An AI Probe product
AI Testing Probe
AI drafts. An expert decides.
A learning test engineering platform that moves your manual test assets to automation. AI Testing Probe imports your manual test cases, learns from your team’s approvals, keeps improving prompt quality and automates the cases.
It runs on top of your existing tools, in your own cloud environment.
AI Testing Probe · manual case import
- Manual cases imported2,480 cases · Excel, Word, ALMDone
- StandardizationSteps, expected result, test dataDone
- Duplicate cases312 merge suggestionsIn review
- Link to requirementsMatched to Jira storiesDone
- Automation fit61% candidates · by priorityAwaiting approval
- Prompt qualityv14 · acceptance 74% → 81%Improved
Import. Learn. Improve. Automate.
Enterprises hold thousands of manual test cases built up over years: in different formats and tools, most of them not linked to a requirement. AI Testing Probe turns that backlog into an organized, traceable asset that is ready for automation.
- 01
Import
Reads manual cases from Excel, Word and legacy tools, converts them to a standard format, flags duplicates and imports them into your test management system.
- 02
Learn
Every analyst approval and rejection reason feeds back into the rule and prompt library. The platform learns your team’s quality standard.
- 03
Improve
Prompts that work become templates, and the acceptance rate of every prompt version is measured. Case quality improves over time.
- 04
Automate
Prioritizes the cases that suit automation and generates web, desktop, mobile and API automation code. Code arrives for review as a pull request.
Every approval and every rejection improves the platform.
When an analyst rejects a draft, they pick a reason: missing precondition, wrong expected result, out of scope, duplicate case. These reasons feed the rule and prompt library, so the next draft does not repeat the mistake.
The model is not retrained and your data is not used for training. Learning happens in a versioned library kept in your environment; every change is traceable and reversible.
- 01
Draft
The AI drafts a case with the rules and prompts in the library.
- 02
Approve or reject
The analyst decides against a checklist and picks a rejection reason.
- 03
Feedback
Rejection reasons become rule and prompt updates.
- 04
Measure
The acceptance rate of every library version is tracked and reported.
From manual case to result: who does what?
The roles of the AI and the expert are fixed at every step. Nothing the AI produces reaches the test management tool, the code repository or a release decision without approval.
| # | Step | What the AI does | What the expert does | Where it is written |
|---|---|---|---|---|
| 1 | Import manual cases | Reads cases from Excel, Word and legacy test management tools; splits them into steps, expected results and test data; flags duplicates. | The test analyst checks the import on a sample and approves it. | ALM Octane, Jira, Xray |
| 2 | Link to requirements | Matches cases to the related Jira or Azure DevOps stories and shows uncovered acceptance criteria. | The analyst verifies the matches and the coverage gaps. | Test management tool |
| 3 | Draft new cases | Drafts positive, negative and boundary cases from new requirements, with a risk tag. | The analyst approves each draft against a checklist or rejects it with a reason. | ALM Octane, Jira, Xray |
| 4 | Learn | Feeds approval and rejection reasons back into the rule and prompt library so mistakes are not repeated. | The quality manager reviews library changes. | Versioned library |
| 5 | Pick automation candidates | Suggests which manual cases suit automation based on repetition and risk. | The automation lead prioritizes the list. | Automation backlog |
| 6 | Automate | Generates web, desktop, mobile or API automation code from the approved case. | An engineer reviews the code as a pull request; it passes security and quality gates. | GitHub, Azure Repos |
| 7 | Run | The AI does not run at this step. | The CI pipeline runs only approved, deterministic scripts. | Azure Pipelines, GitHub Actions, Jenkins |
| 8 | Analyze | Classifies failed tests as environment, data, application defect or flaky test, and suggests locator repairs. | The test team validates the classification. | Jira, ALM Octane |
Why AI Testing Probe?
What matters to enterprise teams is not how autonomous a tool is, but how controllable it is and how well it uses what you already have.
- Your existing work is kept. Manual cases written over the years are not thrown away and rewritten; they are organized, linked to requirements and moved to automation.
- It runs on your existing tools. Jira, ALM Octane, Ranorex, Selenium, Appium, Playwright and your CI pipelines stay in place.
- It runs in your own cloud. It is deployed as containers in your environment; identity, network and security controls are yours.
- No lock-in. Every case and script it produces already lives in your tools.
- It learns your team’s standard. Not a generic model: a library shaped by your analysts’ approval and rejection decisions.
- Expert approval is mandatory. Every case, every code merge and every release decision goes through a person.
- No AI in CI. The AI works at design and maintenance time; test runs are deterministic.
- Adopt at your own pace. Each capability can be switched on or off per project without affecting the test service.
Benefit is measured before it is accepted.
Benefit is not a promise; it is a target tracked against values measured at the start. A capability is not rolled out before it passes the acceptance gate.
- ≥ 70%
- of AI-drafted cases accepted with minor edits; pilot acceptance gate
- ≥ 60%
- of AI-generated scripts compile and run at first review; pilot acceptance gate
- 30–50%
- targeted reduction in case preparation time per story; confirmed after the baseline measurement
- 20–25%
- targeted reduction in automation script development time per case
Targets are measured by having different analysts work the same story sets with and without AI, and are reported monthly.
Ready to use today.
Every AI Testing Probe capability is ready to deploy.
| Capability | Status | What it delivers |
|---|---|---|
| Manual case import and a learning library | Ready | Standardizes existing cases and links them to requirements; learns from approvals and rejections. |
| Test case generation from requirements | Ready | Faster case drafting; positive, negative and boundary coverage with traceability. |
| AI-assisted automation development | Ready | Framework-aligned code generation and maintenance; review through pull requests. |
| Failure classification and clustering | Ready | Separates environment and data failures from real application defects and flaky tests. |
| Locator repair suggestions | Ready | Reduces locator maintenance; assertion changes remain defects. |
| Test management and Jira connections | Ready | Automatic transfer of approved cases, run results and defects; a Jira ticket from every failed run. |
| AI-assisted exploratory testing | Ready | Explores paths outside the scripted cases and passes findings to expert review. |
| Risk-based regression selection | Ready | Selects the riskiest regression set for a change; switched on once it catches at least 95% of the defects of a full run for 4 sprints. |
Works with the tools you already use.
It connects to your team’s tools, from work tracking and design to test management, device clouds and app distribution. Your existing cases and automation suites are kept.
- Requirements and work tracking
- Jira, Azure DevOps, GitHub Issues, Confluence
- Test management
- OpenText ALM Octane, OpenText ALM, TestRail, Zephyr Scale, Xray, Azure Test Plans
- Manual case sources
- Excel, Word and all the test management tools above
- Design
- Figma, Zeplin
- Web
- Playwright (MCP), existing Selenium and Playwright suites
- Desktop
- Ranorex for execution, FlaUI-MCP for AI support
- Mobile and device cloud
- Appium (iOS and Android), Sauce Labs real devices
- App distribution
- Firebase App Distribution, Bitrise, Azure DevOps
- API and packaged applications
- REST, SOAP, OpenAPI; RestAssured, Postman; SAP and Salesforce flows
- Code and CI/CD
- GitHub, Azure Repos; Azure Pipelines, GitHub Actions, Jenkins
- AI clients
- Teams and Copilot Studio, GitHub Copilot, Claude Code
- Reporting
- QA Portal, Power BI
Want to test your AI agents and chatbots? Explore AI Agent Probe
In use in the field.
AI Testing Probe’s automation support is used in an FMCG customer’s multi-country mobile regression and in customer environments in retail and banking.
- 13
- country-language configurations in an Android regression, on 7 real devices in 51–56 minutes
- 0.9%
- infrastructure-related failure rate over 1,143 runs on 37 device models in 21 days
- 6
- retail and banking customer environments where it runs test processes
Free your test team from writing test cases.
Let’s start with one application and a few real requirements, and measure together how many AI drafts pass approval.