CQI-38: Implementing and Assessing AI Vision Inspection Systems

AI-powered camera systems are increasingly taking on complex inspection tasks in automotive manufacturing. They recognize features, classify defects and evaluate component conditions. To ensure that inspection results remain reliable over time, organizations must systematically control not only the technical equipment, but also AI models, training data, process conditions and subsequent changes.

The CQI-38 “AI Vision Inspection Application and Assessment Guideline” provides a standardized and practical framework for this purpose. Published in June 2026, the AIAG guideline helps organizations plan, implement, validate, monitor and assess AI-powered vision inspection systems throughout their entire lifecycle.

TopQM-Systems supports organizations with the technical interpretation and practical implementation of CQI-38 as well as with the preparation of corresponding assessments.

What Is CQI-38?

CQI-38 is a guideline published by the Automotive Industry Action Group (AIAG) for the application and assessment of AI-based vision inspection systems in automotive manufacturing.

Its full title is:

CQI-38: AI Vision Inspection Application and Assessment Guideline

The first edition was published in June 2026. It is the first AIAG CQI publication dedicated specifically to AI-powered vision inspection systems.

CQI-38 provides a standardized, industry-focused approach that enables organizations to assess and safeguard the capability, accuracy and reliability of AI-based inspection applications. The guideline considers the entire system lifecycle – from planning and data management through system and process acceptance to monitoring, change management and continual improvement.

Is CQI-38 a Standard or a Guideline?

CQI-38 is not an independently certifiable standard. It is an AIAG guideline with a corresponding assessment methodology and workbook.

The guideline was developed to supplement IATF 16949 requirements when AI-powered vision inspection systems are used. It applies fundamental requirements relating to process control, validation, monitoring and measuring resources, data, change management and effectiveness to AI-based vision inspection applications.

The corresponding CQI-38 assessment helps organizations systematically evaluate the maturity and effectiveness of their applications and identify opportunities for improvement.

Which Inspection Systems Does CQI-38 Apply To?

CQI-38 applies to AI-based vision inspection systems that use image data to recognize features, classify defects or evaluate component and process conditions.

Typical applications include:

  • detection of surface defects,
  • completeness checks,
  • identification of missing or incorrectly assembled components,
  • detection of contamination and damage,
  • assessment of welds, coatings or joining connections,
  • classification of different defect types,
  • position, presence and status checks,
  • automated pass/fail decisions.

The guideline focuses on learning and data-driven inspection applications. Traditional rule-based vision inspection and precise dimensional measurement are not its primary focus.

Which Organizations Is CQI-38 Relevant For?

CQI-38 is particularly relevant for:

  • automotive manufacturers,
  • Tier 1 and Tier 2 suppliers,
  • other organizations throughout the automotive supply chain,
  • manufacturers of vision inspection and testing systems,
  • machinery and equipment manufacturers,
  • system integrators,
  • quality management and quality assurance teams,
  • production and process owners,
  • developers and operators of AI inspection applications,
  • internal auditors and supplier auditors.

The guideline applies both to organizations that develop AI-powered vision inspection systems and to companies that purchase, operate or assess these systems at supplier locations.

Why Is a Specific Guideline Needed for AI Vision Inspection?

Traditional vision inspection generally works with predefined features, thresholds and decision rules. AI-powered systems, by contrast, derive their inspection results from trained models and the image data used to develop them.

This creates additional questions and risks:

  • Is the training data complete and representative?
  • Have all relevant defect types been considered?
  • How reliably does the system identify defective components?
  • How are false-positive and false-negative results evaluated?
  • Does the model continue to perform reliably under changing production conditions?
  • Which changes require revalidation?
  • How are data quality, access and cybersecurity controlled?
  • How is system performance monitored over time?

CQI-38 helps organizations address these issues systematically and integrate AI-powered vision inspection into their existing quality management system.

What Does CQI-38 Cover?

CQI-38 follows a total lifecycle approach and considers all essential phases of an AI-powered vision inspection application.

System Planning and Equipment Acceptance

Before purchasing or developing an inspection system, the inspection task, defect characteristics, acceptance criteria, risks and responsibilities must be clearly defined.

The technical equipment must be suitable for the intended application. This includes cameras, lenses, lighting, component positioning, environmental conditions, interfaces and integration into the manufacturing process.

AI Model Training and Validation

The reliability of an AI model depends significantly on the quality of the data used. CQI-38 therefore considers aspects including:

  • selection and representativeness of training data,
  • labeling and classification of image data,
  • inclusion of relevant defect types,
  • separation of training and validation data,
  • definition of appropriate performance metrics,
  • evaluation of incorrect classifications,
  • documented validation and approval of the AI model.

Process Acceptance and Capability Maintenance

In addition to the AI model itself, the application must be assessed under actual manufacturing conditions. Following process acceptance, system performance must be monitored regularly.

Deviations, changes or a gradual deterioration in detection capability must be identified and evaluated at an early stage.

Data Management and Cybersecurity

Image data, model data and inspection results are essential elements of the system. Controlled processes are therefore required for:

  • data quality and data integrity,
  • storage and traceability,
  • access rights,
  • version control,
  • protection against data loss and manipulation,
  • handling of sensitive manufacturing data,
  • cybersecurity risks.

Change Management

Changes to the product, process, equipment or operating environment can affect the reliability of an AI-powered vision inspection system.

This may include changes to:

  • cameras, lenses and lighting,
  • equipment and process parameters,
  • AI models and algorithms,
  • training and reference data,
  • product characteristics and defect types,
  • environmental and operating conditions,
  • software and IT infrastructure.

Relevant changes must be evaluated and documented. Depending on the associated risk, partial or complete revalidation may be required.

Continual Improvement

Findings from incorrect classifications, manufacturing deviations, customer complaints, assessments and process data should be used to continually improve the application and maintain reliable performance.

CQI-38 Assessment and Workbook

CQI-38 includes an assessment methodology and workbook that organizations can use to evaluate the implementation and effectiveness of their AI-powered vision inspection applications in a structured manner.

The CQI-38 assessment can be used for:

  • internal gap assessments,
  • evaluation of new AI inspection applications,
  • preparation for system and process acceptance,
  • periodic effectiveness reviews,
  • evaluation of changes,
  • internal assessments,
  • supplier assessments,
  • identification and tracking of improvement opportunities.

The standardized assessment structure supports the consistent and transparent evaluation of different applications, manufacturing areas and suppliers.

Relationship Between CQI-38 and IATF 16949

CQI-38 does not replace IATF 16949. The guideline was developed to supplement the quality management system when AI-powered vision inspection systems are used.

There are important interfaces with the following areas:

  • risk-based thinking,
  • process development and process approval,
  • monitoring and measuring resources,
  • competence and responsibilities,
  • documented information,
  • change management,
  • monitoring and measurement,
  • traceability,
  • supplier management,
  • control of nonconforming outputs,
  • continual improvement.

Organizations should therefore not treat CQI-38 as an isolated technical document. Its requirements and assessment criteria should be integrated into the processes of the existing quality management system.

What Are the Benefits of Implementing CQI-38?

A structured implementation of CQI-38 helps organizations:

  • identify risks associated with AI inspection systems at an early stage,
  • define clear requirements for equipment and AI models,
  • maintain control over training and validation data,
  • demonstrate the reliability of inspection results,
  • systematically evaluate incorrect classifications,
  • implement changes in a controlled manner,
  • better address customer-specific requirements,
  • prepare effectively for assessments,
  • increase confidence in automated inspection decisions,
  • maintain reliable system performance over time.

CQI-38 in the Official AIAG Store

The English-language first edition of CQI-38 is available directly from AIAG as a printed publication and as an electronic document.

The guideline also includes a downloadable assessment file that organizations can use to systematically evaluate their AI-powered vision inspection applications.

We are an AIAG licensed training and distribution partner in Europe.

Implementing CQI-38 with TopQM-Systems

The implementation of AI-powered vision inspection systems is not merely a technical task. Quality management, production, IT, engineering, purchasing and supplier management must work together to ensure that requirements, data, responsibilities and evidence are consistently controlled.

TopQM-Systems supports organizations in integrating CQI-38 into their existing processes and management systems.

Depending on the organization’s requirements, our support includes:

  • technical interpretation of CQI-38,
  • analysis of existing AI-powered vision inspection applications,
  • gap analysis against the guideline,
  • definition of processes and responsibilities,
  • integration into the existing quality management system,
  • preparation and support for assessments,
  • supplier assessment and supplier development,
  • action planning and implementation support,
  • qualification of the employees involved.

As an official AIAG partner, TopQM-Systems combines the requirements of AIAG publications with practical experience in quality management, automotive processes, audits and supplier development.

REQUEST CQI-38 SUPPORT

Frequently Asked Questions About CQI-38

What does CQI-38 mean?

CQI-38 is the designation of the AIAG publication “AI Vision Inspection Application and Assessment Guideline.” It provides an industry-focused approach for applying, managing and assessing AI-powered vision inspection systems in automotive manufacturing.

Has CQI-38 already been published?

Yes. The first edition of CQI-38 was published in June 2026.

Is CQI-38 mandatory?

CQI-38 is not a legal requirement or an independently certifiable standard. However, its application may become mandatory through customer-specific requirements, contracts or supply chain requirements. Organizations may also apply the guideline voluntarily to systematically safeguard their AI-powered inspection systems.

Does CQI-38 supplement IATF 16949?

Yes. CQI-38 was developed to supplement IATF 16949. The guideline applies existing quality management requirements to the specific characteristics and risks of AI-based vision inspection systems.

What is the difference between AI vision inspection and traditional machine vision?

Traditional machine vision generally uses predefined rules, features and thresholds. AI-powered vision inspection systems learn from image data and trained models. This results in additional requirements for data quality, validation, model changes and the ongoing monitoring of detection performance.

Does CQI-38 include an assessment?

Yes. CQI-38 includes an assessment methodology and workbook. Organizations can use them to systematically evaluate the effectiveness and control of their AI-powered vision inspection applications.

Who should use CQI-38?

CQI-38 is intended for automotive manufacturers, suppliers at all tiers, vision inspection equipment manufacturers, machinery and equipment manufacturers, and system integrators. It is relevant to organizations that develop, purchase, operate or assess AI-powered vision inspection systems.