AI Testing Services
Four Agents Embedded in the Full Testing Workflow
From requirement analysis, test case generation and script generation to automated execution and Bug analysis, ZD Technology embeds four agents into the entire testing workflow to make testing faster, more accurate and less labor-intensive.
* Data measured on smart cockpit testing projects; actual benefits vary by project scenario.
Service Overview
Facing persistent challenges such as rapid feature iteration, heavy workloads in test case and script writing, and time-consuming Bug localization and fix verification, ZD Technology runs four agents - test case generation, test script generation, test execution and Bug analysis - through the entire process from requirement analysis to Bug root-cause analysis and fix recommendations, helping customers improve testing efficiency and delivery quality. The solution has been implemented in real smart cockpit testing projects and can be generalized to vehicle network, function integration, OTA and other testing scenarios, with strong cross-project reusability.
Industry Challenges We Solve
The capacity bottleneck of a test team often lies not in execution itself, but in the labor-intensive stages of design, scripting and defect handling.
Fast requirement iteration, heavy case and script workloads
Feature versions update frequently, and test design and script writing consume large amounts of manpower, becoming a bottleneck for delivery pace.
High maintenance cost of automated scripts
When UI elements or the runtime environment change, script locators easily break, and maintenance effort keeps accumulating.
Bug localization depends on individual experience
Problem analysis requires combining descriptions, logs, screen recordings and source code; manual investigation is time-consuming and conclusions are hard to replicate.
Test assets are hard to accumulate
Test cases, scripts and analysis experience are scattered across individuals and single projects, making it hard to form organizational assets reusable across projects.
Four Testing Agents
A capability matrix built around design → scripting → execution → defect handling; each agent can be used independently or chained into a complete closed loop.
Test case generation agent
Parses requirement documents, identifies test points and breaks down steps to generate structured test cases in batches.
Output: UI / functional / interface / performance test cases with IDs
Test Script Generation Agent
Generates automated scripts based on test case content and environment characteristics, and continuously refines scripts and locating strategies using execution feedback.
Output: Executable automated test scripts
Test Execution Agent
Computes execution steps automatically according to the current GUI/VUI environment and judges results, supporting automated execution of click input and voice interaction.
Output: Execution process records, logs / videos / test results
Bug analysis agent
Triggered asynchronously by the defect platform; after aggregating multi-source information it localizes the cause and distills recommendations, then writes the results back to the defect system automatically.
Output: Problem attribution, fix recommendations
Measured Results
The following data was measured on smart cockpit testing projects.
← Swipe left or right to view the full content →
| Stage | Benefit |
|---|---|
| Case generation | Large-scale generation per project: 1000 test cases; test design cycle shortened from 5 person-days to 2 person-days |
| Script Generation | Generates 100 automated scripts; script writing cycle shortened from 7 person-days to 2 person-days |
| Test Execution | Automated execution supporting GUI click input and VUI voice interaction, with automatic recording of process and results |
| Bug Analysis | Bug Analysis Efficiency Improvement 60% or more; analysis conclusions and fix recommendations are written back to the defect system automatically |
| Voice Interaction | Voice interaction test coverage 90.3%, test time reduced by 60% |
Design Stage
Reduces the time cost of manually decomposing requirements and writing test cases.
Development Stage
Automatically generates script skeletons to improve automation coverage efficiency.
Execution Stage
Perceives the environment automatically and completes test execution and result recording.
Defect Stage
Analyzes problems by combining multi-source information to support rapid localization and fixes.
Technology Foundation and Knowledge Assets
← Swipe left or right to view the full content →
| Dimension | Description |
|---|---|
| Large Model Capabilities | Built on mainstream open-source large models and code-generation large models, flexibly selectable and replaceable according to project and compliance requirements |
| Platform and Devices | Cloud services + agent platform; the central control device adopts a central control unit + in-house software and algorithms, customizable per project |
| Input Formats | Requirement documents, test cases, GUI / VUI interfaces, XML, images, logs, screen recordings, source code |
| Output Formats | Test cases, automated test scripts, execution results and process records, problem attribution and fix recommendations |
| Knowledge Assets | Smart cockpit test case library, automated script assets, Bug analysis knowledge base; agent outputs are recorded into the knowledge base automatically, empowering subsequent cockpit and whole-vehicle testing projects |
Our Strengths
01Proven in Real Projects
The four agents run in real smart cockpit testing projects and deliver quantifiable benefits - not a concept demo.
02Closed-Loop Capability, Not a Point Tool
Script generation goes beyond writing code: it forms a generate → execute → refine closed loop, in which execution feedback feeds back into script tuning and locating-strategy optimization.
03Reusable Knowledge Assets
Test cases, scripts and analysis capabilities are stored into the library automatically and reused across projects, empowering subsequent cockpit and whole-vehicle testing projects.
04Generalizable Scenarios
The same agent capabilities can be transferred to vehicle network, function integration, OTA and other testing scenarios.
05Delivered Together with Existing Testing Services
Agent capabilities are embedded in ZD's testing service delivery process and can be combined with on-site and automated testing services.