Lesson 1 of 85 min read
What AI assistants can and cannot do for testers
Claude and GitHub Copilot are fast helpers, not testers. Knowing where they help saves hours; knowing where they fail keeps you safe.
Key takeaways
- 1Claude (by Anthropic) is a chat assistant: great for reading requirements, drafting test ideas, summarising logs and writing bug reports.
- 2GitHub Copilot lives inside your code editor: it suggests code as you type and has a Chat panel for writing, explaining and fixing test code.
- 3Both can sound confident and still be wrong. You stay responsible for every test case and every line of code you keep.
A good split of work
You read the story and decide what matters most. Claude drafts 20 test ideas in a minute. You delete 6 weak ones and add 3 edge cases it missed. Copilot then helps you turn the top 5 into Playwright tests, and you run and review them.
Go deeper
Expert notes beyond the basics
- 01
Chat assistant vs coding assistant
Claude works best with text: requirements, notes, logs, emails, documents. Copilot works best with code because it sees your files in the editor. Many testers use both: Claude to think and plan, Copilot to build.
- 02
Where AI is weakest
It doesn't know your product's hidden rules, your users or last month's production incident. It can't tell which bug would hurt the business most. That risk-based thinking is the tester's job.
- 03
Give it context, not just a question
An assistant answers based only on what you tell it in the conversation plus general knowledge; it has no memory of your product unless you paste in relevant detail. 'Write test cases for the login page' gets generic output. 'Write test cases for our login page, which uses email plus OTP, locks after 5 failed attempts, and supports Google sign-in' gets something useful immediately, because the assistant now has the constraints that actually matter.
- 04
Treat data privacy as a first question, not an afterthought
Before pasting anything into a chat assistant, ask whether it contains real customer data, credentials, internal URLs, or anything confidential. Many companies have rules about which AI tools are approved and what can be shared with them. Default to masking or removing identifying details, and check your company's AI usage policy before relying on a tool for day-to-day work with real data.
Common mistakes to avoid
- - Copying AI output straight into a test plan without reading it.
- - Using AI for tasks you don't understand, so you can't spot its mistakes.
- - Giving a prompt with no context about the actual product and expecting specific, useful output.
- - Pasting real customer records or credentials into a prompt without checking company policy first.
Ready-to-copy prompts
Copy, replace the [square brackets] with your details, and paste into the tool shown. Remove passwords, keys and customer data first.
Split a task between you and AI Claude
I am a QA tester working on [feature, e.g. checkout with coupons]. List the testing tasks for this feature in two groups: 1. Tasks an AI assistant can draft for me (and what I must review in each) 2. Tasks I must do myself because they need product, user or risk judgement Keep it to 5 items per group, one line each.
Learn a testing concept fast Claude
Explain [concept, e.g. regression testing] to a beginner tester in under 150 words. Then give one real-world example from an e-commerce app and one common interview question about it with a short model answer.
Try it yourself
List two tasks you'd give AI and one you'd always keep yourself.
Write your answer on paper or in a sheet first, then compare.
Interview questions
How to answer like a confident QA engineer
- Q1
Will AI replace testers?
Model answer
AI speeds up repetitive work like drafting test cases and boilerplate code, but it does not understand the business, the users or the risk the way a tester does. Testers who use AI well become faster and more valuable; the judgement part of the job stays human.
- Q2
Where does AI help most in QA today?
Model answer
Drafting test ideas from requirements, writing and refactoring automation boilerplate, explaining errors and logs, generating test data, and turning rough notes into clear bug reports.
- Q3
Why does context matter so much when using an AI assistant?
Model answer
The assistant only knows what you tell it plus general knowledge; it has no access to your product unless you describe it. Providing the relevant rules, constraints and examples turns a generic answer into a specific, usable one.
- Q4
What should you check before pasting data into an AI tool at work?
Model answer
Whether it contains customer personal data, credentials, secrets or confidential business information, and whether your company has approved that specific tool for that kind of data. When in doubt, mask or remove the sensitive parts first.
Knowledge check
+5 points
Who is responsible for a test case Claude wrote?