Hiring QA Engineers in the AI Era: What Actually Matters

1. What changed in QA between 2023 and 2026

Three years ago, the QA automation workflow looked roughly the same as it had for a decade. A QA engineer would examine a feature, design test scenarios, manually write test scripts in Selenium, Cypress, or Playwright, and spend a significant portion of their week maintaining those scripts as the application evolved. The skill that mattered most was the ability to write reliable, maintainable test code.

Today, AI tools can handle the mechanical parts of that workflow. Tools like Assrt automatically discover test scenarios by crawling an application and generate Playwright test files without manual scripting. Other tools convert natural language descriptions into executable tests. Self-healing frameworks automatically update broken selectors. The tedious parts of QA automation, the parts that consumed 60% to 70% of a QA engineer's time, are increasingly automated.

This does not mean QA engineers are unnecessary. It means the value they provide has shifted. When AI can generate a test script in seconds, the ability to write test scripts manually is no longer a differentiating skill. The differentiating skills are now test strategy, risk assessment, edge case identification, and the ability to evaluate whether AI-generated tests actually cover what matters.

The parallel to software development is instructive. AI code generation tools like Copilot did not eliminate the need for developers. They eliminated the value of typing speed and syntax memorization while increasing the value of architecture, design, and code review skills. The same transformation is happening in QA.

2. Skills that matter now vs. before

Declining in value: manual test scripting, selector craftsmanship, framework boilerplate knowledge, and the ability to reproduce a bug by manually walking through a specific set of browser interactions. These skills are not worthless, but they are no longer what separates a great QA engineer from an average one.

Increasing in value: test strategy and risk assessment. Can the candidate look at a feature and identify which test scenarios provide the most coverage for the least effort? Can they prioritize testing based on business impact rather than technical convenience? Can they identify the edge cases that AI tools consistently miss (race conditions, cross-service interactions, data migration scenarios)?

AI tool proficiency is a new skill category that did not exist three years ago. This means the ability to configure, prompt, and evaluate AI testing tools effectively. A QA engineer who can set up Assrt to discover test scenarios, review the generated tests for coverage gaps, and enhance them with business-specific assertions is more productive than one who writes every test from scratch.

Systems thinking matters more than ever. As applications become more complex (microservices, event-driven architectures, real-time data pipelines), the ability to reason about system behavior holistically is crucial. AI tools are good at testing individual workflows but struggle with cross-system interactions, eventual consistency scenarios, and failure modes that span multiple services.

Communication skills have become more important, not less. A QA engineer who can clearly articulate test strategy to stakeholders, explain risk tradeoffs to product managers, and collaborate with developers on testability is extremely valuable. AI handles the code; humans handle the judgment and communication.

3. The right split between AI and human testing

The question is not "should we use AI for testing?" but "what should AI handle and what should humans handle?" Getting this split wrong leads to either wasted human effort on tasks AI can do better, or insufficient coverage because AI missed critical scenarios.

AI should handle: initial test generation for standard CRUD workflows, selector maintenance and self-healing, regression test suite updates when UI changes, visual regression screenshot capture and comparison, and boilerplate setup/teardown code. These are high-volume, repetitive tasks where AI is consistently faster and more accurate than humans.

Humans should handle: test strategy and prioritization, edge case identification, security testing scenarios, accessibility evaluation, performance testing design, and the final review of AI-generated tests. These require business context, risk judgment, and creative thinking that AI models do not reliably provide.

A practical ratio for most teams: AI generates approximately 70% of E2E test code, and humans write the remaining 30% (complex scenarios, edge cases, integration tests). Humans review 100% of AI-generated tests before they enter the permanent test suite. This is similar to how developers use AI code generation: let the tool handle the straightforward parts, write the tricky parts yourself, and review everything.

The review step is non-negotiable. AI-generated tests can have subtle issues: assertions that check the wrong thing, selectors that target a similar but incorrect element, or test flows that skip important intermediate steps. A human reviewer catches these issues in minutes; an unreviewed test can mask bugs for months.

4. How to evaluate QA candidates in 2026

Traditional QA interviews focus on writing test scripts live, often in Selenium or Playwright. This evaluates a skill that AI can now do in seconds. Here is a more effective interview structure for the current landscape.

Test strategy exercise (45 minutes).Present the candidate with a realistic product feature (e.g., "We are adding a group payments feature to our fintech app"). Ask them to design a test strategy: What scenarios would they test? How would they prioritize? What are the highest-risk edge cases? What would they automate vs. test manually? This evaluates their ability to think strategically about quality, which is the core skill that AI cannot replace.

AI tool evaluation (30 minutes). Give the candidate access to an AI testing tool (Assrt, or another tool your team uses) and a test application. Ask them to generate a test suite, evaluate its quality, and identify gaps. This evaluates their ability to work with AI tools effectively and their judgment about test quality.

Test review (30 minutes). Show the candidate a set of AI-generated tests with deliberate issues: flaky waits, incorrect assertions, selectors targeting the wrong element, missing edge cases. Ask them to review the tests and identify problems. This evaluates their ability to serve as the quality gate for AI-generated output, which is a primary responsibility in the new workflow.

System thinking scenario (30 minutes).Describe a complex bug that spans multiple services (e.g., "Orders are sometimes duplicated when the payment service times out"). Ask the candidate how they would investigate and design tests to prevent recurrence. This evaluates their ability to reason about distributed systems, which AI tools consistently struggle with.

5. Restructuring your QA team for AI

The shift toward AI-assisted testing changes not just individual roles but team structure. Here is how leading organizations are adapting.

Fewer test scripters, more test strategists. Instead of five QA engineers each writing and maintaining tests for their assigned feature area, organizations are moving toward two or three senior QA strategists who design the overall test approach, supplemented by AI tools that handle test generation and maintenance. The strategists review AI-generated tests, identify coverage gaps, and design the complex scenarios that require human judgment.

QA engineers as tool builders. Some teams are redefining the QA automation role as a tooling role. Instead of writing individual tests, QA engineers build and configure the AI testing infrastructure: setting up tools like Assrt, creating custom assertion libraries, building CI pipeline integrations, and developing internal testing frameworks that amplify the productivity of the entire team.

Embedded quality ownership. As AI handles more of the E2E test scripting burden, some organizations are distributing quality ownership to development teams. Developers use AI tools to generate E2E tests alongside their feature code, and a smaller central QA team provides strategic oversight, maintains the testing infrastructure, and handles cross-cutting concerns like performance and security testing.

The cost factor matters. Enterprise testing services like QA Wolf cost approximately $7,500 per month for managed E2E testing. For teams considering whether to hire another QA engineer or subscribe to a managed service, open-source tools like Assrt (free, MIT licensed) offer a middle path: AI-powered test generation without the managed service cost, suitable for teams with enough engineering capacity to run Playwright in their own CI.

Regardless of structure, invest in continuous learning. The AI testing landscape is evolving rapidly. Tools that are cutting-edge today may be obsolete in 18 months. QA engineers who stay current with new tools, understand their tradeoffs, and can evaluate new approaches objectively are the most valuable long-term hires.