Computer vision-guided testing for a Malta-based testing laboratory.

A laboratory that ran out of eyeballs, not ambition.

The client is a Malta-based software testing laboratory that runs quality assurance for gaming and consumer software vendors. Most of their work involves running regression suites against client builds and manually verifying visual output, a category of testing that does not lend itself neatly to traditional automation frameworks.

The bottleneck wasn't ambition. It was time. Every new client build required a tester to physically watch the UI, compare it against an expected state, and flag differences. This is work that AI handles better than humans, but only if the tooling is set up correctly and the edge cases are addressed honestly. Most off-the-shelf visual testing tools fall over on dynamic content, localisation, or animated elements. The lab had tried several. None survived production use.

Three constraints that made this harder than it looked.

• Client diversity. Each client build had its own visual language, layouts, and edge cases. The system had to generalise without requiring per-client retraining.
• False positive tolerance. Testing labs live and die on trust. A flaky test that cries wolf is worse than no test at all. We had to get precision high before recall.
• Integration with existing workflows. The lab's engineers already had test cycles, reporting, and escalation paths. The system needed to slot in, not replace.

Four stages. Standard Elevate AI playbook.

The engagement ran through our standard four-stage framework. Each phase had a defined deliverable and a milestone gate.

Discover (weeks 1–2). We embedded with the lab's senior testing engineers to map their current manual QA workflow end to end. We identified which categories of test actually benefited from visual AI, and which were better left to traditional assertion-based tooling. Not everything was a nail.

Co-create (weeks 3–6). Rapid prototype built against one client's regression suite. We used computer vision frameworks to handle structural comparison, plus targeted LLM calls for semantic interpretation where pixel-perfect comparison would have produced false positives. Reviewed with the client's engineers every Friday.

Build (weeks 6–10). Production integration. Connected into the lab's CI pipeline, test management system, and reporting layer. Hardened the confidence scoring so the system only raised flags it was sure of. Added a human-in-the-loop escalation path for edge cases.

Scale (weeks 10–12). Rolled out across additional client suites. Documentation, handover, internal training for the lab's engineers. Monitoring and observability in place before we stepped back.

[Note: placeholder attribution. To be updated with a real, consented quote from the client before publication.]

Why this engagement became a pattern.

This was one of the three engagements that shaped the studio's early thesis. Each one involved a regulated or compliance-heavy operator. Each one involved integrating AI capability into workflows that existing vendors had failed to handle. And each one produced a pattern that's now informing our next ventures.

Test automation at this level of visual complexity, done right, is a product category of its own. Watch this space.