In 2020, we wrote about managing Git repositories from a QA perspective. The article covered rebasing, branch cleanup, and staying in sync with development. In 2026, those fundamentals still apply, but the way QA engineers interact with code has transformed completely. AI coding assistants are now standard tools in our workflow.
The modern QA engineer's toolkit
Our QA team now operates with AI assistants integrated directly into their development environment. These aren't just chatbots - they're context-aware tools that understand codebases, test frameworks, and project history.
Claude Code (Terminal/CLI)
Anthropic's Claude Code runs directly in the terminal alongside Git. It reads files, understands project structure, and can write test cases, debug flaky tests, or refactor test utilities. QA engineers describe what they need in natural language, and Claude generates the code.
GitHub Copilot
Inline code completion trained on test patterns. Particularly useful for writing assertions, generating test data, and completing repetitive test setup code. Understands Playwright, Cypress, Jest, and pytest conventions.
OpenAI Codex / ChatGPT
Broader reasoning for complex test scenarios. Useful for explaining production bugs, analyzing stack traces, and designing test strategies for new features.
Cursor IDE
VS Code fork with native AI integration. QA engineers use it to navigate unfamiliar codebases, understand component relationships, and generate tests that match existing patterns.
MCP: connecting AI to your tools
The Model Context Protocol (MCP) is the breakthrough that made AI assistants truly useful for QA. MCP lets AI models connect to external systems - databases, APIs, browsers, and testing tools.
Browser automation via MCP
Claude can control Chrome or Playwright sessions, take screenshots, inspect elements, and debug failing UI tests. The AI sees exactly what the test sees.
Database inspection
MCP servers expose read access to test databases. AI can query for test data, verify state after test runs, and help debug data-related failures.
CI/CD integration
AI assistants can read build logs, analyze test results, and suggest fixes for pipeline failures. No more copy-pasting error messages.
Documentation access
MCP connects to Confluence, Notion, or internal wikis. AI can reference test plans, requirements, and historical decisions when writing tests.
AI-assisted Git workflows for QA
The fundamental Git operations haven't changed, but how we execute them has. Here's how our QA team uses AI with version control:
| Task | 2020 Approach | 2026 Approach |
|---|---|---|
| Understanding changes | Read diff line by line | "Summarize what changed in this PR and what tests might be affected" |
| Resolving conflicts | Manual merge in editor | AI explains both versions and suggests correct resolution |
| Writing commit messages | Manual, often inconsistent | AI generates conventional commit message from staged changes |
| Debugging failures | Search logs, Stack Overflow | AI analyzes error + code context, proposes fix |
| Creating test branches | git checkout -b test/XX-123 | Same, but AI scaffolds test files automatically |
Practical workflow: debugging a flaky test
Here's how our QA engineers use AI tools to fix a flaky Playwright test:
Step 1: Describe the problem
This test fails 20% of the time on CI but passes locally. Here's the test file and the last 3 failure logs.
Claude reads the test, understands the assertions, and identifies the race condition - a network request completing after the assertion runs.
Step 2: AI proposes fix
Claude suggests adding await page.waitForResponse() before the assertion, or using expect.poll() for retry logic. It shows the exact code change.
Step 3: Validate and commit
QA engineer reviews the fix, runs the test 10 times locally, then commits with an AI-generated message explaining the root cause and solution.
Test generation from requirements
One of the most powerful applications: generating test cases from user stories or requirements documents.
Input
A Jira ticket describing a new feature: "Users should be able to export their data as CSV. The export should include all transactions from the selected date range. Files over 10MB should be split into multiple parts."
AI output
A complete test file covering: basic export, date range filtering, empty results, large file splitting, file format validation, download link expiration, and permission checks. Each test includes setup, actions, and assertions.
The QA engineer reviews, adjusts edge cases, and adds the tests to the suite. What took 2 hours now takes 20 minutes.
Prompting patterns for QA
Effective AI usage requires good prompts. Our team has developed patterns that consistently produce useful results:
| Pattern | Example Prompt |
|---|---|
| Context + Task | "This is a React component using React Query. Write Playwright tests for the loading, success, and error states." |
| Constraint specification | "Generate test data for the user registration form. Email must be unique, password must meet complexity requirements, age must be 18+." |
| Pattern matching | "Here's an existing test for the login page. Write a similar test for the registration page following the same patterns." |
| Debugging request | "This test times out on line 45. The selector exists in the DOM. What could cause the click to hang?" |
What AI can't replace
Despite the productivity gains, human QA judgment remains essential:
- Risk assessment - AI doesn't understand business impact or user frustration
- Exploratory testing - Creative, curiosity-driven testing that finds unexpected bugs
- Test strategy - Deciding what to test, how much coverage is enough, and what to skip
- User empathy - Understanding how real users will interact with the product
- Communication - Explaining bugs to developers, negotiating priorities with product
AI accelerates execution. Humans provide direction and judgment.
Setting up your AI-powered QA environment
For teams looking to adopt these tools:
Start with code completion
GitHub Copilot or Cursor. Low friction, immediate productivity gains for writing assertions and test data.
Add terminal AI
Claude Code for complex debugging and test generation. It understands your full project context.
Connect MCP servers
Browser automation, database access, and CI/CD integration. This is where AI becomes truly powerful for QA.
Build team prompts
Document effective prompts, share templates, and establish conventions for AI-assisted testing.
Want to modernize your QA workflow?
Our team has integrated AI tools into testing workflows for companies ranging from startups to enterprises. We can help your team adopt these practices effectively.
Book a consultationNeed help with software testing?
BetterQA provides independent QA services with 50+ engineers across manual testing, automation, security audits, and performance testing.
