Marketing Content Generator

Redesigned a fragmented AI content platform from research to shipped product — replacing a cluttered one-page setup with a modular, context-aware experience that cut campaign setup time by 40%.

Role
Lead UX / Product Designer
Employer
American Express
Tools
Figma · Framer · Miro
Focus
AI · Design Systems · UX
40%
Reduction in campaign setup time
80%
Task success rate — nearly double prior performance
60%
Of users' real workflows unaccounted for in v1
01 · Context

Joining mid-flight — with a mandate to reimagine

When I joined the MCG project at Amex, my first task was to design the UI for early conceptual ideas and wireframes. I used the Amex design system as a foundation and created additional components and patterns tailored to the app's needs — a decision that proved vital as the product evolved well beyond its original scope.

The original concept was overloaded with settings and options. My goal wasn't just to apply a visual layer — it was to reimagine the experience from the ground up, using research as the primary input rather than assumptions baked into the original brief.

Research → design → ship Amex Design System Component extension Generative AI product Multi-channel scalability
Original wireframes from requirements
F.1 — Early wireframes built from the original requirements, before user research revealed critical workflow gaps
High fidelity mockup of original concept
F.2 — High-fidelity mockup of the original concept: settings-heavy, sidebar-first layout that testing later invalidated
02 · Research & Evidence

Listening first — then designing

I began by identifying core pain points through stakeholder interviews and moderated user testing. Feedback revealed critical issues: excessive scrolling, misaligned settings, and a layout that didn't reflect how 60% of users actually worked. Rather than iterating on the existing design, the evidence called for a structural rethink.

Live user testing session
F.3 — Live user testing with stakeholders present, synthesizing behavioral patterns in real time
Behavioral personas derived from research
CM
Campaign Manager
Launches 8–12 campaigns/month. Values speed and predictability. Skips settings they don't immediately recognize.
Pain: Too many irrelevant options upfront
CA
Content Author
Focuses on copy quality and brand tone. Uses AI generation as a starting point, not a final answer.
Pain: Settings interrupt the creative flow
Persona documentation
F.4 — Behavioral personas grounded in user data, used to prioritize which pain points to address first
Finding Evidence Design decision
60% of users had different workflows Interview patterns across 3 user segments Role-aware, modular setup flow vs. one-size layout
Excessive scrolling to reach key settings Task completion heatmaps, session recordings Embedded, context-aware controls — no sidebar
Settings order didn't match mental model Think-aloud sessions; confusion peaked at step 2 Step-by-step setup with channel-conditional branching
Low trust in AI output on first use User sentiment interviews, fear replacement by AI Inline AI with transparent reasoning + edit-in-place
03 · Design Decisions

From sidebar to embedded — the core UX shift

The biggest structural decision was removing the sidebar-based settings panel entirely and embedding controls contextually within the content creation flow. This wasn't aesthetic — it came directly from the research finding that users ignored or misread settings they had to navigate away from the canvas to adjust.

I introduced a modular, step-by-step setup flow that guides users through key decisions without overwhelming them. Each step is purpose-built, with options surfacing conditionally based on selected channel — so a social campaign doesn't surface email-specific settings.

Layout evolution: sidebar model → embedded model
Before — sidebar model
Mockup example
Settings panel detached from the content canvas
After — embedded model
Mockup example
Controls surface inline with the work in progress
Modular setup flow — conditional steps based on channel selection
1 · Channel
2 · Audience
3 · Tone
4 · Generate
5 · Review
Steps 2–4 surface different options depending on selected channel — reducing irrelevant settings per user type
Step by step setup flow
F.5 — The modular setup flow: each step purpose-built, conditional on channel selection, reducing cognitive load without sacrificing flexibility
Role-aware UI design
F.6 — Role-aware UI balancing competing user needs within a scalable, component-driven architecture

04 · AI as a Design Surface

Generative AI embedded in the workflow — not bolted on

The MCG product used generative AI as its core output engine. My design challenge was making that feel trustworthy and steerable — not like a black box. I designed the AI interaction layer so users could guide generation, review reasoning, and edit outputs inline without ever leaving the creation canvas.

This required inventing new design patterns: streaming output states, confidence indicators, regeneration controls with preserved context, and an edit-in-place model that respected the AI's role while keeping the human clearly in control.

Prompt engineering in UI Inline generation patterns Edit-in-place Streaming output states Trust through transparency
Role-aware UI design
"The AI feels like a collaborator, not a slot machine. I can actually steer it now." — User testing participant, post-redesign

05 · Design Systems Thinking

Extending Amex DS — components built to scale

I used the Amex design system as the foundation and extended it with new patterns specific to AI content workflows: a generation panel component, channel-conditional form controls, and an admin role management surface. Each component was architected to accommodate future channel additions without requiring layout rework.

Delivered structured design annotations that clarified component states, interaction logic, and handoff specifications — significantly reducing developer Q&A cycles and keeping the team aligned across a fast-moving build.

Amex Design System Component extension Token architecture Dev handoff annotations Multi-channel scalability
Design annotations
Delivered annotations clarifying component states, design decisions, and handoff specifications — streamlining the development workflow

06 · Project Dashboard & Management

Making past work findable, shareable, and manageable

Beyond the creation flow, I designed a project dashboard aligned to real user mental models — making it easy to locate past campaigns, manage bookmarks and favorites, and support key collaboration tasks like sharing and role-based access.

I also designed a dedicated admin settings surface that kept role management separate from the creation flow — giving admins the control they needed without adding complexity for everyday users.

Project dashboard
F.7 — Dashboard structure mapped to user mental models: browse, favorite, share, and manage campaigns from a single surface
Admin settings
F.8 — Admin role management: role creation, user assignment, and permission scoping — a dedicated surface that doesn't leak complexity into the creation flow

07 · Post-Launch Impact

Measurable results from research-backed decisions

Post-launch results showed the redesigned experience not only improved usability but delivered measurable impact across every tracked dimension. The modular setup flow, role-aware UI, and embedded AI interaction model each contributed to a fundamentally faster and more confident user experience.

40%
Faster campaign setup — modular flow vs. cluttered one-page design
80%
Task success rate in usability testing, nearly doubling prior performance
Improvement vs. prior baseline across key measured tasks
Users described the experience as "streamlined," "intuitive," and "much more manageable" — unprompted, post-launch.