Artificial intelligence is moving into the quality department. The question is no longer whether AI can analyse quality data, but whether it can help organizations prevent problems before they become defects.
By Yelna Yuristiary
For decades, quality management has largely been about finding problems.
An inspector discovers a defect.
An auditor identifies a nonconformity.
A customer submits a complaint.
An engineer investigates the root cause.
A corrective action is then introduced.
This familiar cycle has helped organizations control quality for generations.
But artificial intelligence is beginning to change the sequence.
Instead of waiting for a defect to appear, organizations are increasingly exploring whether data can be used to identify patterns, anomalies and potential failures before they occur.
The idea sounds simple.
If a company already has thousands—or millions—of quality records, could an AI system recognize the warning signs that humans might miss?
That question is becoming increasingly relevant as organizations move toward digital Quality Management Systems.
And in 2026, the discussion is no longer purely theoretical.
The ISO 9001 Auditing Practices Group published a paper in June 2026 specifically addressing auditing a QMS that uses Artificial Intelligence.
At the same time, ISO has published ISO 9001:2026, an updated edition of its quality management standard that places greater emphasis on strategic decision-making, risks and opportunities, leadership, quality culture and the changing digital business environment.
The result is a new question for quality professionals:
What happens when the Quality Management System can learn from its own data?
From inspection to prediction
Traditional quality management often follows a familiar pattern:
Process → Inspection → Defect → Analysis → Corrective Action
AI potentially introduces another layer:
Process → Data → Pattern Detection → Prediction → Preventive Action
This does not mean that AI eliminates inspection.
Instead, it can change where organizations focus their attention.
Imagine a construction project producing thousands of inspection records.
The records might contain:
concrete test results;
temperature measurements;
slump results;
curing records;
material test certificates;
inspection requests;
NCRs;
weather conditions;
subcontractor performance;
equipment information;
rework records.
Individually, each piece of information may appear relatively ordinary.
But when these data are analysed together, patterns may emerge.
For example, an AI model might identify that certain combinations of:
high temperature + delayed concrete placement + particular supplier + specific curing conditions
are associated with a higher frequency of later quality problems.
The engineer still needs to investigate.
But the system has provided an early warning.
That is the fundamental difference between reactive quality management and predictive quality management.
AI does not "know" quality
There is an important misconception to avoid.
AI does not automatically understand what quality means.
An AI system learns patterns from the data and objectives it is given.
If the underlying data are incomplete, inaccurate or biased, the output can also be unreliable.
This makes data quality a central issue.
ISO/IEC 5259-4:2024, for example, provides a standardized process framework for managing data quality in analytics and machine learning, including issues such as data labelling, evaluation and lifecycle management.
This creates an interesting paradox:
The future of AI-driven quality management depends on the quality of the data used to manage quality.
Poor data can produce poor predictions.
And a sophisticated algorithm cannot automatically compensate for fundamentally unreliable information.
The new Quality Data Chain
A useful way to understand AI in quality management is to think of five stages.
1. Data collection
The organization collects information from its operations.
Examples include:
inspection results;
test results;
equipment data;
customer complaints;
audit findings;
NCRs;
supplier performance;
environmental conditions.
2. Data quality
The organization verifies whether the data are reliable.
Questions include:
Is the data complete?
Is it accurate?
Is it consistent?
Is it traceable?
Was it collected correctly?
3. Data analysis
AI or advanced analytics searches for relationships and patterns.
4. Prediction
The system identifies situations associated with increased probability of a quality problem.
5. Human decision
An engineer, manager or quality professional evaluates the information and decides what action is appropriate.
This final stage remains critical.
AI can support a decision.
It should not automatically become the organization’s unquestioned decision-maker.
From NCRs to early warnings
Consider the traditional NCR.
A nonconformance report tells the organization:
Something has gone wrong.
That is valuable information.
But it is also historical information.
An AI-enabled QMS could potentially identify a pattern before an NCR occurs.
For example:
| Traditional QMS | AI-enabled QMS |
|---|---|
| NCR occurs | Warning generated |
| Defect identified | Anomaly identified |
| Root cause investigated | Risk pattern detected |
| Corrective action | Preventive intervention |
| Historical analysis | Predictive analysis |
| Periodic reporting | Continuous monitoring |
The objective is not to eliminate NCRs completely.
That would be unrealistic.
The objective is to increase the organization's ability to detect emerging problems earlier.
Can AI predict construction defects?
Construction provides an interesting test case.
Unlike manufacturing, construction projects are highly variable.
Projects differ in:
location;
weather;
materials;
design;
contractors;
subcontractors;
workforce;
equipment;
sequencing;
project management;
site conditions.
This makes prediction challenging.
But it also creates enormous amounts of potentially useful data.
Imagine a project where the quality team has historical data from 20 similar projects.
The database contains thousands of observations about:
concrete;
steel reinforcement;
waterproofing;
finishing;
earthworks;
pavement;
marine works;
suppliers;
inspections;
defects;
rework.
An AI model could potentially identify relationships between project conditions and quality outcomes.
For example:
Which conditions are associated with higher waterproofing defect rates?
Which suppliers show recurring quality patterns?
Which work packages generate the greatest rework risk?
Which inspection findings tend to precede major defects?
The answers would still require engineering judgement.
But the data could help direct that judgement.
The same idea applies to manufacturing
Manufacturing may provide an even more structured environment for predictive quality.
A production system can generate continuous information about:
temperature;
pressure;
vibration;
machine speed;
production cycle time;
material properties;
equipment condition;
product dimensions.
AI can analyse these variables to identify abnormal patterns.
Suppose a machine normally operates within a certain range.
A traditional system might wait until a product fails inspection.
A predictive system could detect a gradual change in machine behaviour and alert the quality or maintenance team.
The intervention may happen before the defective product is produced.
This is where quality management and predictive maintenance begin to overlap.
What about healthcare, laboratories and services?
AI-driven quality management is not limited to physical products.
Hospitals, laboratories, universities, financial institutions and service organizations also produce large amounts of operational data.
For example, a service organization might analyse:
complaint patterns;
response times;
transaction errors;
customer behaviour;
employee workload;
service interruptions.
The "defect" may not be a cracked component.
It may be:
a delayed service, an incorrect transaction, a missed requirement or a dissatisfied customer.
The principle remains similar:
detect → analyse → predict → intervene → learn.
AI and the Quality Auditor
One of the most interesting developments is what AI means for auditors.
The ISO 9001 Auditing Practices Group's June 2026 publication on auditing a QMS that uses AI is an indication that AI is becoming a practical issue for QMS auditing.
The auditor may increasingly need to ask questions such as:
What AI system is being used?
What quality decision does it support?
What data does it use?
How was the system validated?
How is its performance monitored?
What happens when the AI produces an incorrect result?
Who has authority to override it?
These questions introduce a new dimension to quality auditing.
The auditor is no longer examining only human-designed procedures.
The auditor may also need to understand AI-supported processes and controls.
The danger of trusting the algorithm
There is another side to the AI revolution.
An AI prediction can be wrong.
This is particularly important when the prediction affects safety, compliance, customer requirements or major financial decisions.
Consider a system that predicts a low probability of a concrete defect.
Should the project team skip the inspection?
No.
The prediction is information—not proof.
This distinction is fundamental.
AI can identify a pattern.
An engineer must still determine whether the pattern is meaningful in the actual context.
This is why the future of quality management is unlikely to be simply:
Human versus AI.
It is more likely to be:
Human + AI.
The importance of explainability
Another challenge is understanding why an AI system produced a particular prediction.
Imagine a dashboard that says:
HIGH QUALITY RISK
A Quality Manager may immediately ask:
Why?
If the system cannot provide useful information about the factors contributing to the prediction, it becomes difficult to use responsibly.
For quality management, an AI system should therefore ideally provide more than a score.
It should help users understand:
what changed;
which variables matter;
what historical pattern was detected;
how confident the prediction is;
what evidence supports the warning.
This is especially important when the output is used in an audit, engineering decision or corrective action.
AI creates a new quality problem: AI itself must be controlled
There is a fascinating implication here.
If AI becomes part of the QMS, then AI becomes part of the quality system that itself needs to be managed.
Organizations may need to consider:
data quality;
model performance;
validation;
access control;
cybersecurity;
version control;
human oversight;
monitoring;
changes to the AI model;
error handling;
records and traceability.
In other words:
The organization cannot simply install AI and assume that quality will improve.
The AI application itself becomes another process that requires appropriate controls.
ISO 9001:2026 and the AI era
ISO 9001:2026 does not turn ISO 9001 into an AI standard.
The standard remains a general quality management system standard applicable across sectors. ISO describes the 2026 edition as an evolution of the established framework, with greater emphasis on leadership, quality culture, strategic alignment and risks and opportunities.
But these principles provide a useful environment for discussing AI.
For example:
Leadership
Management needs to understand how AI affects quality decisions.
Risk and opportunity
AI can introduce new risks while creating opportunities for better prediction and efficiency.
Competence and awareness
Employees need appropriate understanding of AI-supported processes.
Performance evaluation
Organizations need to determine whether AI actually improves process performance.
Improvement
AI models and workflows may need to be continually evaluated and improved.
This means AI can become part of the QMS without replacing the fundamental principles of quality management.
The rise of "Quality Intelligence"
A new concept may emerge from this transformation:
Quality Intelligence
Traditional Quality Management asks:
What happened?
Quality Intelligence asks:
Why did it happen?
Predictive Quality asks:
What is likely to happen next?
Prescriptive Quality asks:
What should we do about it?
This creates a possible evolution:
Quality Control
↓
Quality Assurance
↓
Quality Management
↓
Data-Driven Quality
↓
Predictive Quality
↓
Quality Intelligence
The boundaries between these stages are not absolute.
But they illustrate the direction in which digital quality management is moving.
What should Quality Managers do now?
The first step is not necessarily to purchase an AI platform.
It may be to examine the organization's existing data.
Ask five questions:
1. What quality data do we already have?
Many organizations have more data than they realize.
2. Is the data reliable?
AI cannot solve fundamentally poor data quality.
3. What quality problems occur repeatedly?
Recurring problems are particularly useful candidates for analytical approaches.
4. Can we identify leading indicators?
Look for signals that appear before the final defect.
5. Where should humans remain in control?
Define which decisions AI can support and which decisions require professional judgement.
These questions can provide a more realistic starting point than simply asking:
"How can we use ChatGPT in Quality?"
The future quality professional
The role of the Quality Manager may gradually change.
The traditional skill set includes:
standards;
auditing;
inspection;
corrective action;
root-cause analysis;
process control.
The emerging skill set may additionally include:
data literacy;
statistical analysis;
digital QMS;
AI awareness;
predictive analytics;
data governance;
model validation.
The Quality Manager of the future may therefore spend less time manually searching through spreadsheets and more time interpreting information generated by digital systems.
But the core responsibility remains.
Quality decisions still require context, judgement and accountability.
The real question is not whether AI can replace the Quality Manager
The more useful question is:
Can AI help the Quality Manager see problems earlier, understand them better and make better-informed decisions?
That question has a much more practical answer.
AI can potentially help organizations process large volumes of information, identify patterns, detect anomalies and support prediction.
But the effectiveness of an AI-driven QMS depends on the quality of its data, the suitability of its models, the controls surrounding its use and the people responsible for interpreting its outputs.
The technology is therefore only one part of the equation.
The future can be represented simply:
Quality Data + AI + Human Expertise + Governance = Intelligent Quality Management
From detecting defects to preventing them
For much of its history, quality management has been built around a powerful but fundamentally reactive question:
"What went wrong?"
The next generation of quality management may increasingly ask:
"What is beginning to go wrong?"
And eventually:
"What can we do now to prevent it?"
That is where AI could make its greatest contribution to quality management.
Not by replacing the inspector.
Not by replacing the auditor.
Not by replacing the engineer.
But by giving them something they have always needed:
better information, earlier.
The future of Quality Management System may therefore not be about creating a system that simply records failures.
It may be about creating a system that learns from data, detects emerging patterns and helps people prevent failures before they become costly problems.
And that could transform the meaning of quality itself—from finding defects to anticipating them.
