How Crystal Intelligence cut regression from weeks to hours

Company Overview

• Company: Crystal Intelligence.
• Industry: Crypto compliance / Financial intelligence.
• Business Type: Blockchain analytics & risk intelligence platform.
• Size: 10–50 employees.
• Headquarters: UK / Europe.

Key Use Cases

• Complex canvas & data visualization testing.
• Automated UI regression testing.
• Visual color verification (e.g., Risk Scoring).
• CI/CD pipeline integration (PR quality gates).

The Challenge: Manual regression and canvas testing

Crystal Intelligence provides a blockchain analytics and risk intelligence platform used to investigate crypto transfers and trace potential money laundering. While their backend APIs were well-covered by automated integration tests, their UI testing remained entirely manual.

This manual approach created a severe bottleneck as the engineering team pushed to increase their deployment velocity. For instance, some of the core updates required a full manual regression of the app that took long days of heavy effort, ultimately stretching the release cycle to four weeks.

Furthermore, automating their UI was technically daunting. The platform's most critical feature is a visual investigation tool built on an HTML canvas. Because canvas elements lack standard HTML IDs, legacy DOM-based automation tools like Playwright could not reliably interact with the visual nodes or verify crucial details – such as whether a crypto wallet's risk score was correctly colored green, yellow, or red.

The Solution: AI agents for testing complex environments

To achieve their goal of releasing new features every two to three days without blindly scaling their manual QA headcount, Crystal Intelligence partnered with QA.tech. Their team focused on:

• Testing the canvas: Instead of relying on hidden code selectors, QA.tech's AI agents evaluate the application visually. The agent can actually "see" the canvas, interpret the graphs, and interact with the UI elements just like a human compliance officer would.
• Goal-oriented user stories: Instead of writing brittle scripts, the QA team prompts the AI chat using natural language user stories. The agent autonomously navigates the platform, searches for specific wallet addresses, adds transactions to the graph, and verifies data.
• Securing internal access: To maintain strict security protocols over their sensitive data, Crystal Intelligence utilizes QA.tech's secure SSH tunnel to safely run these AI agents directly against their internal, non-production development environments.

The Results: Speeding up regression from weeks to hours

By adopting QA.tech, Crystal Intelligence is transforming a painful, weeks-long manual regression cycle into a rapid automated process.

Dmitry Anisimov highlighted the unique capability of QA.tech for their needs:
"The key requirement for us was the ability to test canvas. QA.tech can recognize visual differences and interact with canvas elements – something we haven’t seen in any other tool."

Ultimately, QA.tech helped Crystal Intelligence to integrate AI automation for UI checks directly into their CI/CD pipelines. By catching visual bugs and verifying complex canvas workflows on every pull request, their developers can ship crucial crypto compliance features with speed and confidence.

Highlights

• AI agents capable of visually interpreting and interacting with complex HTML canvas elements.
• Eliminates manual testing bottlenecks, transforming weeks-long regression cycles into hours.
• Validates visual cues natively, such as checking risk score colors (green, yellow, red) on transaction graphs.