Before you go — grab our most downloaded resource.
QA cheat sheets used by 30,000+ testers.
🔒 No spam, ever. Join 30,000+ QA engineers. Unsubscribe anytime.
No thanks, I'll skip the free cheat sheetsThe SDET role has changed. This is the complete, practical guide to using AI agents in test automation and to the questions that now decide AI-SDET interviews — with a question bank and tools to practise on.
For a decade, the SDET job was a fairly stable stack: a language (usually Java), an automation library (Selenium, now often Playwright), API testing, a framework you designed yourself, and a CI/CD pipeline to run it all. That stack has not disappeared — but in 2026 a second layer sits on top of it. Interviewers now expect you to know how to work alongside AI agents: to generate tests with them, to review what they produce, and to test the AI-powered features your own company is shipping.
The important shift is one of judgement, not tools. The market has moved from "can you write a test?" to "can you decide what to test, direct an agent to write it, and catch the agent when it is wrong?" AI removed some of the typing; it raised the bar on design and review. That is the lens to bring into every answer below.
The single most common AI question in an SDET screen now is some version of: "How do you use AI coding agents for test automation?" The strong answer is not a tool name — it is a pattern. The pattern that has emerged for using agents on a test suite has three roles:
Decides what to test and in what order — reads the story and acceptance criteria and proposes coverage: happy path, negative, boundary, security, accessibility.
Writes the actual test code from the plan — page objects, API assertions, data setup — one file at a time so a human can review each.
Watches runs, and when a locator or assertion breaks because the app changed, proposes a repair instead of leaving a red pipeline.
Why this pattern impresses: it shows you know where AI helps (volume, boilerplate, first drafts, self-healing brittle locators) and where a human still owns the decision (what is worth testing, whether the generated assertion is meaningful, whether a "heal" is hiding a real bug). The failure mode interviewers listen for is the candidate who lets the generator run unchecked and ships duplicated or meaningless tests.
data-testid so generated and healed locators stay stable.MCP (Model Context Protocol) is an open protocol that lets any LLM call any tool through one standard interface, instead of every model needing custom glue for every tool. It has become common enough that "explain MCP" is now its own interview question for AI-adjacent SDET roles.
For a tester, MCP matters in two directions. First, it is how an AI agent reaches your world — your CI, a browser, a database, your test-management tool — through MCP servers that expose those as callable tools. Second, and more interview-relevant, when you are asked to test an MCP-connected agent, you are really testing three things:
You do not need to have built an MCP server to answer well. You need to show you understand that an agent with tools is a system with a new attack surface, and that testing it means verifying the boundaries as much as the happy path.
This is the concept that separates candidates who have used AI from candidates who have only read about it. Traditional tests use a fixed assertion: input X gives exactly output Y. But an LLM is non-deterministic — the same prompt can give different, still-valid answers — so a hard assertEquals does not work. The answer is evals: the unit tests of AI systems.
An eval scores an AI's output rather than checking it for exact equality. The main approaches:
Build a curated dataset of 50–200 known-good input/output pairs — a "golden set". Run the AI over it and score how close each answer is to the reference. This is your regression suite for a model or prompt: run it on every change and watch the score.
Use a second model to grade the first model's output for quality, correctness or reasoning, against a rubric you write. It scales far past what humans can grade by hand — but you validate the judge itself with periodic human spot-checks so you are not trusting a grader you never checked.
For agentic systems you go beyond "was the final answer right?" and score the process:
Grounding (does the answer cite the source it was given?), correctness, relevance, toxicity, latency per step, and cost per call. In production you also track escalation rate, resolution rate and user-satisfaction score. Being able to reel off two or three of these signals real hands-on experience.
Generating test cases from a user story is now a practical exercise you may be asked to do live. The reliable structure is a four-part prompt — Role, Context, Task, Format:
What turns this from a party trick into an engineering skill is the review discipline around it: run a self-critique pass, check the output against a rubric (are negatives and boundaries actually covered? any duplicates? any missing security or accessibility case?), and only then fold the survivors into your test-management tool. The prompt drafts; you decide.
More products now ship an LLM inside them — a chatbot, a summariser, a "generate with AI" button. Testing those features is a distinct skill, because the usual assumptions break:
The mental model to bring: you are no longer testing a function with one right answer, you are testing a probabilistic system against a set of properties and risks. Frame it that way and you will stand out.
The senior signal in an AI-SDET interview is judgement about risk. Expect at least one question on governance: how do you keep an AI-assisted testing process trustworthy? Points worth making:
You do not need to have written a governance policy. You need to show you would not let an unreviewed agent silently rewrite the test suite — that instinct is what separates a senior answer from a junior one.
You cannot fake hands-on AI experience in an interview, so build a little of it:
Drill the fundamentals and run a full timed mock — the AI questions land best on top of solid basics.
Sample questions across the topics above, with tight answers you can adapt. Click to reveal.
data-testid, run a self-critique pass, and reject duplicated or meaningless tests. The point is directed use with human review, not letting it run unchecked.Agentic AI testing has two sides. First, using AI agents to help create and maintain tests — an agent that plans coverage, generates test code, runs it, reads failures and repairs the tests. Second, testing agentic systems themselves — evaluating an autonomous, tool-using AI on whether it plans correctly, calls the right tools, follows the intended process and stays safe under adversarial input.
It is a three-role way to use AI coding agents for test automation. A planner decides what to test and in what order, a generator writes the actual test code, and a healer watches for failures and repairs brittle locators or assertions when the app changes. Naming this pattern in an interview shows you understand where AI helps and where a human still reviews.
MCP (Model Context Protocol) is an open protocol that lets any LLM call any tool through a standard interface. For testers it matters because it is how AI agents connect to your test tools, CI, browsers and data — and because testing an MCP-connected agent means verifying tool discovery, permission checks and safe tool use.
Evals are the unit tests of AI systems. Instead of a fixed assert, you score an AI's output against a golden dataset of known-good answers, or use a second model as a judge, and track metrics like grounding, correctness, tool-use accuracy and toxicity. Evals run on every prompt or model change to catch regressions.
Yes. AI augments the SDET role; it does not remove the fundamentals. You are expected to know automation (Selenium or Playwright), a language (usually Java), API testing, framework design and CI/CD, and then layer AI skills on top. The strongest candidates pair solid fundamentals with hands-on AI-agent experience.
Keep your core SDET skills sharp, then build one real project using an AI coding agent end to end, learn how to evaluate an LLM or agent (golden sets and LLM-as-judge), understand MCP at a high level, and practise prompt patterns for generating test cases. Be ready to discuss where AI failed you and how you caught it — that judgement is what interviewers probe.
Agentic AI in testing moves fast — this guide is reviewed periodically. Tool names change; the patterns (Planner-Generator-Healer, evals, MCP) are the durable part.