Charlotte Truong Portfolio

← Back

TL;DR

Challenge

Solutions

Research

Research Methods

Process

Key Insights

Reflection

END-TO-END DESIGN

LAUNCHED 2026

Envisioning agentic AI & trust with AWS

AI-Powered Feedback Intelligence System

Building trust between players and game studios through the feedback pipeline.

MY ROLE

Lead UX Researcher

Product Designer

TEAM

Charlotte Truong

Angela Shen

Bernice Sun

Yun Cho

AWS PARTNERS

Principal AI Strategist

Head of Design

SKILLS

Service Design

Product Design

User Research

TL;DR

“Hey Ghost”

Every day, millions of players notice something wrong in their game, report it, and hear nothing back. Every day, studios sit under thousands of those reports with no reliable way to tell what actually happened.

Ghost AI is the feedback intelligence system that collects players' reports by voice and turns thousands of scattered reports into a ranked, actionable list. So players are heard, and developers know where to start. No one gets ignored.

FINAL DESIGN

Concept & Features

Ghost AI captures player reports in context, consolidates scattered feedback for studios, and closes the loop when issues are resolved.

Smart Capture

Players say "Hey Ghost" to report an issue without breaking their flow. The system captures the screen, system setup, and timestamp, then transcribes and structures what the player actually said.

Feedback Intelligence

Ghost clusters reports and qualitative feedback from multiple channels into issues and themes, ranks them by impact, and lets studio teams examine trends through the AI assistant.

Impersonation Mode

Instead of interpreting a vague report, a developer can inspect the report game state, including environment, player level, experience, and items, to understand what the player experienced.

Human-controlled action

Teams can move between tasks and feedback views, adjust AI-supported synthesis, and prioritize the work. When a fix or an update is made, the player is notified. Testing made issue routing a requirement for the next iteration.

AI should suggest, not decide

This principle came directly from SME interviews and co-design.

Users wanted AI to reduce manual synthesis and surface meaningful patterns, but warned that overreliance on analytics could erode creative intuition. Usability testing reinforced that prioritization and routing require role-specific context and accountable human judgment. Ghost AI therefore organizes evidence and recommends next steps while people retain control over interpretation, prioritization, and action.

THE CHALLENGE

Feedback isn’t missing. Translation is.

Players report bugs through a dozen channels (Discord, Reddit, X, or the website). The studio manually sorts, triages, and organizes them by priority, with no shared system connecting players' voices to a developer's next move. The gap isn't collection. It's everything that happens after.

The research behind every decision (across players, gaming communities, and studios)

200+ survey responses

across 15+ gaming communities

200+ survey responses

across 15+ communities

6 SME interviews

Indie founders to VP at Triple-A game studio

17 Reddit threads

collected & analyzed

5 co-design sessions

with game developers & designers

5 co-design sessions

with developers & designers

4 usability sessions

Dashboard + in-game

RESEARCH PHASES

An integrated, decision-led research process

I triangulated player behavior, community discourse, and studio operations. Each phase solved a different uncertainty and shaped what we investigated next.

Desk Research & Opportunity Scoping

Main question: Where is AI useful, and where is it merely crowded in the gaming industry?

My team and I mapped out six areas where AI was already reshaping gaming: matchmaking, onboarding, NPC behavior, content generation, anti-cheating, and LiveOps.

LiveOps, the ongoing post-launch management and updates after launch, is the only area with real player pain points and no dedicated AI tooling. This gap is where Ghost AI started.

To understand where the gap lived, I examined how Riot, Xbox, Valve, Bungie, and Pearl Abyss described their feedback and bug-triage processes. The same pipeline kept appearing:

Community intake

↓

Classification

↓

Investigation

↓

Prioritization

↓

Engineering

↓

Player-facing closure

KEY FINDING

The opportunity was not another way to collect feedback. It was the missing translation layer between what a player reports and what a studio can act on next.

Quantitative Player Research

Main question: Where is AI useful, and where is it merely crowded in the gaming industry?

I designed a 10-question survey and distributed it through 15+ gaming communities (Discord servers, LinkedIn, and personal networks). To reduce bias, I placed open-ended questions before structured ones, randomized multi-select answers, and required a single choice for the quit-trigger question.

Among 200+ responses, 57% cited core gameplay as the reason they kept playing. Developer trust was the strongest open-ended theme. Personalization had the strongest relationship with feeling rewarded (r = 0.457, n = 191), and 66% of self-reported churn was actionable according to respondents.

I then interviewed 6 players who had taken the survey, to deepen what the survey could not capture, including four FOMO subtypes and the “safety to leave” retention pattern associated with games such as Warframe.

What keeps players playing, n=206 — core gameplay led at 56.8%

Main reason players stopped, n=206 — most cited reasons were actionable by the studio

KEY FINDING

The opportunity was larger than issue collection. Players' continued engagement was tied to whether they felt the studio listened and responded to their feedback.

Semi-structured SME interviews

Main question: Why do studios struggle to act on feedback they already have?

I planned and conducted six interviews across studio sizes and roles. Each guide was customized to the participant and iterated on prior sessions, so later interviews could probe contradictions and gaps rather than repeat a fixed script.

Indie designer/producer on a 2-person team
Business/product lead at a 3-person studio
Systems designer on a major publisher’s MMO
Tools engineer with sandbox and mobile experience
Lead producer at a VR studio using AI
VP of Product for an 8-year-old live-service title

I recorded, transcribed, and coded interviews around feedback collection, business-versus-design tension, trust, tooling gaps, and attitudes toward AI, allowing the team to compare a 2-person studio with a publisher serving 20M+ monthly players.

Indie designer/producer on a 2-person team
Business/product lead at a 3-person studio
Systems designer on a major publisher’s MMO
Tools engineer with sandbox and mobile experience
Lead producer at a VR studio using AI
VP of Product for an 8-year-old live-service title

CONVERGENCE

Every single participant, independently and without prompting, described some version of the same structural gap: feedback is not hard to collect, but translating it into prioritized, cross-functional action is almost entirely manual.

This convergence across radically different studio scales was the strongest evidence signal of the entire research phase.

Synthesis and Affinity Mapping

Main question: Which patterns hold across players, communities, and studios?

I combined survey results, Reddit findings, and all six interview transcripts in one affinity map. This made it possible to examine where evidence converged, where it conflicted, and which needs were consequential enough to shape the product thesis.

Affinity map

KEY FINDINGS

Feedback is abundant, but the pipeline is fragmented.

The real bottleneck is manual monitoring, grooming, and interpretation—not collection.

Feedback only becomes actionable when paired with context.

Prioritization is where feedback becomes political and cross-functional.

Player trust is simultaneously the outcome and the risk of the entire feedback process.

THESIS

Connect fragmented player signals to context-rich, cross-functional studio action, while keeping player trust central to the system.

Game player

Player reporting journey

Game studio

Studio triage journey

Co-Design Sessions

Main question: How much can analytical depth help teams without displacing creative judgment?

Before finalizing a design direction, I facilitated structured co-design sessions with some SME participants, walking them through the envisioned experience, including an in-game report prototype, then giving them unstructured time to explore a studio-side dashboard based on their own experience.

The participant cautioned that over-reliance on analytics dashboards could become a "slippery slope" that erodes creative intuition.

The participant cautioned that over-reliance on analytics dashboards could become a “slippery slope” that erodes creative intuition.

DESIGN CHANGE

Use a simplified default view with progressive disclosure into deeper data rather than presenting a dense dashboard all at once.

Moderated Usability Testing

Main question: Can players easily report in-game, and can studio teams turn the result into action?

With a working prototype, I planned and moderated (individual and joint dual-participant sessions), walking participants through both halves of the product: the in-game voice-activated report flow and the studio-side dashboard, including a novel “impersonation mode” letting developers step directly into a player’s reported game state. Sessions used think-aloud protocol throughout, plus targeted scenario tasks (e.g., “200 players reported a similar bug over the weekend—walk me through how you’d make sense of it”). ( individual and joint dual-participant sessions), walking participants through both halves of the product: the in-game voice-activated report flow and the studio-side dashboard, including a novel “impersonation mode” letting developers step directly into a player’s reported game state. Sessions used think-aloud protocol throughout, plus targeted scenario tasks (e.g., “200 players reported a similar bug over the weekend—walk me through how you’d make sense of it”).

Findings meaningfully changed the design rather than merely validating it. Multiple participants, independently, identified that a single unified dashboard was trying to serve incompatible roles.

With a working prototype, I planned and moderated (individual and joint dual-participant sessions), walking participants through both halves of the product: the in-game voice-activated report flow and the studio-side dashboard, including a novel “impersonation mode” letting developers step directly into a player’s reported game state. Sessions used think-aloud protocol throughout, plus targeted scenario tasks (e.g., “200 players reported a similar bug over the weekend—walk me through how you’d make sense of it”). (one individual, one joint dual-participant session), walking participants through both halves of the product: the in-game voice-activated report flow and the studio-side dashboard, including a novel “impersonation mode” letting developers step directly into a player’s reported game state. Sessions used think-aloud protocol throughout, plus targeted scenario tasks (e.g., “200 players reported a similar bug over the weekend—walk me through how you’d make sense of it”).

Findings meaningfully changed the design rather than merely validating it. Multiple participants, independently, identified that a single unified dashboard was trying to serve incompatible roles.

" … very few companies have the same person fixing bugs and investigating data.” - One participant opinion during usability testing

The same participant’s single sharpest, most emotionally charged pain point — bugs routinely assigned to him that he couldn’t fix, with no system to route issues to the correct owner — became a headline feature requirement for the next design iteration.

DESIGN CHANGE

Design for role-aware workflows and issue routing. Across all sessions, reinforce one non-negotiable constraint: AI may surface and suggest, but it must never decide or act autonomously.

OUTCOME

The Outcome: a Defensive Direction, not just a Polished Concept

The work followed a clear line from evidence → decision → product response, moving the team from an unscoped prompt to a tested product direction supported by quantitative, qualitative, participatory, and evaluative evidence.

Research gave the team a foundation to build, a model for what to build, and constraints for how AI should behave.

REFLECTION

What I carried forward

The most important methodological decision in this project was triangulation across radically different vantage points—a 2-person indie team, a mid-size mobile studio, and a publisher managing millions of monthly players—rather than optimizing for depth within a single studio profile. It was the convergence across that range, not any single strong data point, that gave our design direction its credibility.

The usability testing phase reinforced a lesson I try to carry into every project: the sharpest, most valuable insights are rarely the ones you asked a direct question to get. Our single strongest feature requirement—bug routing—emerged unprompted, mid-conversation, from a participant working through a scenario task. It was a reminder that think-aloud protocol and scenario-based testing consistently outperform direct questioning when the goal is surfacing insights participants do not yet know how to articulate as feedback.

CLOSING THOUGHT

Software should not become a barrier between people. It can help build connection and trust.

CHARLOTTE TRUONG

WORK

PLAY

ABOUT

Resume

Charlotte Truong

WORK

PLAY

ABOUT

Resume

Charlotte Truong

WORK

PLAY

ABOUT

Resume