How artificial intelligence is reshaping geotechnical skills

There is increasing reliance on underground engineering. The global tunnel construction market alone is predicted to grow from £88bn to more than £160bn over the next decade. At the same time, expansion in sectors such as renewable energy and offshore wind is creating a new reliance on geological exploration, putting additional pressure on companies to deliver faster designs.

As geotechnical companies expand their capabilities to meet this demand, artificial intelligence (AI) tools are being deployed to assist with mundane desk tasks and accelerate some of the most time-consuming processes, such as manual extraction of data from borehole logs, laboratory reports, and field notes. In addition, new specialist roles are emerging that require skills from a variety of disciplines.

This steady adoption of AI is changing the way geotechnical contractors and consultancies work and the skills they expect from their engineers. But as intelligent systems take over human tasks, this change raises ethical questions about how far companies can push technology.

Give shape to your daily work

Over the past five years, AI adoption has more than doubled among geography professionals. According to a 2025 survey of more than 1,000 international professionals in geotechnical work by geoscience software company Seequent, 51% of respondents use AI in their work, a significant increase from 19% in 2020.

Additionally, 48% say they currently use data science scripting (2020: 30%), followed by machine learning (36% (2020: 25%)) and advanced analytics (35% (2020: 25%)). Just a quarter (23%) of respondents said they never use these tools at work (2020: 45%).

Geotechnical engineers routinely use large language models (LLMs) such as ChatGPT, Claude, and Microsoft Copilot to automate workflows, analyze large datasets, build predictive models, and improve quality control processes. Machine learning (ML) and artificial neural networks are also being deployed for classification and regression tasks.

“The reality is that AI is becoming more or less ubiquitous within organizations, consciously or not,” said Jim de Weer, elected executive committee member of the British Geotechnical Association (BGA).

At Mott MacDonald, our geotechnical engineering practice incorporates AI throughout data processing, design optimization, and daily workflows. Engineers use a generative AI assistant called “EMMA” (an acronym for “Every Mott MacDonald Answer”) to search corporate data and instantly retrieve relevant information, reducing the time it takes to search for documents, extract parameters, or perform repetitive actions.

In parallel to EMMA, a dedicated sandbox team will test openly available LLMs such as Gemini, OpenAI, Google and Meta to develop bespoke tools.

“This team is made up of a small number of highly specialized users who can train young, enthusiastic people to use the tools wisely, especially in a geotechnical environment, because not all tools work the same,” said Zandos Orazarin, senior geotechnical engineer at Mott MacDonald.

According to consultant A-squared Studio, the introduction of LLM based on transformer architecture is being promoted in a second phase from 2022 onwards. Managing director Domenico Lombardi and directors Angelo Fasano and Tony Suckling say this enables advanced text analysis, reporting, knowledge retrieval and decision support.

A-squared Studio’s geotechnical engineers use LLM on a daily basis to help prepare deliverables, capture and manage information, and support coding tasks.

“While early models required high computational power, new, smaller, optimized models can now be deployed on local or private servers, improving data privacy, cost efficiency, and regulatory compliance,” Lombardi added.

However, the use of AI in geotechnical engineering goes far beyond the LLM. De Waele says some companies in the UK are also innovating with ML. He said geotechnical experts have begun collecting installation data from shipboard drilling rig instruments. By collecting installation data, analyzing optimal cycles, and feeding the results back to the control system, the rig learns how to reduce drilling time with each new site.

“We should expect widespread use of this type of machine learning,” says De Waele.

Back at Mott MacDonald, consultants have been using ML for several years to digest the vast amounts of data collected through sensors on construction equipment operating in deep excavations and geotechnical structures.

“Historically, there was so much data that a lot of it wasn’t fully analyzed,” explains Tony O’Brien, global leader of Mott MacDonald’s geotechnical practice. “Now we can understand it, but that alone doesn’t add much value. The value comes when you combine that technology with tools like numerical modeling and real-time inverse analysis.”

He adds that a “bonus” is that his team applies adjustments to observation techniques during construction, something Mott MacDonald does on UK and international projects. “If we apply that technology at scale, there is great value for our clients,” he says.

new AI jobs

The widespread use of AI tools for routine data entry and simple calculations will inevitably make some skills unnecessary. This doesn’t necessarily mean the entire role goes away. Rather, human input is shifting from time-consuming, repetitive tasks to higher-order decision-making.

In the consulting industry, AI is seen as an “augmentation tool,” says A-squared’s Fasano. Professional judgment, responsibility, and responsibility remain firmly with the engineer.

“The regulatory framework reinforces this human-centered approach,” Fasano says. G.E.. High-risk AI systems in safety-critical fields need effective human oversight so that experts can understand their limitations, detect anomalies, and intervene when necessary.

Suckling added: “Similar trends can be seen in other regulated professions, such as law, where the adoption of AI is increasing with strict governance. Across these sectors, there is a growing expectation to declare when AI is used in formal technical and advisory work.”

At the same time, AI is creating new roles in the geotechnical profession. Companies are now hiring “intelligent field survey” experts to operate AI-enhanced drones and sensors that automatically extract and format geotechnical data, and AI-enabled geotechnical analysts use ML and deep learning to predict soil behavior parameters from large subsurface datasets.

Mott MacDonald recently introduced a new role to support AI-powered methodologies across its geotechnical operations. The ‘Principal Geotechnical Data Engineer’ combines traditional geotechnical expertise with data science, AI and software engineering skills.

This new position brings together hybrid skillsets that were previously spread across the organization, such as engineers working on scripting and data analysis on the side. The move is part of Mott MacDonald’s broader strategic plan to support AI-enabled design, real-time inverse analysis, and observational techniques, complemented by the hiring of a data scientist to lead data-driven workflows. .

ethical dilemma

As AI tools become more sophisticated and their adoption becomes more common across industries, some professionals may understandably wonder if their role will still be needed in 10 years.

This may be especially true for new and junior roles, where AI could eventually take over easier tasks that require less human oversight. As AI continues to develop, some companies may be tempted to ignore younger engineers, Lombardi said.

But this could hurt these companies in the long run.

“Experienced engineers develop their knowledge over several years by performing routine tasks many times in different situations and on different ground conditions,” explains Lombardi. “In that regard, over time, there is a risk that effective AI governance will become impossible without smart geotechnical engineers.”

These concerns were echoed by BGA’s De Waele. He says that while AI is far from replacing geotechnical engineers at this point, fewer engineers will likely be needed in the future, especially those in the early stages of their careers. But when companies cut junior jobs to cut costs, they can create knowledge gaps and unknowingly jeopardize their company’s future, and the future of their industry.

“The tools available today have increased efficiency,” he says. “But it also raises questions because there are fewer opportunities to train, experiment, mentor and advance knowledge by working on problems under the guidance of experienced engineers,” says De Weer. “It may be convenient to use AI to create sketches, calculations, and even summary reports instead of junior engineers, but who will train the experienced engineers of the future?”

One area where De Weer warns that AI could be misused is graduate recruitment. The AI ​​application filters currently widely used by businesses have embedded biases that often result in the rejection of great candidates.

“This is destructive to graduate student morale, as talented applicants with good interpersonal skills are overlooked by the algorithm,” De Weer added. “Interestingly, some senior executives report that they have hired alumni who have been approached directly or have hired contacts within their networks to fill roles.”

Misuse of AI is a real risk that geotechnical companies must consider when developing and deploying new tools. This is especially important when working on infrastructure projects where public safety is at risk.

To ensure it is implemented ethically, Suckling highlights four considerations. First, responsibility and responsibility must remain clearly with human experts. While AI systems can assist with decision-making, they should not replace expert judgment in safety-critical situations.

The second is transparency. “Engineers and clients need to understand when AI will be used, how the output will be generated, and the limitations of the system to avoid over-reliance and automation bias,” he says.

The third point to consider is awareness of bias and data quality. Geotechnical datasets can be incomplete, site-specific, or historically biased, which can lead to misleading or unsafe AI output if not properly validated.

Finally, data privacy and security. These two considerations are essential, especially for sensitive projects.

“The implementation of AI must ensure that the technology enhances professional practice without compromising responsibility, trust, and public welfare,” Suckling concludes.