Skip to content

QA and Testing Services

Software testing that keeps pace with AI-built code

Our QA practice covers manual and automated functional and regression testing, accessibility testing, load, stress and soak testing, static code analysis, and expert-led secure coding validation for AI-generated code. It runs inside our delivery and managed support work, so the team that finds a problem also fixes it.

Why testing has to change when AI writes the code

AI tools produce plausible code faster than anyone can read it line by line, so quality has to come from layers of checks, each built to catch a different kind of fault, with a named person accountable for each. Automated checks prove exactly what each test asserts. People judge the design, the usability and whether the feature does what the business needed.

That is how we work. In our own delivery, 91.6% of our code is AI-authored, and 100% of it is reviewed by senior engineers and validated by ISTQB-qualified QA. The example below follows one story through the four stages of testing, from the build to sign-off.

An example of how one user story is tested: tagged tests for every acceptance criterion, test cases approved by the QA analyst, triage, and a manual session that ends in sign-off. Illustrative data for a fictional client.

What our QA practice covers

The same checks apply to code we write and to code we did not build, including code written with AI tools.

Function: Functional testing

Manual and automated functional testing as standard. Every acceptance criterion gets automated tests that prove it, manual cases designed to challenge it, and time-boxed exploratory sessions aimed at the riskiest changes.

Function: Regression testing

Automated regression suites check that what used to pass still passes. Where they exist on a system we support, we maintain them as part of managed support.

Security: Static code analysis

Static code analysis with GitHub Advanced Security, extended with our own in-house rules, runs automatically on every pull request.

Security: Secure coding validation

Senior engineers validate AI-generated code against the OWASP Application Security Verification Standard (ASVS) and our own suite of checks, built from the findings of penetration tests on previous projects. The reviewer is never the person who wrote the change.

Accessibility: WCAG 2.2 AA testing

Automated accessibility testing against WCAG 2.2 AA, plus standard techniques and tools.

Performance: Load, stress and soak testing

We use Azure Load Testing to find and fix scalability problems, design scaling strategies such as Kubernetes autoscaling, and size infrastructure and cloud SKUs for performance and cost.

Two kinds of testing sit outside our practice. We do not carry out penetration tests: when you need one, commission it from a CREST-accredited tester, and we fix its findings alongside our own. Our accessibility testing is automated testing plus standard techniques and tools, not a manual audit with assistive-technology users. Our guide to penetration tests and readiness reviews explains which to do first.

How a change is tested, from criteria to release

Cheap, fast checks run first and slow, thorough ones later, so the people at the end spend their time on judgement rather than on problems a machine could have found. The exact suites vary by project.

How a software change is tested: acceptance criteria, automated checks, code review, manual testing and UAT
The seven stages in which a software change is tested, and who owns each. 1, agree the acceptance criteria: the business analyst agrees each story's criteria with the client's product owner, and they stay in the work-item system, where delivery tooling can read but not rewrite them. 2, build it with tests: the engineer has automated tests written with the code, each tagged with the story and criterion it proves. 3, check every pull request, highlighted: automatic checks run a release build, unit tests, static analysis with GitHub Advanced Security, and a coverage check that fails the build when a criterion has no covering test. A gate follows: an engineer reviews every pull request before it merges. 4, draft and approve test cases: happy, negative and boundary cases for each criterion, drafted with AI and signed off by the QA analyst. 5, run the slower suites: automatic database integration and end-to-end tests before release, and the full regression suite overnight. 6, test it by hand, highlighted: the QA analyst's manual and exploratory testing on the test environment, with the business analyst judging whether the feature makes sense to its users. 7, user acceptance testing: the client's team tests the feature against the agreed criteria, and requested changes go back through analysis before the criteria are rewritten.

What testing found on a change we measured

We measured one change closely: two complex, linked user stories on one of the booking platforms we build, which went from open questions to two merged pull requests in 15 hours 35 minutes. Across the session, the unit suite of about 2,400 tests ran twice with no failures, and the combined smoke and story tests, around 160 of them, passed in full.

Testing still found a fault the build gates had missed. During QA on the second story, a smoke test from an earlier, unrelated story failed. The first story had introduced a new kind of history entry, the second story's test created one, and the older test did not recognise it. Neither the story's own tests nor the smoke tests on their own exposed it. When a fault gets past a layer like this, we fix it and then change the layer that should have caught it, and several of the rules and guard tests in our delivery harness started that way.

The full account is in Anatomy of an AI Change, and our post on how we keep quality high when AI builds at speed sets out who owns each of the twelve layers that checked it.

Seen enough? Let's talk through your requirements.

Book a free consultation

Frequently asked questions

What testing do you offer?

Our QA practice covers manual and automated functional and regression testing, accessibility testing, load, stress and soak testing, static code analysis, and expert-led secure coding validation for AI-generated code. We run it as part of our delivery and managed support work, so the team that finds a problem can also fix it.

Can we buy testing on its own?

We do not offer testing as a standalone resourcing service. If you need an independent view of software someone else built, including AI-generated code, the Production Readiness Review is a fixed-fee, two-week review at £5,500 that ends in an evidence pack. Ongoing testing then runs through managed support or a development engagement.

Who does the testing?

Three roles share it. Engineers direct the automated tests that prove each acceptance criterion. Our ISTQB-qualified QA analyst designs the approach to test cases, signs off every case and runs the manual and exploratory sessions. A business analyst agrees the criteria with your product owner and joins the manual session to judge whether the feature makes sense. In our own delivery, 91.6% of our code is AI-authored, and 100% of it is reviewed by senior engineers and validated by ISTQB-qualified QA.

How do you test AI-generated code?

The same way we test any code, with two extra safeguards. The acceptance criteria live in the work-item system, where our delivery tooling can read them but not rewrite them, so the code is always measured against something the AI did not write. And senior engineers validate AI-generated code against OWASP ASVS, with a reviewer who never wrote the change. Our post on keeping quality high when AI builds at speed shows every layer on a measured change.

Do you carry out penetration testing?

No. Our secure coding suite is built from the findings of penetration tests on previous projects, so issues that real tests have found are checked for on every review. If a customer or insurer needs a penetration test, commission it from a CREST-accredited tester, and we can fix its findings alongside ours.

Do you carry out accessibility audits?

We test against WCAG 2.2 AA with automated tools, plus standard techniques and tools. That is not the same as a manual audit or testing with assistive-technology users. If a buyer or regulator asks for one of those, commission it from a specialist accessibility auditor, and we can fix what it finds.

Can you load test our system?

Yes, as part of delivery or managed support. We use Azure Load Testing for load, stress and soak testing, to find and fix scalability problems, design scaling strategies such as Kubernetes autoscaling, and size infrastructure and cloud SKUs for performance and cost.

See how AI is changing how we build software

Our quarterly AI Velocity Report tracks real delivery metrics from live projects: how much of our code is AI-authored, how delivery timelines compare to baseline, and which tools are making a measurable difference. No marketing spin, just honest data from a team that builds software every day.

Read the latest AI Velocity Report

Talk to us about testing your software

Book a consultation and we will tell you how we would test your system, what we would check first, and what it would cost.

Book a consultation

or call 01202 006729