aiAI TESTINGHUB
Enrol now
AI AGENT TESTER · IN DEVELOPMENT

From requirements
to release evidence.

A human-controlled AI testing agent designed to help teams plan, automate, execute and manage testing for traditional software and AI applications.

Book a demonstration Explore the workflow
OUR AI AGENTS

Tools for the testing workflow

Explore our available demonstration and the testing capabilities currently in development.

LIVE DEMONSTRATION

Test Case Generator

Explore our dedicated demonstration for generating test cases. Review generated outputs before use and do not submit confidential information without approval.

Open demonstration

Opens in a new tab.

IN DEVELOPMENT

AI Agent Evaluator

Planned evaluation support for AI agents, including datasets, quality metrics, safety checks and evidence-based scoring.

Access not yet available
IN DEVELOPMENT

AI Agent Tester

Designed to coordinate test design, automation, execution, defect drafting, retesting and release-readiness evidence in one controlled workflow.

View planned capabilities
THE AI AGENT TESTER

One workflow for software and AI quality

The AI Agent Tester is being developed by AI Testing Hub to support the complete testing lifecycle. It is intended for websites and business systems as well as AI agents, chatbots, LLM applications and RAG systems.

The agent will use authorised project information to understand requirements, identify risks and recommend the testing needed for the agreed scope.

RequirementsUser storiesApplication linksExisting test casesBusiness documentsAPI specificationsAI prompts & knowledge basesAuthorised source code
PRODUCT WALKTHROUGH

See the AI Testing Agent in action.

Explore how the workspace brings test generation, automation and AI evaluation into one controlled environment.

AI TEST AGENT · PRODUCT PREVIEW DEMONSTRATION
AI Test Agent dashboard showing test case totals, generation runs, applications tested, priority distribution and test type distribution
01 / MONITORDashboard and coverage overviewTrack generation activity, application coverage, priorities and test-type distribution.

Development preview: Screens show the current product environment with demonstration data. Features and interface details may change as the AI Testing Agent develops.

Open the live demonstration
HOW THE AGENT WORKS

Two applications. One controlled test process.

The agent adapts its test design and quality checks to the system being tested while keeping evidence and human decisions traceable.

01
Web and business application testingFrom requirements and application access to release-readiness evidence
PROJECT INPUTS

Application context

  • Website and login
  • Requirements documents
  • User stories
  • Past test cases
  • API specifications
  1. 01Analyse requirementsMap journeys, rules and risks
  2. 02Generate test casesFunctional, UI, API, smoke and accessibility
  3. 03Build automationCreate or update the approved test repository
  4. 04Execute and inspectRun tools such as Playwright and capture evidence
  5. 05 · HUMAN CHECKPOINTReview failuresApprove, amend or reject draft defects before logging
  6. 06Retest and releaseRegression checks and release-readiness evidence
02
AI agent, chatbot and RAG testingFrom AI context and risks to measurable evaluation and release evidence
AI SYSTEM INPUTS

Agent context

  • Agent workflows and prompts
  • System rules
  • RAG knowledge base
  • Tools and MCP connections
  • Authorised source code
  1. 01Analyse risksMap tasks, expected behaviour, data and failure modes
  2. 02Create evaluation dataGolden datasets, edge cases and adversarial scenarios
  3. 03Define quality gatesMetrics, thresholds, red-team tests and evaluation code
  4. 04Run and scoreEvaluate responses, retrieval, citations, tools and security
  5. 05 · HUMAN CHECKPOINTReview failuresConfirm risks and approve evidence-backed draft defects
  6. 06Re-evaluate and releaseSafety regression, accepted exceptions and evidence pack

The exact test scope, tools, metrics and approval points are agreed with each organisation. Planned capabilities are shown for demonstration purposes while the AI Agent Tester remains in development.

PLANNED WORKFLOW

Testing from discovery to release

Each stage is designed to retain traceability between requirements, tests, results, defects and release evidence.

  1. 01

    Understand

    Analyse the supplied requirements, user stories, business processes and technical information.

  2. 02

    Design

    Identify risks and generate suitable test cases, datasets, evaluation criteria and expected results.

  3. 03

    Automate

    Prepare approved tests for execution using Playwright or other suitable UI, API and testing tools.

  4. 04

    Execute

    Run tests against authorised environments or evaluate AI workflows and connected repositories.

  5. 05

    Review

    Update results against test cases and prepare evidence-backed draft defects for human approval.

  6. 06

    Release

    Retest fixes, run regression checks and produce a release-readiness evidence pack.

TEST COVERAGE

Designed around the application

Testing is selected according to the product, its risks and the organisation's agreed requirements.

WEB & BUSINESS APPLICATIONS

Software testing

  • Functional and regression testing
  • Integration and API testing
  • Smoke and end-to-end testing
  • UI and accessibility testing
  • Business-process validation
AGENTS, CHATBOTS & RAG

AI evaluation and testing

  • Golden datasets and expected answers
  • Adversarial and edge-case scenarios
  • Factual correctness and faithfulness
  • Groundedness, relevance and citation accuracy
  • Tool-call and workflow-completion accuracy
  • Prompt injection and red-team security tests
HUMAN IN THE LOOP

Automation with accountable decisions

When a test fails or an anomaly is identified, the agent is designed to prepare a draft defect with the expected and actual results, available screenshots or logs, reproduction details and a suggested severity.

A human reviewer can approve, amend or reject the draft before it is formally logged and triaged. Human review also remains part of risk acceptance and release approval.

RETESTING & RELEASE READINESS

A traceable evidence pack for every release

After fixes are delivered, the planned agent workflow can retest affected areas, run regression suites, repeat AI quality and safety evaluations, and compare results with agreed release criteria.

PLANNED OUTPUT

Release-readiness evidence

  • Tests executed and their results
  • AI evaluation scores and evidence links
  • Open risks and accepted exceptions
  • Agreed scope and release criteria
  • Human approvals and decisions

Important: Where requested, the evidence pack may include a release confidence certificate confirming completion of the agreed testing scope. It supports an organisation's release decision but does not replace human approval, regulatory assessment or independent certification.

BOOK A DEMONSTRATION

See how the workflow could support your testing process.

Book a 15-minute call

For QA teams and organisations developing web applications, AI agents, chatbots or RAG systems.