Certifying the Human in the Age of the Algorithm

Nondestructive testing (NDT) certification has always rested on a fundamental premise: a trained, qualified human makes the call. The technician reads the data, applies judgment, and puts their name on the result. That premise isn’t disappearing—but it is being tested.

As artificial intelligence (AI) and machine learning (ML) tools move deeper into inspection workflows, the industry faces a question that certification bodies cannot afford to sidestep: What does it mean to qualify and certify a technician when an algorithm increasingly assists—or performs—the interpretation?

This isn’t a theoretical exercise. NDT professionals in the field are already encountering AI-assisted systems. The certification frameworks governing their qualifications were not designed with those tools in mind. Closing that gap is among the most consequential challenges ASNT faces today.

A Pivotal January

The industry signaled loudly in January 2026 that it understands the urgency. Two significant events unfolded in the same week, in the same city, and together they mark a turning point.

NDT Week, held 18–23 January at the American Welding Society (AWS) headquarters in Miami, Florida, convened ASNT, ASTM, and AWS—the industry’s three leading standards organizations— under one roof for focused, collaborative work that will impact the industry. Committees including the ASTM E07, ASNT Certification Management Council, ASNT Standards Council, and AWS Standards groups met in person to begin building the kind of integrated framework that the pace of technological change demands. Among the outcomes: a second public comment period has opened on a new standard addressing AI and ML in NDT and nondestructive evaluation (NDE) applications—a concrete step the industry should engage with.

That same week, ASNT convened the inaugural Thought Leaders Summit—an invitation-only gathering of senior executives from across the NDT ecosystem. Conducted under Chatham House Rules, the gathering was designed not as a technical conference but as a facilitated peer dialogue. Leaders from organizations that use and supply NDT products and services spoke candidly about workforce disruption, AI, and economic uncertainty. As ASNT CEO Neal Couture noted, the goal was “to create space for meaningful conversation among leaders navigating these issues in real time.” What those conversations reinforced is that the industry is asking the same questions from boardrooms to job sites—and that the answers will require collaboration, not just individual organizational initiative.

The industry has moved from asking whether to address AI to asking how to address it. That shift matters.

The Certification Gap

Here is the core problem: current qualification and certification frameworks were built on the assumption that the technician is the primary interpreter of the data. The technician performs a test on the component, evaluates the indications, and renders a judgment. AI changes that assumption—not by eliminating the technician, but by inserting a layer between the data and the decision.

As AI tools process inspection data, the examiner’s role shifts toward interpreting advanced analytics, validating algorithm outputs, and making critical safety and quality decisions based on information they did not generate themselves. That is a meaningfully different skill set from what today’s certification schemes assess. It demands new competencies: the ability to interrogate an algorithm’s output, understand its constraints and failure modes, recognize when results warrant skepticism, and know when to override the system entirely.

None of that is straightforward to certify. Traditional NDT qualification has always involved demonstrating a specific technical skill—for example, interpreting a radiographic image, conducting an ultrasonic scan, or evaluating indications from a magnetic particle examination. AI literacy and algorithmic oversight are harder to define, test, and standardize across methods and sectors. That difficulty is not a reason to defer the work. It is the work.

There is also a liability dimension the industry has not fully reckoned with. When an AI system contributes to an inspection call that proves incorrect, where does the responsibility reside? With the algorithm developer? The equipment supplier? The technician who accepted the output? The certification scheme that credentialed that technician? These questions are coming, and they will be answered one way or another— either by deliberate frameworks built in advance through industry collaboration, or by litigation absorbed afterward.

What Certification Services Is Thinking

Let me be direct about where ASNT stands. Our role is not to certify AI systems. Validating algorithms, auditing training datasets, and establishing performance benchmarks for AI models—that work belongs to standards bodies and developers. ASNT’s role is to certify the humans and organizations who use those systems responsibly.

That means we are actively examining what competency looks like in an AI-assisted inspection environment. Some of this may be addressed through updated exam content—adding AI literacy components to existing certification schemes. But we are also asking a deeper question: Do we need new certification pathways designed from the ground up for AI-integrated roles? A technician whose primary function is supervising and validating automated inspection systems may require a fundamentally different qualification model than one performing manual contact ultrasonics or other conventional methods.

This is not work ASNT can or should do in isolation. It requires input from employers deploying these tools, from technicians using them, from standards bodies shaping the frameworks, and from the broader membership. The Thought Leaders Summit demonstrated that executives across the industry are navigating these pressures simultaneously. The certification community needs to be part of those conversations—not downstream of them.

Invitation to Engage

The new AI/ML in NDT/E standard is out for public comment, and I encourage every member to engage with it. Standards for AI systems and standards for the people who use them are two distinct bodies of work—but they must develop in dialogue with each other. What standards the community defines as acceptable AI performance in inspection will shape what certification must demand of the technicians overseeing those systems.

ASNT has always been the institution that defines what it means to be a qualified NDT professional. That definition is due for an update. The algorithm is not replacing the technician, but it is changing the job. Our task is to ensure our certification schemes reflect the work technicians are actually being asked to do.

The conversations that began in January are a foundation. ASNT Certification Services will build on them, and we will do so in collaboration with industry. Watch for updates from ASNT in both these areas, standards, and certification development for a new age.