The Challenge
UI testing was manual, with no automation platform to start from. Writing one test took about two days, and choosing test data needed a bioinformatics specialist. In a regulated lab environment, any automation also had to leave a reviewable trail.
The Approach
We co-built an agentic Playwright framework with the quality engineers: an agent chain with three human gates (acceptance criteria, plan, merge) and a reviewable artifact at each step. A train-the-trainer model meant the team, not Liatrio, ran and taught the workflow.
What We Built Together
Test-workflow engine plus TestRail and Jama provider plugins (provider-agnostic)
Agent chain with three human gates and an artifact at every step
Test-data skill: selection ~2 hrs → minutes, no specialist needed
Rerunning an existing case: ~15 min → under 20 sec
Beyond scope: 269 legacy UI tests migrated from Java to TypeScript
Business Outcome
What began as one team's pilot became shared infrastructure. Two teams outside the original scope adopted the framework without being asked, and quality engineers on a further team now write acceptance criteria and build tests on it directly.
ROI Achieved
$73 AI cost per ticket against a $96/hr labor basis
$119 to $1,271 net saved per test, depending on hours saved
Break-even after ~46 min of saved effort
FROM THE READOUT, MATH CHECKED BY LIATRIO
Break-even 73 ÷ 96 = 0.76 h ≈ 46 min
Low 2 h × $96 − $73 = $119
High (16 − 2) h × $96 − $73 = $1,271
