Quality assurance · Production

How Mercado Libre's accessibility team uses AI to scale support, ticket enrichment, and review workflows

The problem

Mercado Libre's accessibility team needed to scale support for hundreds of designers and developers across questions, reviews, and continuous improvements without proportional growth in team size.

Workflow diagram · grounded in source
1
A11Y assistant triggered in support channel
trigger
“Activates when mentioned in the support channel”
2
RAG-based answer generation
ai_action
“Consults internal documentation, training materials, historical queries, previously reported accessibility tickets, and our design system. Uses a large language model (LLM) with Retrieval-Augmented Generation (RAG) to deliver reliable an…”
3
AI enriches accessibility tickets
ai_action
“For each generated ticket, AI automatically adds a contextual note covering three key aspects”
4
Handoff assistant analyzes screen image
ai_action
“Starting from a screen image (desktop web or native mobile), the assistant analyzes the visual context and platform type. It then generates a descriptive visual map”
5
Daily resolved ticket pull
trigger
“Pulls all accessibility tickets resolved that day”
6
AI agent reviews tickets and PRs
ai_action
“Instructs the AI agent to: Analyze ticket comments and solution evidence. Identify linked GitHub pull requests (PRs) and review their technical content. Assess the clarity, relevance, and documentation of the solution.”
7
Traffic-light fix quality classification
validation
“Classifies the ticket with a traffic-light emoji (🟢 green, 🟡 yellow, 🔴 red) based on fix quality”
8
Results stored in shared spreadsheet
output
“Stores the results in a shared spreadsheet”
Reported outcome

Multiple AI initiatives were launched to automate responses, enrich accessibility tickets with context, and assist design handoffs and ticket reviews, enabling the team to scale impact and reduce manual effort.

Reported metrics
Work speedspeed up work
Team independenceincrease independence
Understanding scalesignificantly scales understanding
Manual review timereduce manual review time
Show all 5 reported metrics
work speedspeed up work
team independenceincrease independence
understanding scalesignificantly scales understanding
manual review timereduce manual review time
team time freedgain time to focus on the next strategic challenges
Reported stack
large language model (LLM)RAGAxeJiraGitHubFuryFigma
Source
https://medium.com/mercadolibre-tech/how-we-are-using-ai-in-mercado-libres-accessibility-team-e960b83283a9
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Multiple AI initiatives were launched to automate responses, enrich accessibility tickets with context, and assist design handoffs and ticket reviews, enabling the team to scale impact and reduce manual effort.

What tools did this team use?

large language model (LLM), RAG, Axe, Jira, GitHub, Fury, Figma.

What results were reported?

Work speed: speed up work; Team independence: increase independence; Understanding scale: significantly scales understanding; Manual review time: reduce manual review time (source-reported, not independently verified).

How is this quality assurance AI workflow structured?

A11Y assistant triggered in support channel → RAG-based answer generation → AI enriches accessibility tickets → Handoff assistant analyzes screen image → Daily resolved ticket pull → AI agent reviews tickets and PRs → Traffic-light fix quality classification → Results stored in shared spreadsheet.

Related quality assurance cases