Serenity Vanilla Candle
BL-007 Cozy Hearth Cinnamon Candle
BL-005 Citrus Bloom Hand-Poured Candle
BL-006 Midnight Rose Luxury Candle
BL-008 Tranquil Lavender Spa Candle
BL-009 Golden Sandalwood Essence Candle
1 filter added
Cancel
Apply Filter
Action Status
To-Do & 1
Search Value
Value
Select all
Clear all
To-Do
In Progress
Done
Saved For Later
1 filter added
Cancel
Apply Filter
Overall Validation Score
>=
70%
Enter Value
>=
70
%
1 filter added
Cancel
Apply 2 Filter
You
3:27pm
I’m trying to diagnose sales changes that happened in Q3. Focus on market trends and competitive factors
Categories:
Dog Food
Time:
01 Apr 2025 → Jun 30, 2025
Compared to:
01 Apr 2024 → Jun 30, 2024
Frequency:
Just once
You
3:27pm
I’m trying to diagnose sales changes that happened in Q3. Focus on market trends and competitive factors
You
3:27pm
SKUs with multiple content issues
Did conversation rate coincide with the drop in availability?
Blake
Hi, I'm Blake, your AI teammate for sales performance insights. I can help you dig into your numbers and uncover what's really driving your results. What would you like to explore?
Blake
Thought for 3 minutes...
Here’s what I found...
Tempus leo eu aenean sed diam urna tempor. Pulvinar vivamus fringilla lacus nec metus bibendum egestas. Iaculis massa nisl malesuada lacinia integer nunc posuere. Ut hendrerit semper vel class aptent taciti sociosqu.
Blake
Thought for 3 minutes...
Here’s what I found...
Tempus leo eu aenean sed diam urna tempor. Pulvinar vivamus fringilla lacus nec metus bibendum egestas. Iaculis massa nisl malesuada lacinia integer nunc posuere. Ut hendrerit semper vel class aptent taciti sociosqu.
Blake
Thought for 3 minutes...
Here’s what I found...
Tempus leo eu aenean sed diam urna tempor. Pulvinar vivamus fringilla lacus nec metus bibendum egestas. Iaculis massa nisl malesuada lacinia integer nunc posuere. Ut hendrerit semper vel class aptent taciti sociosqu.
We’re sorry you didn’t find this helpful! Please let us know why?
Inaccurate
Unclear
Not actionable
What could have made this more helpful?
Submit
Skip
Blake
Thought for 3 minutes...
Hi, I'm Blake, your AI teammate for sales performance insights. I can help you dig into your numbers and uncover what's really driving your results. What would you like to explore?
Added SKUs with highest unavailability
Blake
Memory updated today at 3:27pm
Blake
Thinking...
Clarify the scope
Segment Sales Δ into Units Δ and Price Δ
Investigate Market Trends Impacting Traffic
Assess Competitive Organic Presence
Examine Competitive Paid Activity
Check Conversion Drivers Influenced by Competition
Incorporate Market-Level Events
Synthesize Findings
Blake
Working on your request...
Thinking...
Clarify the scope
Gather PIM Content Data
Collect Retailer PDP Data
Match PIM and Retailer Content
Detect Content Gaps and Mismatches
Evaluate Content Quality
Generate Recommendations
Summarize and Prepare Report
Design System Contributions
I was also shaping our new Tailwind-based component library, so Jessica’s design had to scale. Every layout, table, and action pattern I created was built to be reusable across future agents and support large, complex SKU workflows.
The Effort Behind the Outcome
This project came with tough timelines and plenty of all-nighters. It pushed me outside my comfort zone, but also accelerated my growth. Looking back, the progress I made and the craft I developed give me genuine satisfaction and confidence in my abilities.
