# QA Orchestrator: One Clear Verdict Before Any Change Ships > The QA Orchestrator is the AI quality assurance agent in the Ai1 platform by MyZone AI that checks a change before it goes out, from code and pages to agent instructions, content, server setup and data, sizes the check to the risk, brings in the right specialist agents and gives one verdict with evidence. Canonical page: https://myzone.ai/pages/agents/ai-qa-agent Part of Ai1, by MyZone AI. Book a QA walkthrough: https://calendly.com/d/ct6h-tcy-8qf/ai1-demo?a1=QA%20Orchestrator&utm_source=myzone.ai&utm_medium=agent-page&utm_content=ai-qa-agent-final Last updated: 2026-10-05. Reviewed by the MyZone AI team. Any change, one check sized to the risk, one verdict with evidence. ## At a glance - **What it checks:** Code, pages, agent instructions, written content, server setup and data - **How deep:** One of four depths; payments and logins always get a deeper check - **Who runs the checks:** Specialist agents, such as the Tester and the Code Review Agent - **The verdict:** Pass, pass for a limited scope, changes requested or fail - **Every verdict names:** The exact version checked and anything not tested - **What it never does:** Fix or ship the change: fixes go to the agent that owns them ## What the QA Orchestrator does - **Sizes the check to the risk:** A typo fix gets a quick look; a change to payments or logins gets a deep one. It picks one of four depths based on how risky the change is and what kind of change it is. - **Brings in the right specialists:** It assembles the checks a change needs, such as automated tests, code review, visual checks, accessibility and speed, security, and content and tone, and runs each through the agent that owns it. - **Counts only verified findings:** Every finding must be backed by something that can be checked. Claims that cannot be verified do not count, so the verdict is not built on guesswork. - **Gives you one verdict:** You get one answer: pass, pass for a limited scope, changes requested or fail. Findings are ranked, and where specialists disagree you see both sides. - **Says what it did not test:** Each verdict names the exact version it checked and lists anything it did not cover, so a partial check is never mistaken for a full one. - **Works on more than code:** It also checks screens and pages, written content, server setup, data changes and AI agent instructions, testing an agent's behaviour against test cases and reading its tone. ## How the QA Orchestrator differs from a pile of separate test reports Separate test reports leave you to add them up. The QA Orchestrator sizes each check to the risk, brings in the specialist agents a change needs, counts only findings backed by evidence and gives one verdict: pass, pass for a limited scope, changes requested or fail. It names the version it checked, lists what it did not test, and never fixes or ships the change itself. How this differs from the Tester (https://myzone.ai/pages/agents/ai-tester-agent): the Tester runs browser checks and test suites and the Code Review Agent reads the code, while the QA Orchestrator sizes the check, brings them in and gives one verdict. ## How it works 1. **A change comes in** (When a change is ready): You, or the agent that built the change, hand it over for checking. 2. **Size the check** (Before any check runs): It weighs the risk and the kind of change, then picks the right depth. 3. **Run the checks** (Side by side): Specialist agents run tests, reviews and other checks side by side. 4. **Verify the evidence** (Before a finding counts): Each finding is checked against real evidence before it counts toward the verdict. 5. **One verdict** (Once the evidence is in): A pass goes back to whoever built the change. Anything that fails goes to the right agent to fix, then gets checked again. ## When to use it - **A new page is about to go live:** It runs a standard release readiness check covering tests, layout, accessibility and code review, and gives you one verdict. - **A change touches money, logins or customer data:** It picks a deeper check and brings in security and edge-case testing before giving its answer. - **You updated the instructions for one of your AI agents:** It tests the agent's behaviour against test cases and reviews its tone before you rely on it. - **You need to know exactly what was covered:** It records the version it checked and lists anything it did not test. ## What you get - One verdict: pass, pass for a limited scope, changes requested or fail - Findings ranked by severity, each with the evidence behind it - The exact version that was checked - A list of anything that was not tested - Both sides shown where specialists disagree ## Example: QA verdict on a checkout change Example with a fictional company. Names, people and figures are invented to show the agent's output. Any resemblance to a real company or person is unintended. What it was asked: Run QA on our checkout change (yearly billing at 20% off plus a new coupon field) and tell us if it is safe to release. The QA Orchestrator rated the change for a deep check because it touches payments, planned six checks and briefed the Tester agent, which ran four of them on a synthetic copy of the checkout: a build check, the team's own tests, five checks written from the brief, and a browser check at 1280 and 375 px wide. The QA Orchestrator then re-checked the Tester's evidence and re-ran every failure three times. Code review and security review were planned for round 2 but not run, and the live payment provider, tax and accessibility were not tested. Run date: 30 September 2026. ### What it found - Yearly billing and a coupon both applied, so a yearly Clinic plan with a 15% coupon cost CAD $399.84 instead of CAD $470.40, losing CAD $70.56 per sale. - A coupon that ended on 1 September was still accepted on 30 September, because the code compared the dates as text rather than as dates. - For one coupon order the page showed CAD $16.15 but sent a charge of CAD $16.14, so the receipt and the card charge would disagree by a cent. ## Guardrails - It checks work; it never fixes or ships it. Fixes go to the agent that owns them, and releasing is the DevOps Agent's job. - Findings count only when they are backed by evidence that can be checked. - Every verdict names the version checked and lists what was not tested. - Reports are cleaned of private details before they are shown anywhere public. ## Frequently asked questions ### What goes into a release readiness check? The checks a change needs before it goes out, sized to its risk. For a new page, the QA Orchestrator in Ai1 by MyZone AI plans tests, layout, accessibility and code review, then combines the results into one verdict that names the version checked and lists anything not tested. Besides code, it also checks screens and pages, AI agent instructions, written content, server setup and data changes. ### How much QA testing does a small change need? Less than a risky one. The QA Orchestrator weighs how risky the change is and what kind of change it is, then picks one of four depths: a typo fix gets a quick look, while changes that touch payments or logins always get a deeper check. That keeps everyday work quick without waving risky changes through. ### What should a software QA verdict tell you? Whether the change can go out, and how far you can rely on that answer. The QA Orchestrator gives one of four verdicts: pass, pass for a limited scope, changes requested or fail. Findings are ranked and backed by evidence, the exact version checked is named and anything not tested is listed. You do not sign off every verdict. A verified pass goes straight back to the agent that asked, and failures go to whoever fixes them. ### Does the QA Orchestrator fix or release changes, and is it available? No. It reports the problems it finds, sends them to the right agent to fix, then checks the change again. Releasing and deploying belong to the DevOps Agent. The QA Orchestrator is ready to use on your Ai1 server. ## More on this topic - [Multi-Agent Development: Why the Future of Software Is Built by AI Teams](https://myzone.ai/pages/blog/multi-agent-development): Discover how multi-agent AI systems are transforming software development. Teams of specialised AI agents build products 10x faster at lower cost. ## About Ai1 Ai1 is the AI operations platform by MyZone AI, where each client runs on its own private server. The QA Orchestrator is included on every Ai1 level, including Developer Core, with no per-agent charge. It works alongside the other Ai1 agents on your account. Pricing: https://myzone.ai/pages/services/ai1-pricing. Security: https://myzone.ai/pages/security.