Solutions
QA Testing Automation
For engineering leaders whose releases wait on manual regression testing, and whose oldest code has the thinnest coverage. We help your team use AI to write tests that actually check something.
You're likely a fit if…
Releases wait on manual regression testing.
Every release means clicking through the same flows by hand, and something still slips through.
Your most important code has the fewest tests.
The oldest parts of the system pay the bills, and developers avoid touching them.
AI-written tests passed and checked nothing.
Your team tried it and got tests that confirm whatever the code does today, bugs included.
Customers find bugs a test should have caught.
And each one costs you time and trust.
What we build
Test generation grounded in your codebase
Agents write tests that follow your conventions and target real behavior, from a spec that names the code involved. An engineer reviews every one.
Learn more: Test generation grounded in your codebaseCharacterization tests for legacy code
Today's behavior pinned down before anyone changes it, so refactoring stops being a gamble.
End-to-end tests for key user flows
Browser tests for the journeys that matter, run on every pull request and kept current as the interface changes.
Independent review of test quality
A separate reviewer, with fresh context, checks that each test would fail if the code were wrong.
Manual test plans where automation isn't worth it
Step-by-step plans drafted from the ticket and the code change, for your testers to run.
Scope
Included, and not
A pilot on one repository: the harness, test patterns, CI hooks and team practices, then a rollout plan. Your engineers keep every merge decision. Not included: outsourced QA staffing or replacing your CI platform.
What we need from you
An engineering lead as owner, repository and CI access, a staging environment, and time from the people who test today. Your code stays in your repositories, under your access controls.
What drives cost
The number of repositories and teams, the state of your test infrastructure, how many flows need end-to-end coverage, and how many people we coach. Builds are scoped and priced after the assessment.
How we'd start
Start with an Assessment, or a focused review of one repository and how your team tests today if engineering is the whole question.
Run a pilot: one repository, one representative feature, one team practice.
Set up the harness: grounded specs, independent review, evidence and a human merge gate.
Hold a weekly review of where the agents went wrong, and fix the harness.
Hand your team the skills, documentation and harness to run on their own.
Common questions
Can AI write useful tests for legacy code?
Yes, with the right harness. The risk is tests that lock in today's bugs, so tests are grounded in a spec, reviewed independently and approved by an engineer.
Does this replace our QA engineers?
No. It moves them toward the judgment work: deciding what to test, exploratory testing and reviewing what agents produce.
Do we have to use PRFlow?
No. PRFlow is an open-source project I lead, and a working reference for this approach. We can use it, adapt it, or apply the same practices with the tools your team already has.
How long does it take?
A pilot on one repository is usually S (1–3 weeks). Rolling out across several teams is typically M (1–2 months) or L (3–6 months).
Industries where this applies
Let's find where AI pays off in your business.
Thirty minutes with Daniel. No pitch deck: you talk about your business, I tell you honestly where AI fits and where it doesn't.