How GENI Works

GENI is not a single AI model — it is a stack of AI technologies working together. GENI uses multiple specially tuned transformer models and a proprietary code generator to convert English test cases into executable scripts.

Note: GENI Script Generation does not use Blueprint / Digital Twin technology. The Digital Twin (formerly “Blueprint”) is built by AISG, a separate module — see AI Script Generation (Autonomous Testing) for details.

The GENI process

A business analyst writes step-by-step test cases in plain English — the same way manual testers document them. GENI’s transformer models understand the natural language, page layouts, and visual cues in those steps. A proprietary code generator produces executable test scripts — in bulk, hundreds of test cases at a time.

Two distinct phases

Phase

What happens

Characteristics

1. Generation

GENI reads each English test case, resolves the elements and actions it describes, and writes the script.

Compute-heavy. Runs each test case in parallel on its own isolated instance. Batch runtime is roughly constant (~2 hours) whether you submit 1 test case or 100.

2. Execution

The finished script is played back against the application — no LLM involved, pure playback.

Speed is bounded by how fast the application responds. Scripts run on demand or on a schedule, inside AIQ or in your CI/CD.

Because every test case runs in parallel on its own machine instance, capacity scales linearly: running 100 test cases takes no more wall-clock time than running one. That constant-time behavior is what makes batch onboarding practical — upload your existing manual suite, and the entire set converts in a single pass.

Inputs: what GENI needs from you

You can bring test cases in two ways: an Excel file, or an integration with a test management system (TMS). TestRail is supported today, with Azure and other standard TMS integrations rolling out. GENI maps the columns, validates the batch, and reports exactly how many test cases are ready before you generate.

Excel format

Column

Required?

Purpose

Step Description

Yes — always

What the tester does in this step.

This is what GENI interprets.

Test Data

Yes — always

The actual data entered: URLs, usernames, passwords, inputs. GENI needs at least a URL and a step description, even for a single case.

Test Script Name

Yes, if more than one test case

Names each resulting script.

Expected Result

Optional

Your validations. Leave the column blank if you have none.

Test Case ID / Name / Preconditions / Step Number

Optional (metadata)

Helpful context; auto-mapped when present.

Tip: column mapping is auto-detected and can be overridden manually. After mapping, GENI shows the count of ready test cases (for example, “Test cases ready: 20”) before you click Transform Test Scripts.

From Excel to script

Once the batch is mapped and validated, a single action — Transform Test Scripts — starts the generation. Each test case is processed in parallel, and progress is visible in the GENI UI with per-script status.

Status

Meaning

Pending

(yellow)

Queued, waiting for an available instance.

Running

(blue)

Actively being translated into a script.

Failed

(red)

Generation could not resolve this case — re-run it, or adjust the test case.

Completed

(green)

Script generated and ready to execute.

Drilling into a completed entry shows when it started, the original steps pulled from the spreadsheet, the expected results used as validations, and the test data applied. Finished scripts appear in the Test Scripts area, where you can open any script to inspect its steps and its execution history.

Execution, scheduling, and reporting

Execution evidence. Each execution captures a screenshot of every state the script visits — every page, every filled form, every action — and, on failure, a screenshot of the page where it failed. Screenshots are expandable, so a failed run is diagnosable at a glance: see the step, the data, and the exact screen state.

Scenarios: built-in CI/CD. Generated scripts are grouped into scenarios. A scenario behaves like a CI/CD pipeline built into the GENI platform: add the scripts you want, then run the scenario manually or schedule it — daily, weekly, monthly, or on an advanced cadence, in any time zone.

Reports and dashboards:

Download execution reports per scenario. Export results to the built-in open-source dashboard, or import the data into the dashboard your team already uses — designed for easy CI/CD integration. Use the built-in Allure Report for rich, shareable execution reporting.

Did you know? GENI harnesses Appvance’s two million lines of code and deep application ML/AI models. Per CEO Kevin Surace: true AI — deep AI harnessing multiple transformer models and LLMs — makes GENI the first 100% AI-first QA platform that dispenses with legacy scripting and tools.

Data privacy

The LLM is invoked only during the generation phase. Execution is pure script playback — no AI inference, and no data leaving for model calls. Customer data is encrypted at rest and in transit; stored data remains encrypted even to infrastructure access. GENI models are private and fixed — they do not learn from your application or data. No data is shared between clients, and no PII is stored. Generated scripts are yours: downloadable (Playwright / Java / JavaScript) and editable, for portability into any pipeline. All models run on a GPU-centric AI server, with on-prem and cloud deployment options.

Sources: AIQ AI and Data Privacy (docs.appvance.net) · Appvance announces GENI (Mar 18, 2025) — press release · GENI Script Generation — Technical Overview

Last updated: September 10, 2026