Technology

Autonomous AI review, combined with human-in-the-loop validation, offers a practical pathway to addressing one of the most persistent challenges in project delivery. (Image source: Adobe Stock)

Wassim Ghadban, global SVP, AI & Digital Engineering at Kent, discusses the concept of autonomous AI-driven review, combined with human-in-the-loop validation, as a means to remove structural bottlenecks in project delivery, and how the convergence of AI capabilities across engineering and operations can enable earlier production, reduce capital inefficiencies, and redefine how projects are executed

In the delivery of energy projects, the industry has long focused on optimising engineering productivity, refining project controls, and improving procurement and construction strategies. Despite these efforts, delays remain a persistent challenge. These delays are often attributed to complexity, uncertainty, or resource constraints. However, a closer examination reveals that the underlying issue is more structural in nature.

Engineering outputs are developed rapidly, supported by sophisticated tools and experienced teams. Yet, the progression of these outputs through the project lifecycle is governed by validation processes that are inherently sequential and fragmented. Documents and models move through multiple layers of review, often across different disciplines, organisations, and geographies. Each step introduces latency, not necessarily because of the effort required, but because of coordination, alignment, and the need to reconcile inconsistencies.

In this context, the constraint is not the ability to produce engineering work, but the ability to validate and integrate it efficiently.

The nature of the bottleneck

As projects evolve from conceptual design to detailed engineering and construction, the volume and granularity of deliverables increase significantly. This expansion is accompanied by a corresponding increase in interfaces between disciplines. Mechanical systems must align with structural supports, electrical systems must integrate with control architectures, and all must comply with operational and safety requirements.

Traditional review processes address this complexity through iterative, discipline-specific validation. While effective in ensuring technical integrity, these processes are limited in their ability to scale. Reviews are conducted sequentially or in loosely coordinated parallel streams, and issues identified at later stages often require revisiting earlier decisions. The result is a cycle of iteration that extends project timelines and introduces inefficiencies.

This dynamic is further compounded by the fact that review cycles are not purely technical activities. They involve communication, coordination, and decision-making across multiple stakeholders. As such, the duration of a review cycle is often driven more by process than by technical effort.

The emergence of AI in engineering contexts

Advances in AI have reached a level of maturity that allows systems to interpret, analyse, and generate engineering content with increasing accuracy. AI models can process structured and unstructured data, identify relationships between elements, and detect patterns that may not be immediately visible to human reviewers.

In practical terms, AI can now extract information from engineering documents, generate drawings based on defined parameters, and evaluate the consistency of designs across multiple domains. It can compare specifications against standards, identify deviations, and highlight potential conflicts between systems.

These capabilities extend beyond isolated tasks. AI can operate across interconnected datasets, enabling a more integrated view of engineering systems. This creates the foundation for a new approach to validation, one that is not constrained by discipline boundaries or sequential workflows.

Autonomous review with human-in-the-loop validation

The concept of autonomous AI review builds on these capabilities by applying them to the validation process itself. Rather than relying on multiple rounds of manual review, AI systems can evaluate entire engineering packages in a single pass, checking for completeness, compliance, and cross-discipline consistency.

Crucially, this approach does not eliminate the role of the engineer. Instead, it redefines it. Engineers remain responsible for final validation and decision-making, but their role shifts from conducting detailed manual reviews to assessing and confirming the outputs generated by AI. This human-in-the-loop model ensures that accountability and professional judgment are preserved, while significantly reducing the time required for validation.

The impact of this shift is not limited to efficiency. By enabling rapid, comprehensive review, it reduces the likelihood of late-stage issue discovery. Inconsistencies and conflicts can be identified earlier, when they are less costly and easier to resolve.

Implications for project execution

The introduction of autonomous review has broader implications for how projects are executed. By reducing the duration and frequency of review cycles, it allows engineering activities to progress more continuously. Dependencies between disciplines become less restrictive, as validation can occur in parallel rather than in sequence.

This creates the conditions for a more dynamic and responsive execution model. Decisions can be made earlier, with greater confidence, and adjustments can be implemented without significant disruption.

The traditional concept of rigid stage gates begins to give way to a more fluid process, in which validation is embedded within the workflow rather than applied after the fact.
The cumulative effect is a compression of the project timeline. Engineering completion is achieved sooner, procurement can be initiated earlier, and construction activities can proceed with fewer interruptions.

Enabling earlier production

For owner-operators, the most significant consequence of this transformation is the potential to achieve first production earlier than originally planned. The value of this acceleration extends beyond cost savings. It directly impacts revenue generation and the overall economic performance of the asset.

By reducing delays in engineering and execution, autonomous review contributes to a more predictable and shorter path to operational readiness. This not only improves capital efficiency but also enhances the strategic flexibility of operators, allowing them to respond more effectively to market conditions.

In this sense, the primary value of AI in project delivery is not merely in reducing costs, but in unlocking time. Time, in turn, is directly linked to value creation in the form of earlier production and extended asset utilisation.

Extending AI into operations

The same principles that enable autonomous review can be extended into the operational phase. AI systems are increasingly capable of monitoring plant conditions, interpreting sensor data, and applying predefined operational philosophies to optimise performance.

In environments where safety, efficiency, and reliability are critical, AI can support or even execute operational decisions within defined boundaries. This reduces the reliance on manual intervention and enhances the consistency of operations.

In certain scenarios, AI can effectively operate a plant by following established control strategies and responding to real-time conditions. While human oversight remains essential, the role of operators evolves toward supervision and exception management.

Conclusion

Autonomous AI review, combined with human-in-the-loop validation, offers a practical pathway to addressing one of the most persistent challenges in project delivery. By transforming how engineering outputs are validated and integrated, it removes a key constraint that has historically limited project performance.

The resulting benefits extend beyond efficiency gains. They include earlier production, improved capital utilisation, and a more responsive and resilient execution model. When combined with AI-enabled operations, this approach forms the foundation of a new paradigm in which projects are not only digitally enabled but increasingly autonomous.

The future of project delivery will not be defined by how quickly engineering work can be produced, but by how effectively it can be validated, integrated, and executed. In this context, AI is not simply a tool for optimisation, but a catalyst for fundamental change.

Maintaining consistent pig velocity when pipeline flow conditions are variable, constrained or absent is a persistent challenge for operators. (Image source: Expro)

Energy services provider Expro has launched Velonix, an optimised pipeline pig control system that addresses the challenge of maintaining consistent pig velocity by automating and stabilising pigging operations

Pipeline integrity depends on accurate inline inspection (ILI) and effective pigging, both requiring pigs to travel within a critical velocity range. Deviations can cause stalled tools, missed contaminants, or incomplete data - compromising safety and compliance.

Velonix automatically controls pig velocity to help reduce re-runs, improve inspection data quality and enhance safety performance across low, high and no flow pipeline environments, combining state of the art SONAR measurement with automated closed loop flow control. The system integrates three key components: a clamp-on ActiveSONAR meter for continuous direct velocity measurement, computer-controlled throttling valves, and a purpose-built control unit for closed-loop regulation of pipeline flowrates and pressures. Combined with Expro's own Data to Desk platform, operators gain immediate visibility and control from any location or device, allowing for faster and more informed decision making and project visibility. This approach enables accurate velocity control under varying conditions, including high or low flow pipelines and alternative propellant scenarios. The system automatically adjusts pipeline flow through a digitally controlled skid to maintain pigs within the optimal velocity window throughout the run.

Maintaining consistent pig velocity when pipeline flow conditions are variable, constrained or absent is a persistent challenge for operators. By removing reliance on manual adjustment, the system reduces operational uncertainty, helps prevent inspection failures caused by speed excursions and avoids the need for costly and disruptive re runs, enabling customers to complete pipeline integrity campaigns more efficiently and with greater confidence.

“Pig velocity is one of the most critical factors in successful pigging operations, and Velonix provides operators precise, automated control over it,” said Andrei Ion, vice president of Well Flow Management at Expro. “After extension field testing, Velonix has consistently demonstrated its ability to deliver smooth, stable pig runs across a range of pipeline environments, maintaining target velocities, avoiding speed excursions and eliminating the need for costly re runs, increasing the reliability of pipeline intervention through data intelligent services.”

The introduction of Velonix reinforces Expro’s continued investment in intelligent automation and digital technologies to improve the safety and environmental performance of pipeline integrity operations

Now that the focus has moved from what AI can do to how it can be leveraged to create value, oil and gas companies are accelerating the deployment of AI to transform their operations

ADNOC, for example, is deploying the first AI-enabled fully automated walking island rig, which enables faster and more consistent offshore well delivery, while TotalEnergies is deploying agentic AI solutions to improving the detection of fugitive emissions and target the highest-emitting equipment.

One of the undisputed leaders in AI deployment is Aramco. At LEAP 2026 in Riyadh, Ahmad O. Al-Khowaiter, the company's executive vice president Technology & Innovation, discussed how AI is changing the way the company works, explaining that Aramco looks at AI through three lenses: intelligence, trust and scale.

Intelligence

Aramco collects more than 10 billion plus data points from its operations every day, from wells, pipelines, refineries, terminals, and laboratories.

“But this data on its own does not create value,” stressed Ak-Khowaiter.

“That value comes from combining it with huge computing power and world class technical expertise to put into models which help our engineers, operators, and technologists make better decisions.

“This is why we are moving AI deeper into the core of our business.

“We are using it to enhance exploration, optimise production, improve reliability, accelerate engineering, strengthen maintenance, and even keep our people safe.

“It detects anomalies, recommends responses, and automates workflows, empowering our people to focus on higher-value decisions.”

Aramco’s in-house large language models process millions of requests a day and are estimated to have saved hundreds of thousands of working hours every year, he added.

Trust

In the age of agentic AI and autonomous operations, trust is more important than ever. Cyber security, digital ethics and governance all need to be part of the foundations of how AI is developed, deployed, and scaled, al Khowaiter stressed.

“Autonomous systems need strong guardrails.”

Secure cloud environments, resilient connectivity, digital ethics, and AI-enabled monitoring are a focus, with AI-enabled monitoring significantly reducing the time needed to resolve IT service issues.

“Furthermore, our CyberMind security system demonstrates how AI is strengthening cybersecurity at scale, automating millions of investigations while processing over a billion potential threats every year,” Al-Khowaiter said.

“It enhances our resilience, improves response times, and provides the secure digital foundation we need for the next generation of industrial AI to succeed.”

Scale

The real value of AI projects will come from scaling them across enterprises, industries, and ecosystems, Khowaiter said, noting that Aramco is no stranger to scale given it operates some of the largest and most complex industrial systems in the world

“We have the data and we have the infrastructure, but most importantly we have the people. Engineers, scientists, operators, geologists, and technologists who understand the difference between a promising model and a practical solution.

“That’s why AI is not replacing our industrial knowledge, it is amplifying it.”

Al-Khowaiter went on to stress Aramco’s leadership role in AI, with Aramco Ventures now having a portfolio that includes around 300 start-ups from around the world, while the TecShift initiative aims to attract early-stage start-ups from across the world to Dhahran, bringing them together with industrial partners to work on solving real world challenges.

Concluding, Al-Khowaiter said, “The future of industry will inevitably become more intelligent. But it must also be trusted and scalable. That is the opportunity we see at Aramco.

“Combining nearly a century of industrial expertise with cutting edge technology and world-class talent.

“Our innovation has moved to implementation, creating value not just for our company, but for the Kingdom and the world.”

See also: https://oilreviewmiddleeast.com/technical-focus/aramco-highlights-its-ai-leadership-at-world-economic-forum

DUG Insight provides a single workspace where geoscientists can work together. (Image source: DUG)

Margarita Kongawoin, senior vice president of software & HPC at DUG Technology, on how DUG Insight brings the complete geoscience workflow into a single, integrated package — and why that's changing the way teams work

What do you see as the biggest bottleneck in geoscience software today? 

Fragmentation seems to be a major issue slowing geoscience teams down. They often have to work with many different software programs. Moving data between those applications requires data exchange and format conversion, which takes time and can lead to mistakes. When the data reaches the decision-maker’s desk, time (aka money) and attention have been wasted. This fragmented workflow also prevents people from working together and sharing their different skills in a complementary way.

How does DUG Insight overcome those challenges? 

DUG Insight provides a single workspace where geoscientists can work together. For interpreters, this means interactive 2D/3D/gather visualisation with tools for fault and horizon picking, well management and manipulation, crossplotting and AI-assisted interpretation. They can also generate attributes including dip and azimuth, semblance, curvature, spectral decomposition and RGB blending, all on demand.

For processing geophysicists, it offers a full suite of time-processing and depth-imaging tools scalable to massive data volumes from any acquisition geometry — land, marine or ocean-bottom node — with dedicated capabilities for time-lapse (4D) and multicomponent data. Users are not required to run heavy batch processes to test a parameter. On DUG Insight, it takes just seconds to test and QC workflows.

QI specialists get the fastest path to rock properties and probabilistic lithology and fluid prediction, with tools from statistical rock physics, to traditional AVA inversion, to our revolutionary elastic MP-FWI imaging technology.

DUG Insight is the only geoscience software package on the market that spans the complete workflow from seismic data processing and depth imaging through to interpretation, visualisation and rock-property prediction — all in a single, integrated package.

AI is all the buzz now. How does it enhance DUG Insight? What about users who want to bring their own code? 

AI tools are now being increasingly integrated into DUG Insight’s workflows for maximum efficiency, whether it’s lithology prediction, fault and horizon interpretation, or accelerated convergence of our MP-FWI.

We also don’t want to lock anyone into a closed box. The software’s API gives users the flexibility to bring their own code into the environment, supporting Java, C/C++, C# and Python. Ultimately, we want to ensure that your custom code is treated with the dignity it deserves. No additional licensing is required, and all intellectual property stays with you.

What’s next for DUG Insight? 

We’ve got lots of exciting developments coming soon! Our R&D team continues to embed the latest technology, spanning signal processing to interpretation to algorithmic efficiency. For example we’re now modelling complex acquisition effects like tides, water-column changes, currents and array geometry, which is vital for full-wavefield imaging and for superior time-lapse (4D) results. On land, we’re inverting ground roll to give high-resolution shear-wave velocity, while the new land statics methodology (AMGRT) solves near-surface complexity to correctly position deep targets. Watch this space!