MAL
Company
LG Electronics
Duration
AUG 2023 - JUL2024
Cutting revision cycles by 50% with an AI SaaS verification tool, enabling faster and accurate global workflow.
Overview
LG’s global teams create thousands of contents each year across appliances, apps, and digital products. Before it goes live, it must be verified for accuracy, grammar error, consistency, and brand tone — a process that inconsistent, and heavily manual.
I designed MAL (Machine-Assisted Language), an internal AI-driven verification platform, to make that process scalable. But this wasn’t just a design problem — it was an organizational challenge. I had to navigate legacy workflows, pushback from global teams resistant to change, and massive data audit.
My role was to create the systems — tone, rules, and workflows — that AI could thrive in, while winning trust from stakeholders across the company.
Role
Product Designer
Platform
Internal SaaS tool
Scope
Product Design,
AI Plaform Design, Stakeholder Alignment, Data Analysis, UX Writing
Team
PM, 3 Engineers, Technical Writer
Background
LG’s global teams produce massive amounts of UX writing content across both physical devices and mobile products. The process was manual, inconsistent, and slow — with multiple international teams revising the same strings for clarity, tone, and accuracy.
As a result, verification took weeks, slowed down product launches, and created friction across teams.
Goal
The team’s mission was to design an internal AI-powered SaaS tool that would unify workflows and reduce revision cycles for global user facing contents.
Hypothesis
My first assumption was that the problem was primarily a tooling gap.
Only if we could collect existing contents from our key users and automate verification using AI, they would be thrilled about it and seamlessly adopt the new system into their workflow.
I mean, on paper, it sounded pretty logical.
Step 1.
Train AI on existing content.
Step 2.
Automate verification checks.
Step 3.
Save teams time and reduce errors.
Can't get any simpler than this…right?
The Problem
As I dug deeper, it became clear the problem wasn’t just technical. It was cultural and organizational.
There were so many layers that I overlooked when thinking about the scope of the project.....When I opened the can, I was faced with:
Global complexity: Dozens of international teams, each with their own workflows, guidelines, and cultural nuances.
Content scale: Thousands of strings across physical and digital products.
Workflow friction: Revisions were bottlenecked — it could take weeks to finalize even small batches of content.
Alignment challenges: Engineers, UX writers, and content reviewers had competing priorities and different definitions of “done.”
Resistance from legacy teams: Many groups had long-standing manual processes and were reluctant to add anything new to their workflow.
This challenge went beyond design — it was a product–market fit gap within the organization.
The Journey
The problem wasn’t just about writing quality —
it was about workflow misalignment at scale.
Designers, writers, PMs and QA all used different flows and checkpoints. Content bounced between teams (via the classic email on separate excel attachments...oy), often duplicating review cycles.
The Journey cont.
Through workshops, audits, and stakeholder interviews, I learned...
There was no unified brand tone or language system to normalize data
Each business unit created its own verification flow and rules, which were never shared outside their teams
As a result, verification processes relied on whatever resources were available with the reviewer, at the moment.
Whatever we were building needed to be THE SINGLE SOURCE OF TRUTH that connected all contributors across teams, regardless of their location and role.
Designing MAL
Research showed that, despite the complexity of the project, user needs were straightforward:
We made an early decision to design MAL as a web app rather than a standalone tool. This was driven by two realities:
Security constraints: All data needed to remain within the LG cloud ecosystem.
Adoption concerns: Requiring a separate login would create friction and resistance. By embedding MAL within existing systems, we minimized the sense of “one more thing to manage” and kept workflows seamless.
As a designer, I tried to keep my focus on these three pillars when ideating my solutions.
Solution
Systems Thinking: Building from the Ground Up
My first step was auditing LG’s legacy label database. There were so many patchwork of inconsistent formats, tones, and even conflicting rules across divisions. AI couldn’t be effective without a structured foundation.
I led the effort with UX writers and global stakeholders to establish universal brand guidelines — unifying core standards while accommodating each division’s restrictions, before moving into design.
Seamless AI Integration: Training for Trust
Once we had a consistent dataset, I partnered with engineers to separate the database into two parts:
Label Data (the content being verified)
Rule Data (the standards it must meet)
This structure allowed us to train AI models that could search, verify, and cross-check brand guidelines across divisions.
Non-Disruptive Design: Meeting Users Where They Are
From my earliest interviews, one theme was loud and clear:
This resistance shaped my design philosophy. Instead of introducing a complex interface, I deliberately designed MAL to feel familiar and frictionless.
SEARCH:
A simple search engine (what users are familiar with) — one box, one primary action.
Users can toggle between three options (search sentence, verify, or search branding terms).
By keeping the experience minimal, adoption felt almost effortless.
VERIFY: Clear UI indication to Build Trust
Designed the outputs to be explainable: each revision was flagged with a clear before/after indication, so users understood why something was flagged and why it needed correction. This transparency was critical in building trust with skeptical teams.
ACCESS: Guideline Library for All
No more awkward favors to different teams to gain access on the team's content guidelines. All version controlled guidelines regardless of divisions and teams are now stored in one place as a single source of truth.
The Results
The first version of MAL rolled out across select LG teams — and the results spoke for themselves:kept workflows seamless.
The biggest effect wasn’t just efficiency. It was laying the foundation for AI to scale responsibly across LG. MAL created the structure — tone, rules, workflows — that made future automation possible.
Anddddddddd,
Reflection
This project taught me that designing at enterprise scale is as much about people as it is about tools. I learned how to navigate global workflows, balancing complexity with simplicity so the system was powerful enough to scale yet easy for teams to adopt. I saw firsthand how alignment requires patience and clear communication, especially when priorities conflict across functions and time zones. And I came to appreciate the trade-offs of AI in practice — it could accelerate reviews, but cultural nuance still demanded human judgment. In the end, the real design challenge wasn’t just building a tool, but building trust in the system around it.
PORTFOLIO
CHECK OUT SOME MORE
description
CHRONICLE
CHRONICLE
CHRONICLE
description