Qualis-Lab
Qualis-Lab
Financial Services · Argentina

Banco Hipotecario: test automation with AI agents

Design and evolution of an AI-agent platform for web automation using Playwright MCP, with agent memory centralized in PostgreSQL on OpenShift and test-data generation via APIs.

Banco Hipotecario

The challenge

Starting point

A financial-sector organization wanted to evolve its test-automation strategy to reduce manual intervention by functional teams and increase validation coverage across its web applications.

The solution

Qualis strategy

We implemented an architecture based on AI agents with specialized instructions to use the Playwright MCP, enabling automated flows to run in a simpler, more guided way. The work focused not only on building new agents but also on the continuous improvement of the existing ones, refining the processes so that functional analysts could run tests with less technical knowledge.

As part of the platform's evolution, we carried out the migration of the memory system used by the agents, replacing a LanceDB-based implementation with a centralized solution on PostgreSQL. This migration prepared the infrastructure for deployment on OpenShift, consolidating information into a single database, simplifying administration and improving the scalability and maintainability of the environment.

In parallel, the solution kept integrating with functional processes tied to the banking core, allowing the agents to support increasingly complex business scenarios and progressively expand automated test coverage.

Additionally, we collaborated on an automatic test-data generation project. The banking CRM was initially considered as the generation source but, after detecting stability and performance limitations, the strategy was redefined to generate data directly through APIs. This change delivered a significantly more robust, reliable and efficient process.

Results

Measurable impact
  • Evolution of a Playwright-based AI-agent platform for web automation.
  • Simpler test execution for functional teams.
  • Agent memory centralized in PostgreSQL, with deployment ready for OpenShift.
  • A more scalable and maintainable architecture.
  • Continuous agent improvement to increase automation coverage.
  • A more stable test-data generation mechanism via APIs, removing dependencies on low-reliability systems.

Technologies

  • Playwright
  • Playwright MCP
  • AI Agents
  • PostgreSQL
  • OpenShift
  • APIs
  • Test Automation
  • Data Generation

Want a case like this for your company?