Author: Michelle Unger
Keeping Humans in the Loop: The Real Value of AI-Human Collaboration in ILI Data Analysis
In a nutshell:
Artificial intelligence (AI) is rapidly transforming the way we access information and perform tasks both in our professional and personal lives. From quickly researching topics to receiving tailored recommendations, whether it’s the next movie to watch, a travel destination, or even what to cook for dinner – AI has become an integral part of our everyday routines. While these advancements bring significant convenience and efficiency, they also raise an ongoing debate: to what extent might AI replace human roles?
This article offers a closer look at that question in the context of pipeline integrity. Drawing on insights from conversations Michelle Unger, VP Industry Affairs, held with technical experts closely involved in ROSEN’s AI initiatives, it explores how AI is reshaping in-line inspection data analysis, while keeping human expertise at the center. As part of a broader series on the evolving role of AI in integrity engineering, it highlights how human-AI collaboration is enabling more reliable results and ultimately better decisions.
In the workplace, AI is fundamentally changing the way we work. In our industry, for example, inspections and integrity assessments have traditionally relied heavily on periodic checks, manual data analysis, and the experience and judgment of analysts and engineers. Today, AI enables continuous monitoring, automated anomaly detection, and data‑driven insights that enhance the analysis of inspection data, which in turn supports faster and more informed decision‑making in integrity engineering. Rather than replacing analysts, AI is increasingly positioned as a tool that enhances their capabilities in an increasingly complex and highly safety-critical industry.
Insights from conversations with technical experts involved in ROSEN’s AI initiatives point to a clear conclusion: AI strengthens data analysis by making it more consistent and faster – and ultimately more human‑centered, by freeing experts from routine processes to focus on customer interaction and consultancy. ROSEN is uniquely positioned to lead this development. With decades of integrity engineering experience, access to extensive inspection data, and a deeply rooted safety and validation mindset, the company can ensure that AI solutions are not only innovative, but also reliable, transparent, and grounded in real-world operations.
From noise to signal
One of the most immediate and tangible benefits of AI is its ability to remove noise from ILI data. Historically, analysts spent a significant portion of their time cleaning, filtering and pre‑classifying signals, which is a necessary but repetitive task.
This is already changing in day-to-day analysis. As AI takes over much of the preparation work, the focus is shifting towards interpretation, engineering judgment and customer dialogue.
This shift does not make the process less conservative. Instead, it makes it faster to reach the same level of conservatism. Operators receive the critical information earlier, enabling them to act sooner and improve safety margins. AI is not changing the threshold of what is considered safe, it is shortening the time it takes to get there.
AI is already transforming how we handle inspection data. Tasks like data preparation and pre-classification, which used to take a significant amount of manual effort, are now largely automated. What this means in practice is that our analysts can spend far more time on what really matters – interpreting results, applying engineering judgment, and engaging with customers.
Better pictures, better decisions
Beyond efficiency, AI is enabling richer, more accurate representations of pipeline conditions. Through integrating data from multiple inspections, such as through ROSEN MFL Data Fusion and advanced modelling, engineers can now derive profiles and near-real geometries rather than relying solely on conservative “box” approximations of metal loss. This reduces uncertainty and supports more precise engineering decisions.
AI also makes possible what was previously impractical: large‑scale cross‑technology correlation. For example, checking whether a crack‑like signal aligns with deformation or corrosion across multiple datasets is something humans could only do manually in isolated cases. AI can do it systematically and at scale.
This does not mean taking bigger risks. It means reducing uncertainty with better information. Marc Fischer noted, “We can create a full picture of an anomaly by layering technologies – and that reduces uncertainty.” Better pictures lead to better decisions, and better decisions lead to safer pipelines.
While AI is a powerful tool built on existing data, it is important to recognize that it relies on human oversight to ensure that its results are meaningful and contextually sound. In complex, evolving environments influenced by urban expansion, infrastructure changes, climate impacts, and societal and political factors, human experience and foresight are essential for understanding current and future threats to infrastructure.
Trust, validation, and lifecycle
Every interview reinforced the same principle: trust is earned through validation. AI models used for in-line inspection data analysis are not black boxes; they are engineered systems with documented performance, clear deployment criteria and ongoing monitoring.
ROSEN’s validation process includes:
- rigorous performance analysis
- outlier detection
- independent review
- rejection and retraining when needed
- documented model‑deployment reports
This structured approach reflects a broader understanding of AI – not as a one-time deployment, but as a continuously managed system.
For us, AI is not a ‘train once and deploy’ exercise. Every model goes through a structured lifecycle – from development and validation to continuous monitoring and improvement. We deliberately design our systems so that there is always a human in the loop. AI supports decision-making, but it does not replace accountability.
This discipline is backed by a unique advantage: data. With more than 26,000 in-line inspections in the database and extensive verification (including laser‑scan and computed tomography data), ROSEN has one of the most comprehensive pipeline integrity training datasets in the industry. This breadth of essential variable coverage is a competitive differentiator and a foundation for trustworthy AI.
Lifecycle management is equally important. Models are monitored, refreshed and reassessed as new data arrives. AI is treated as a living system, not a static tool. This approach mirrors the company’s broader engineering philosophy: safety is not a one‑off achievement; it is a continuous process of innovation and improvements.
The next step in this evolution is auditability. By aligning AI validation with the same engineering rigor applied in areas such as test center qualification, ROSEN is preparing for a future in which AI systems are assessed with the same level of scrutiny as traditional technologies – and helping define what that standard should look like.
At this stage, there is limited guidance and formality on auditing AI systems in our industry. That is not a limitation – it is an opportunity to set the benchmark. Our goal is to align AI validation with the same rigorous engineering principles we apply in other areas, such as test center qualification. By doing so, we are preparing not only for future regulatory expectations but also reinforcing confidence in how these technologies are applied.
People first, machines second
AI is often framed as a threat to human roles. Inside ROSEN, the opposite is true. AI is deliberately designed to augment, not replace, human expertise.
Stephan Eule explained the philosophy clearly: “AI decision systems in in-line inspection are always designed around a human in the loop, supporting the assessment rather than automating it.”
AI handles the repetitive, monotonous tasks, the “box‑checking” that drains time and attention. Analysts, in turn, spend more time on interpretation, problem‑solving and customer engagement. This makes the work more meaningful and accelerates skill development. This shift is not only good for analysts, it is good for the industry. A more engaged, more skilled workforce produces higher‑quality evaluations and more confident decisions.
Our core objective remains unchanged: keeping pipelines safe and reliable. AI helps us by taking over repetitive and time-intensive tasks, allowing engineers to focus on higher-value activities like analysis, interpretation, and decision-making. In the end, this leads to better outcomes – for our teams and for our customers – because expertise is applied where it makes the greatest impact.
The key message
ROSEN’s approach is grounded in engineering discipline, validated with real data and shaped by the people who use these tools every day. AI makes:
- Integrity engineering more reliable, not riskier.
- Decisions faster, not less conservative.
- Engineers more effective, not redundant.
- Work more meaningful, not more monotonous.
Operators can trust that ROSEN is deploying AI responsibly, with transparency, human oversight and a commitment to continuous improvement.
Our industry is defined by assets that can operate for over a century, requiring humans that are uniquely capable of identifying anomalies, applying judgment, and imagining future scenarios that extend beyond available data sets. AI will enhance our ability to model and evaluate potential outcomes, but it is ultimately engineers who must interpret these insights, anticipate future conditions, and develop effective risk management strategies.
Looking ahead
AI in data evaluation is not a single product but a platform of practices: data stewardship, model governance, process orchestration and human training. The near term will be about pilots that demonstrate safety and value, and about working with customers and regulators to define audit‑ready validation approaches. The longer term will see richer fusion of technologies, more prescriptive decision support and a workforce that blends domain expertise with data literacy.
The direction is clear: AI will play an increasingly central role in evaluation of in-line inspection data. But it will do so in a way that strengthens, not replaces human judgement.
Because in this industry, trust is earned. And AI must earn it too.
Michelle Unger
Vice President Industrial Affairs, ROSEN Group
Michelle has over 25 years of experience in engineering consultancy, leadership development, and competency strategy in the pipeline industry.
With a focus on aligning technical expertise and organizational performance, Michelle supports the development of future-ready workforces through designing programs, leading certification initiatives, and translating complex technical roles into measurable competencies.
Michelle leverages her extensive international experience to collaborate with organizations across the pipeline industry, fostering leadership capability and integrating people-driven approaches to technical and operational excellence.