AI Launch Analysis
AI Launch Analysis gives each uploaded Qualflare launch a risk level from its pass rate, a summary, failing areas and recommendations, on every plan.
Last updated
In Qualflare, AI Launch Analysis rates each uploaded launch as Low, Medium, High or Critical risk from the pass rate of the tests that ran, and adds a summary, failing areas and recommendations.
| Fact | Detail |
|---|---|
| Where | The AI Analysis card on a launch's Overview tab (Launches) |
| Plans | All plans (available on Starter, Core, Scale, Enterprise, Open Source). Starter includes 10 launch analyses per month and Open Source 100 launch analyses per month |
| Runs on | Launches you upload, for example with qf <identifier> collect (Upload results), never on launches started in the app |
| Requires | Enable AI Features on in the project (Project AI settings) |
| Who | Anyone who can open the launch: Viewer+ · project (requires the Viewer role or higher in the project) |
How AI Launch Analysis works
AI Launch Analysis runs in the background for each launch you upload:
- It is queued at upload, and the card shows Generating AI analysis... until it finishes.
- It reads the launch's platform, OS, status, branch, environment, result counts and failing suites.
- The risk level is computed from the pass rate of the tests that ran, not chosen by the model.
- A launch where no test ran gets a Medium report that starts "No tests ran in this launch", with no model call and without using the allowance.
- If the launch's failures belong to failure clusters that are already named when the analysis runs, the analysis is given those clusters, up to 25. Clustering is a separate, slower job, so a new failure pattern's cluster is usually not named yet and adds nothing.
Where to find the analysis
AI Launch Analysis appears as a card on the Overview tab of the launch's detail page. It is generated automatically for launches created from uploaded results. Launches you start in the app are not analyzed, and their card keeps showing Generating AI analysis....
Read the analysis
When AI Launch Analysis is complete, the card shows up to four sections. Failing Areas and Recommendations appear only when they have items.
Executive summary
A brief, plain-language summary of the launch results, with a risk level badge:
| Risk level | Badge | Pass rate of the tests that ran |
|---|---|---|
| Low | Green | 95% or more |
| Medium | Orange | 80% to under 95% |
| High | Orange | 60% to under 80% |
| Critical | Red | Under 60% |
The pass rate is passed tests divided by passed, failed, error, timeout and aborted tests. Skipped tests don't count, and a launch where nothing ran shows Medium.
Risk assessment
Why the risk level was assigned: the patterns and failure types behind it.
Failing areas
Areas or suites with failures, each with its name, failure count, impact and a short description. Impact is High when the area holds 50% or more of the launch's failures, Medium at 20% or more, and Low below that.
Recommendations
A numbered list of specific, actionable next steps based on the analysis.
Fields and states
An AI Launch Analysis report has these fields, and each launch's analysis is in one of these states.
Report fields
| Field | Description |
|---|---|
| Executive Summary | Plain-language overview of launch results |
| Risk Level | Low / Medium / High / Critical, from the pass rate |
| Risk Assessment | Explanation of why the risk level was assigned |
| Failing Areas | List of areas with failures and impact levels |
| Recommendations | Numbered list of actionable next steps |
Analysis states
| State | What the card shows | What it means |
|---|---|---|
pending | Generating AI analysis... | The analysis is being generated. Refresh the page in a few moments |
completed | The report | The analysis is complete and ready to read |
failed | Failed to generate analysis | The analysis could not be generated. Qualflare retries automatically, so the card can still change. If it stays failed, it can't be re-run. The next launch you upload is analyzed as usual |
skipped | Analysis skipped | The workspace has used this month's launch analyses. They reset on the billing day, and Plan & Billing shows usage |
disabled | AI is switched off for this project | AI features are off for the project. Turn them back on in the project's settings to analyze new launches |
unavailable | The locked AI Launch Analysis card with Open Plan & Billing, or Generating AI analysis... | On an uploaded launch, the workspace's subscription is not active. Every launch starts in this state, so a launch you started in the app keeps it and shows Generating AI analysis... |
completed, unreadable | No analysis available | The analysis finished, but its report could not be displayed |
Limits and plan availability
AI Launch Analysis All plans (available on Starter, Core, Scale, Enterprise, Open Source) has these launch analyses per month:
- Starter: 10
- Core: Unlimited
- Scale: Unlimited
- Enterprise: Unlimited
- Open Source: 100
When the month's analyses are used up, new launches show Analysis skipped until the allowance resets on the billing day. Only the workspace Owner sees usage, on Plan & Billing. See AI usage limits.
Related
- Guides: Analyze launch results · Failure Clusters · Create, triage and close defects
- Reference: Launches · AI Features settings and limits · Plans and limits · Glossary: Launch
- CLI/API: Upload results with qf collect
Failure Clusters
Qualflare groups similar failures from uploaded launches into failure clusters, labels each at 5 failures and opens a defect. Review, merge and export them.
Quo agent
Use Quo, the Qualflare AI agent, to analyze suite coverage, suggest test cases and generate steps. It changes data only after you confirm in the chat.