Process optimization

One quality system. Two delivery models.

EnvironmentQA / Staging
ACTIVE MODELHuman-only
FLOW STEPS12
QUALITY GATES6
RELEASE OWNERHuman approval

Company operating model

Overall QA workflow

One shared workflow for the company. CP and AUR differ only at the planning checkpoint that triggers QA Test Case creation; quality gates and release controls remain the same.

CP

Create Test Cases after Grooming

AUR

Create Test Cases after Dev Planning

01

Requirement

Product / BA

Define business outcome, acceptance criteria, risk, dependencies, and environment needs.

02

Planning checkpoint

Product + Dev + QA

CP: Grooming confirms scope. AUR: Dev Planning confirms the technical approach and dependencies.

03

Create QA Test Cases

QA

CP: create after Grooming. AUR: create after Dev Planning. Link scope, data, expected result, and risk.

04

Build & self-test

Developer

Build in Dev; run unit/API/integration self-tests and required CI checks.

05

QA Staging

QA

Manual functional, exploratory, API, smoke, and approved stable automated regression tests.

06

QA complete

QA

Record Passed/Failed, evidence, known risk, and Go / Conditional Go / No-Go recommendation.

07

Release decision

Product + Engineering

Humans approve release, risk acceptance, feature flags, monitoring, and rollback readiness.

08

PRD smoke & learning

QA + Dev Lead

QA records safe production-smoke result and severity; Dev Lead manages mitigation or rollback.

Defect loopQA finds Fail → QA confirms severity → Dev fixes and self-tests in Dev → New QA/Staging deployment → QA retests the same Test Case.A Failed QA Test Case must never become Passed without new retest evidence.
QA environment boundaryQA executes formal testing in QA/Staging; Production is safe smoke only.
Automation ruleQA activates automated regression only when a Test Case and its requirement are stable.
Release authorityQA recommends; Product and Engineering make the human release/risk decision.

Workflow · click a step

CP (Human) → QA Test Case Repository

Human AI assisted
62%

Common workflow status mapping

CP (Human) → QA Test Case Repository

Common Human workflow: CP controls delivery progress; QA Test Cases hold test execution and evidence.

CP · Human delivery workflowQA · Test Case repository
How to useClick a status pair to see the accountable owner, required action, evidence, and handoff rule in the right panel.Failed QA Test Case always returns through defect/fix and a new QA deployment before retest.

Flow comparison

Human-only vs Human + AI

AI shortens drafting and signal analysis; it never owns a quality or release decision.

FLOW STAGE1. Scope & risk2. Coverage design3. Build & verify4. Decision & learning5. Future NFR
Human-only 12 operating steps
Brief → Refinement → Ready
QA plans manual coverage → Pending stability
Dev unit/self-test → QA Staging manual → stable cases automate regression
QA complete → UAT → release → PRD QA smoke
Roadmap: performance and security testing
Human + AI 15 steps · AI optional
Brief → AI gap draft → Human review → Ready
AI drafts coverage + scaffold in parallel → QA keeps CI activation Pending stability
Dev unit/self-test → AI groups signals → QA Staging validates → QA activates regression when stable
Same human-controlled release path
Same roadmap; AI is optional analysis support
Human control
Product, Dev, and QA approve scope
QA owns scope, scaffold review, and stability decision
QA validates evidence and approves promotion
QA records smoke severity; Dev Lead owns rollback and incident action
Humans set thresholds and approve risk

Time-use simulator

Model recurring QA time before committing automation

Adjust the operational assumptions. This is a planning estimate, not a productivity promise.

Human-only15.0 hper selected regression schedule
Human + AI + automation11.2 hper selected regression schedule
Recurring time released3.8 hfor exploratory and risk work
One-time automation build investment27.0 hEstimated break-even: 8 similar release cycle(s)

Human-only

Advantages

  • Lowest setup cost; works immediately for a manual-first team.
  • Best for new UX, changing requirements, and exploratory risk.
  • Simple accountability and evidence in Jira.

Trade-offs

  • Repeated regression consumes QA capacity.
  • Results vary more by tester and available time.
  • Slower feedback when releases become frequent.

Human + AI + automation

Advantages

  • Less repeat execution; more QA time for exploration and risk.
  • Faster drafts, failure grouping, and evidence preparation.
  • Consistent regression signals in CI for stable critical paths.

Trade-offs

  • Requires initial build time, maintenance, and reliable test data.
  • AI output and automation failures still require human validation.
  • Automating unstable or one-off cases creates cost without value.

Non-negotiable control

AI accelerates the work. Humans own the decision.

Both workflows pass through G1 Ready, G2 Ready for QA, G3 QA Complete, G4/G5 Release, and G6 Production Health.

Company setup blueprint

Minimum operating system for a new company

Start with a version-controlled, low-overhead stack. Add paid test-management or observability platforms only after the team has a measurable need.

Work board

AUR + CP Jira projects

Keep requirements, delivery work, and defects in AUR/CP. Use their existing QA statuses rather than creating a parallel board.

OwnerProduct + QA
Done whenAUR/CP issue has an owner, acceptance criteria, risk, environment, and linked QA Test Cases.
Requirement quality

AUR / CP Story, Task, Bug

AUR follows Ready to test → Testing → Ready to Deploy/Release. CP follows DEPLOYED QA → QA Inprogress → QA Done. Use Jira links rather than duplicate documents.

OwnerProduct + QA
Done whenEvery requirement has testable acceptance criteria and a QA handoff status.
Test case management

QA Jira project · Test Case

Use the existing Test Case lifecycle: To Do → Testing → Passed/Failed → Done. Keep wait for confirm for the current confirmation process only.

OwnerQA Manager
Done whenEvery QA Test Case is linked to its AUR/CP requirement and records the tested build/environment.
Traceability

Jira issue link

Create a standard link between each AUR/CP item and its QA Test Cases. Do not rely only on a key such as [AUR-22] in the Test Case summary.

OwnerQA Manager
Done whenAUR/CP and QA Test Case show reciprocal Jira links.
Source control

GitHub Pull Requests

Protect main branch: PR, one review, passing CI, linked issue, no direct push.

OwnerEngineering Lead
Done whenA test change cannot bypass review.
CI quality gates

GitHub Actions

PR: lint, unit, API/contract test. Main: smoke suite. Nightly: regression. Post-deploy: production smoke test and monitoring links. Store reports as artifacts.

OwnerEngineering + QA
Done whenRequired checks block merge; production smoke has an owner and a recorded result.
Automation rollout

QA Jira + GitHub Actions

Add a separate Automation state field: Manual, AI scaffold review, Pending stability, Automated regression, Retired. AI may draft in parallel, but QA activates CI only after stability.

OwnerQA
Done whenEvery automated regression check links to a stable QA Test Case, has a QA owner, and runs reliably in CI.
API testing

OpenAPI + Bruno

Commit collections and environments without secrets; run critical API checks in CI.

OwnerQA + Developer
Done whenAPI contract has a versioned test collection.
UI testing

Playwright

Automate only stable, critical end-to-end journeys; use data-testid selectors; quarantine flaky tests.

OwnerQA Automation
Done whenSmoke suite runs in ≤15 minutes and produces a report.
Performance roadmap

k6 (future phase)

After stable automated regression, establish a load baseline for core transactions and run it before major releases.

OwnerEngineering + QA
Done whenThreshold, test data, and environment are agreed; no production load test without approval.
Security roadmap

Dependency scan + SAST (future phase)

After CI and test data are stable, add dependency scanning and SAST first; add risk-based DAST/security testing for exposed critical paths.

OwnerEngineering + Security
Done whenFindings have severity, owner, target date, and release-risk rule.
Environment boundary

Dev / QA Staging / Hotfix / Production

Dev owns unit/self-test in Dev. QA validates only in Staging. Hotfix environment is opened only when needed. QA performs smoke only in Production.

OwnerEngineering + QA
Done whenEvery Jira test result records the environment and deployed version.
Release & learning

Release checklist + monitoring

Release notes, rollback owner, smoke checklist, alert links, and RCA action for significant escapes.

OwnerQA Manager + Engineering
Done whenEvery release has a decision record; action items are reviewed next sprint.

Verified Jira mapping

Use the workflow already configured in each project

AUR · AuroraRequirements and product delivery

Requirement Reviewing → Ready to Plan Review → Code Reviewing → Ready to test → Testing → Ready to Deploy / Release

Link every item to QA Test Cases.
CP · Core PlatformRequirements and engineering delivery

Code Review → DEPLOYED QA → QA Inprogress → QA Done → Done

Link every item to QA Test Cases.
QA · Test CaseManual execution and regression evidence

To Do → Testing → Passed / Failed → Done

Add Automation state separately; do not overload the status.

Status flow mapping

Project statuses that represent the QA journey

AUR · Aurora
Requirement ReviewingReady to testTestingReady to DeployRelease
CP · Core Platform
To Do / In ProgressDEPLOYED QAQA InprogressQA DoneDone
QA · Test Case
To DoTestingPassed / FailedDone

Decision support

What actually changes?

AreaHuman-onlyHuman + AI
Requirement & test draftingManual analysis and writingAI produces drafts; human validates
Exploratory testingQA-ledQA-led — protected time increases
Defect documentationQA writes from evidenceAI drafts; QA reproduces and approves
Release decisionHuman onlyHuman only
Data/privacy exposureLowerRequires approved AI, redaction, and access rules

Recommended rollout

Run a 2–4 week pilot on Low / Medium-risk tickets.

Measure QA effort, lead time, escaped Critical/High defects, rework from incorrect AI output, and risk-coverage completeness. Expand only when time improves without weakening quality.