Contents
An agentic coding workshop for QA and test teams shows how coding agents can derive test cases, write test automation and analyse regressions – and why responsibility for test quality becomes more important as a result. When development teams produce code faster with agents, quality assurance has to keep pace. This article covers content, exercises and typical pitfalls.
How agentic coding changes the work of QA teams
Coding agents increase the pace of development. For QA teams, this cuts both ways: more changes arrive in less time and need to be tested – and the same tools are available to create and maintain tests faster. Teams that only experience the first side come under pressure. Teams that use both can strengthen their role: moving away from manually working through test cases towards designing test strategies and quality criteria.
The question of who writes tests is shifting, too. In development teams, agents often generate tests along with the code. Whether those tests check the right things is by no means guaranteed. This is where QA expertise becomes indispensable – as the function that critically assesses test coverage and understands business risks.
Where coding agents fit in quality assurance
The overview below shows where agents relieve QA teams and what responsibility stays with people:
| Area | What agents can do | What the QA team owns |
|---|---|---|
| Test design | Propose test cases and boundary values from requirements and acceptance criteria | Business completeness, risk-based prioritisation |
| Test automation | Write end-to-end, API and component tests and get them running | Meaningfulness, stability and maintainability of the tests |
| Test data | Generate synthetic, realistic test data | No real personal data, coverage of special cases |
| Flaky tests | Analyse causes of intermittently failing tests | Deciding whether the test or the application is at fault |
| Defect analysis | Evaluate logs, stack traces and reproduction steps | Assessing severity and priority |
| Test maintenance | Update tests after UI or API changes | Making sure updates do not hide real defects |
For an overview of methods and tools, see our article Test automation with AI: writing unit, integration and E2E tests faster.
Contents of an agentic coding workshop for QA and test teams
The focus depends on your team: some QA teams already automate heavily, others test mostly manually. Typical modules include:
- Coding agent fundamentals: how agents work, what permissions they need and how they run and evaluate tests on their own.
- From test case to test code: how testers without deep programming skills create automated tests with agents – and where the limits are.
- Using existing test frameworks: binding agents to your existing tools, page objects and helpers instead of letting parallel structures emerge.
- Reviewing tests: how to tell that a generated test is too weak, too narrow or simply wrong.
- AI-assisted exploratory testing: using agents as a source of ideas for edge cases and failure scenarios.
- Collaboration with development: a shared definition of done and test requirements for agent-generated pull requests.
We place AI developers from real-world practice as trainers who use agents for test automation in their own projects.
Hands-on exercise: from requirement to automated test suite
A central exercise takes a real requirement from your project through the entire process:
- Analyse the requirement: the agent reads the story and acceptance criteria and proposes test cases, including boundary values and negative tests.
- Prioritise test cases: the QA team rates the proposals by business risk, adds missing scenarios and removes redundant ones.
- Generate test code: the agent writes automated tests in the existing framework, reusing existing helpers.
- Run a counter-check: deliberately introduce a defect into the application and see whether the tests catch it. A test that never fails protects nothing.
- Check stability: run the tests several times to uncover timing- or order-dependent failures.
- Integrate: add the tests to the pipeline and extend the agent's rules in the context file.
Warning signs in AI-generated tests
Generated tests often look convincing. QA teams should immediately question these patterns in review:
- Tests check implementation rather than behaviour: they break with every refactoring but find no business defects.
- Expected values were copied from the current code: the test then confirms existing bugs instead of exposing them.
- Excessive mocking: when almost everything is simulated, the test hardly checks real interactions.
- Fixed sleeps: pauses instead of waiting for states make tests slow and flaky.
- Adjusted instead of fixed: an agent that "makes a failing test green" by changing the expectation may have hidden a real defect.
- Missing negative cases: only the happy path is covered; invalid input and permission errors are missing.
For reviewing the application code itself, we recommend our code review checklist for agent-generated code. How clients structure acceptance is covered in Acceptance and quality assurance.
Format and cost
Content, format (in-house or remote) and scope are agreed individually – for example as a dedicated session for the QA team or as a joint training with developers so that both sides know the same rules. Pricing is available on request by e-mail; workshops start from EUR 799 net.
Planning an app? Request a free fixed-price quote in 2 minutesConclusion
Coding agents fundamentally change quality assurance, but they do not make it redundant. They speed up test design, automation and defect analysis, while assessing risks, test meaningfulness and business completeness remains the QA team's job. An agentic coding workshop for QA and test teams teaches both: productive work with agents and the critical eye needed to review generated tests.
Frequently Asked Questions
Do testers need programming skills for an agentic coding workshop?
Basic test automation knowledge helps but is not mandatory. Content and exercises are adapted to your team's experience level.
Can coding agents fully replace manual testing?
No. Agents automate many repeatable tests, but exploratory testing, usability assessment and business plausibility checks remain human tasks.
How can you tell whether an AI-generated test is meaningful?
A proven method is the counter-check: deliberately introduce a defect and see whether the test fails. In addition, expected values should come from the requirement, not from the current code.
How much does an agentic coding workshop for QA teams cost?
The price depends on content, format and scope and is quoted on request by e-mail. Workshops start from EUR 799 net.