Quality assurance (QA) prevents defects. It covers the standards, processes and pipeline rules that build quality into how software gets made. Quality control (QC) finds defects. It inspects and tests the actual output, such as a pull request, a build or a release, before it reaches users. QA asks whether your process is set up to produce working software. QC asks whether this specific release works.
What is the difference between QA and QC?#
QA is about the process and QC is about the product. QA works before and around the work, and QC checks what the work produced.
The formal definitions say the same thing. The American Society for Quality defines both terms using ISO-style vocabulary. Quality assurance can be defined as "part of quality management focused on providing confidence that quality requirements will be fulfilled." Quality control can be defined as "part of quality management focused on fulfilling quality requirements." NIST's glossary uses the same QA definition.
ASQ also explains how the two relate: while quality assurance relates to how a process is performed or how a product is made, quality control is more the inspection aspect of quality management. It also notes that QA activities and responsibilities cover virtually all of the quality system in one fashion or another, while QC is a subset of the QA activities.
In practice:
- QA decides that every change needs tests, a security scan and a senior review before it merges.
- QC is those tests, that scan and that review running on change #4,812.
What does quality assurance cover in software teams?#
QA covers every decision that shapes how code gets written, checked and released. These decisions are made once and then applied to every change.
In a software team, QA usually includes:
- Definition of done. What has to be true before a story counts as finished.
- Coding and review standards. Who reviews what, what reviewers check, and which changes need a senior engineer.
- Test strategy. Which layers get tested (unit, integration, end-to-end, contract), what coverage you expect, and which suites block a merge.
- Rules for AI tools. Which assistants are approved, what data they can see, and how AI-written code is reviewed.
- Secure development practices. Dependency policy, secrets handling and threat modeling. NIST's Secure Software Development Framework is a common reference, and NIST has finalized SP 800-218A, Secure Software Development Practices for Generative AI.
- Pipeline design. Which checks your CI/CD pipeline enforces automatically and which ones nobody can skip.
- Learning loops. Post-incident reviews that change the process, not just fix the bug.
A useful test: if an activity changes how the next hundred changes will be built, it's QA. Our Dedicated Teams work this way. Teams agree shared rules for code review, tests, AI tool use and definition of done up front, then check them every sprint.
What does quality control cover, and when does it happen?#
QC is inspecting a specific piece of output against its requirements. In modern teams it happens at several points, not just in one phase at the end.
Typical QC points in a software delivery flow:
- On every commit or pull request. Unit tests, linting, static analysis and dependency scans.
- Before merge. Code review against the agreed standards, plus integration and contract tests.
- Before release. End-to-end and regression suites, performance checks, accessibility checks, and user acceptance where needed.
- After deploy. Smoke tests, canary analysis and production monitoring against service-level objectives.
Each QC point gives a verdict on one thing: pass, fail, or pass with known issues. When QC keeps failing in the same way, that's a sign the QA process needs fixing.
How do QA and QC compare side by side (goal, timing, owner, output)?#
QA is preventive and owns the system. QC is detective and owns the verdict on each release.
| Quality assurance (QA) | Quality control (QC) | |
|---|---|---|
| Goal | Prevent defects | Detect defects |
| Focus | Process: how software is built | Product: what was built |
| Timing | Before and across the work; reviewed periodically | On each change, build and release |
| Key question | "Will our process produce working software?" | "Does this release work?" |
| Typical owner | Engineering leadership, quality engineering lead | The delivery team, enforced by pipeline gates |
| Output | Standards, test strategy, pipeline rules, policies | Test results, review approvals, scan reports, release evidence |
| Failure mode | Rules on paper that nobody enforces | Late, manual inspection that slows releases |
| Example | "No merge without a passing security scan" | The security scan running on this pull request |
Where does testing fit: is it QA or QC?#
Running a test is QC. Deciding which tests exist, when they run and whether they block a release is QA.
That's why the two get mixed up. The same people and tools do both jobs:
- A test automation engineer writing a regression test for a bug that escaped is doing QC (checking this behavior) and QA (making sure the bug can't come back unnoticed).
- Setting a rule that end-to-end tests must pass before deploy is QA. The nightly run of those tests is QC.
- A flaky test suite people have learned to ignore is a QA failure. The process has lost its ability to give you confidence, even though the QC step technically still runs.
If you're structuring a team, don't hire "QA testers" and expect them to fix your process. Give someone explicit ownership of the test strategy and the gates, and treat test execution as something the pipeline does on every change. Our Test Automation practice aims for test suites fast enough to run on every change and stable enough to trust.
How do QA and QC work in a CI/CD pipeline with AI-written code?#
QA defines the gates once, and QC runs them on every change, whether a person or an AI wrote it. AI makes this division more important because it increases the volume of changes that need checking.
The data shows why. In Stack Overflow's 2025 survey, 84% of developers said they use or plan to use AI tools in their development process, up from 76% in 2024. However, 46% of developers said they don't trust the accuracy of the output from AI tools (press release). The survey's AI section adds that experienced developers are the most cautious, indicating a widespread need for human verification for those in roles with accountability.
Security is the sharpest risk. Veracode's 2025 GenAI Code Security Report analyzed 80 curated coding tasks across more than 100 large language models (LLMs), revealing that while AI produces functional code, it introduces security vulnerabilities in 45 percent of cases. Google's DORA research reaches a similar conclusion at the organizational level. Its 2025 report finds that AI's primary role is that of an amplifier, magnifying the strengths of high-performing organizations and the dysfunctions of struggling ones.
The practical result: when code is cheap to produce, QC can't stay a manual phase at the end. It has to be automated, run on every change and backed by rules QA has made non-negotiable. This is how we run our Quality Engineering service. Every release passes three gates:
- Tested. Tests prove it works. AI can help write tests, but engineers review them.
- Secured. Security and dependency checks prove it's safe.
- Reviewed. A senior engineer's review proves it's right.
Whether a person or an AI wrote the change, nothing ships without all three. Each release also produces evidence: a record of which checks ran and who approved the change. That gives QA something to audit and improve.
Everything from the code change to production monitoring is QC. The loop back to standards is QA. If you never update the gates after an incident, you only have half the system. For a step-by-step version of this flow, see our guide on how to verify AI-generated code before production.
Who should own QA and who should own QC on a product team?#
Engineering leadership should own QA: the standards and the gates. The team that ships the change should own QC on its own work, with the pipeline enforcing the checks.
A workable split:
- VP of Engineering or CTO. Owns the quality bar: what "releasable" means, what risk is acceptable, and what gets reported.
- Quality engineering lead. Owns the system: test strategy, pipeline gates, tooling, test stability and release evidence.
- Product team (developers, QA and automation engineers, DevOps). Owns QC on every change: writing and maintaining tests, fixing failures, getting changes reviewed.
- Senior engineers. Own review, especially for risky changes and AI-written code.
- Product managers. Own acceptance criteria, so QC has something concrete to check against.
Avoid a separate QC team that receives finished work over a wall. A testing phase at the end turns quality into a queue and makes developers less responsible for what they ship. Our Product Engineering Pods put a lead, engineers, QA and automation, DevOps and product in one team. The people who build a feature are also accountable for proving it works.
Do you need both, or can test automation replace QC?#
You need both. Test automation doesn't replace QC; it is QC, done by a machine instead of by hand. What automation replaces is repetitive manual checking.
Automation has clear limits:
- It only checks what someone specified. If no test covers a behavior, automation won't notice when it breaks. Deciding what to cover is a QA job.
- It can't judge design. A change can pass every test and still be the wrong approach or hard to maintain, or it can quietly widen access to data. That's what senior review is for.
- It can be wrong. AI-generated tests can assert the buggy behavior the AI just wrote. Tests need review too.
- It decays without ownership. Flaky, slow or ignored suites stop giving you confidence. That's a process failure, and fixing it is QA.
A quick self-check for your team:
- Can you name who owns the quality gates? (QA)
- Does every change pass the same tests, security checks and review before it ships? (QC)
- After an escaped defect, does a rule or gate change, or does only the code change? (QA)
- Can you show evidence of what was checked for your last release? (both)
If you answer no to any of these, that's where to start.
Conclusion#
QA and QC aren't interchangeable, and you can't pick only one. QA builds the system: the standards, test strategy and pipeline gates. QC applies that system to every change and returns a verdict. With AI writing a growing share of code, the pressure falls on QC to run automatically on every change, and on QA to make sure those checks are the right ones and can't be skipped.
If you want to put tests, security checks, review and release evidence on every release, talk to an engineering lead.
Frequently asked questions
Is code review QA or QC?
Reviewing a specific pull request is QC, because you're inspecting a piece of output against a standard. Deciding that every change needs review, who can approve it and what reviewers must check is QA. Most teams need both: a written review policy and reviews that actually happen on every merge.
What is quality engineering, and how is it different from QA?
Quality engineering is a hands-on way of doing QA. Instead of writing process documents, it builds quality into the delivery system itself: automated test suites, pipeline gates, security checks and release evidence. It covers both sides, because engineers design the gates (QA) and the pipeline runs them on every change (QC).
Does a small startup need a separate QA team?
Usually not at first. A small team can treat QC as part of every engineer's job and use CI to enforce tests and checks on each change. What it does need is one person who explicitly owns the quality bar and the gates, so QA doesn't drift as the team and codebase grow.
How do you know if your QA process is working?
Look at outcomes, not activity. Useful signals include defects that reach production, how often releases need a rollback or hotfix, how stable and fast your test suites are, and whether you can show what was checked for any given release. If escaped defects keep repeating the same pattern, the process needs a new rule or gate.
Should AI-generated code go through different checks than human-written code?
It should go through the same gates, applied without exception. AI increases the volume of changes and can introduce security flaws or tests that confirm buggy behavior, so the gates need to be automated and enforced rather than optional. Many teams also add specific rules for AI tool use, such as which tools are approved and what extra review is required.
- Quality Assurance
- Quality Control
- Software Testing
- CI/CD
- AI-Generated Code
- Quality Engineering
Newsletter
Get new articles by email
A short monthly round-up of what we've written on engineering pods, quality and AI. No spam.
