QATES: More support for your testing team

The dominant AI narrative in the technology industry revolves around code generation: copilots that write functions, agents that build entire features, and pull requests that open themselves. But there is an imbalance that almost no one is discussing. If code is produced ten times faster while QA continues operating at the same pace, no real speed has been gained. The bottleneck has simply moved.

The Structural Cost of Manual QA

In any technology development project with a certain level of maturity, QA is not a stage. It is a continuous function competing for time with everything else.

Analyzing user stories, designing test cases, running validations, documenting results, and retesting after every change are activities that scale linearly with product complexity, even as delivery cycles continue to shrink.

This tension creates challenges familiar to any engineering team:

• A growing volume of manual testing that cannot be reduced without increasing risk.

• Increasingly narrow validation windows as releases become more frequent.

• Human error in repetitive tasks, precisely where mistakes are most costly.

• Rework caused by defects detected late in the cycle, after they have already affected development or production.

• Test coverage that deteriorates over time unless someone actively maintains it.

Traditional test automation solved part of the problem, but it comes with a structural limitation: it requires constant manual maintenance and lacks the judgment to adapt when requirements, workflows, or code change. It automates execution, not judgment.

QATES: An Agent Architecture, Not Just Another Assistant

Santex developed QATES (QA Testing Ecosystem) to address this problem at its architectural root.

Instead of relying on a general-purpose model to cover the entire QA lifecycle, the solution distributes the work across specialized agents that collaborate in a coordinated way. Each agent is responsible for one stage of the process and validates the previous agent’s work before moving forward.

The implementation is deliberately lightweight. A single command integrates the orchestrator, five sub-agents, and their workflows into the coding agent the team already uses, without modifying the repository or changing the existing workflow:

$ qates install --agent claude-code --preset acceleration

The agent team includes:

On that foundation, QATES addresses four concrete workflows that engineering teams manage every day:

• spec-to-test: transforms requirements into a test suite.

• pr-guardian: performs automated pull request reviews.

• regression-shield: protects the product against regressions.

• bug-hunter: proactively detects bugs before they reach production.

What QATES Is Not—and Why the Distinction Matters

Precision matters, particularly because this is where expectations around AI-powered QA often break down:

• It is not a new AI agent that teams need to configure. It integrates with the agent they already use.

• It is not a test runner. Tests continue running on the existing infrastructure, such as Playwright or pytest. QATES does not replace that infrastructure; it orchestrates it.

• It does not replace human QA professionals. It changes their role, moving them from executors to agent curators.

• It is not infallible. Its output must be reviewed and adjusted. Human oversight remains part of the system, not an optional step.

• It is not an autonomous tool. It needs context and feedback to continuously improve its results within each project.

This final distinction underpins the entire QATES design: AI speed is combined with specialist judgment, not used as a substitute for it. QA professionals can step away from repetitive execution and focus instead on supervising, validating, and applying business judgment to complex cases. That is precisely where their expertise creates the greatest value.

Measurable Impact

In an analysis of production use cases conducted for a major gas distribution company in Argentina, QATES reviewed 70 AI agent conversations in a single day, logged six bugs, and identified three new failure patterns. These included a critical security finding that a manual process would have been unlikely to detect within the same time window. At the feature-level QA cycle, estimates from an energy-sector project indicate reductions of up to 65% in analysis time and up to 55% in total cycle time compared with the manual process.

Teams with dedicated QA specialists—including the evaluated energy and consumer goods clients—use QATES to accelerate analysis and triage, freeing specialists to focus on higher-risk and more complex scenarios. For teams without a dedicated QA role, QATES works as a “pocket QA” for the development team: it generates tests, reviews pull requests, and triages bugs without requiring additional headcount.

What Comes Next

The QATES roadmap includes a metrics dashboard with historical run data, test coverage trends, and bugs per sprint, moving from estimates to measured results, shared memory across project collaborators, smoke test notifications delivered directly through Slack or Microsoft Teams, Multi-agent support for Cursor, Copilot, and Gemini, in addition to Claude Code, and native CI/CD execution, eliminating the need to trigger the process manually for every pull request.

The Conversation We Still Need to Have

The industry’s focus on code generation has left one question unanswered: Who ensures that the code works at the same speed at which it is produced?

QATES is Santex’s answer to that question—and potentially the missing part of the conversation about AI in software development. 

To assess your use case, schedule a session with our experts. Rather than offering a generic demo, we will conduct an in-depth review of your QA workflow to identify exactly where agent orchestration can create the greatest competitive advantage for your operation.