Better Risk Management with AI Predictive Tools in Infrastructure Projects

The Greenland Ice Sheet, a vast icy expanse spanning approximately 1.7 million square kilometres, plays an important role in our planet’s climate system, influencing sea levels, ocean currents, and global weather patterns.

Recent advancements in artificial intelligence (AI) and satellite technology have revealed an unexpected revelation: the ice loss in Greenland over the past four years has been 20% greater than previously estimated. This revelation, derived from the analysis of over 235,000 satellite images using AI techniques, indicates that ice melt is on the rise and could have a significant impact on ocean currents in the future.

While this clearly isn’t “great news”, the findings are valuable and demonstrate how AI and machine learning can help humankind better understand patterns and associated risks.

AI in Risk Management

It’s not just scientific institutions that are recruiting help from AI. Just as the technology has helped climate researchers understand the impact of glacier reductions, the same predictive analytics can be used in thousands of ways to enhance decision making and reduce uncertainties in risk management.

Specifically, predictive AI allows organisations to forecast risks and proactively mitigate them before they escalate into crises.

By analysing massive datasets, identifying patterns and correlations, predictive AI models can determine the likelihood of future events with higher accuracy compared to traditional statistical models.

In this article, Moataz Mahmoud, Director (Risk) at TBH chatted with our team on the benefits of using AI technology to optimise risk management in infrastructure projects – and what we can look forward to in the future.

How AI enhances decision-making and reduces uncertainties in risk management

The integration of artificial intelligence into risk management within the construction industry has marked a significant shift from traditional methods, addressing challenges that have long plagued the sector.

In the not-so-distant past, risk management was typically very reactive – problems were dealt with as they arose onsite, often leading to costly delays. At best, risk analysis was minimal and/or qualitative.

There are inherent challenges to projects, such as design changes, growth, and clashes which needed to be reworked, sometimes requiring significant investment, or at best requiring reduced functionality.

The loss of institutional knowledge due to the retirement or job transfer of experienced engineers, coupled with the limitations of human observation and manual measurement, has historically constrained the efficiency and effectiveness of risk management practices.

AI, with its capacity to simulate thousands of risk scenarios and enhance design resiliency, (a bit like traditional probabilistic analysis but with an intelligent touch), offers a solution to these issues, enabling a more proactive and informed approach to managing potential threats and ensuring safety.

Optimism and Skepticism

Moataz finds that most people he talks to are in one of two camps regarding AI.

The first group sees it as an instant fix-all, while the second dismisses the technology as hype, struggling to see the practical benefits.

“The reality is probably not so black and white and lies somewhere in between: AI shouldn’t be viewed as a universal solution – but rather as a valuable instrument that complements human judgment to help streamline workflows, improve productivity, and diminish risk. “

He believes that the end goal of AI is not just cheaper, faster jobs, but deliberate integration where human strengths in oversight, quality control and complex decision-making complement AI’s automated efficiencies: using the practise of mounting wearable cameras on a job site as an example.

“These cameras collect critical data that can be used to create augmented reality models to enhance planning, labour deployment, and overall project monitoring. The enhanced ‘project intelligence’ better equips teams to avoid unwelcome variables or surprises that undermine budget and timeline targets and promotes a more productivity-focused perspective, where teams spend less time on transactional tasks and more time on high-value activities.”

Predictive Analytics

Predictive analytics powered by AI algorithms offers interesting potential for improving risk management practices – but the technology still has a way to go.

“AI technology can help predict project performance, project failures, and insurance losses with fairly high accuracy – even though several attempts have been made in the market to reach this level of maturity, it hasn’t been achieved quite yet.” Moataz notes.

Adding that the technology will soon be able to predict when a bridge will need critical maintenance based on minute changes in vibration sensor parameters such as amplitude, frequency, duration, direction, and distribution of vibrational energy.

“Using vibration-based sensing technologies will soon help us predict how sections of road or rail will be affected by storms and incorporate this data into the design or maintenance program once it has been erected. Better still, by continuously re-training predictive models with new data, this predictive intelligence will just get smarter over time and forecast accuracy will similarly improve.”

Natural Language Processing (NLP)

By unlocking insights from unstructured text-based data, natural language processing (NLP) AI can be applied effectively to enhance risk management processes.

“NLP AI can be used to build, uplift, or refine the risk profile in many ways. Models can analyse clauses in contracts and insurance policies to automatically detect terms that indicate project risks, liability exemptions or warranty limitations.”

Additionally, NLPs connected to the internet, like Copilot and Bard, can be used to review regulatory filings, publicly available legal cases, and policy advisories to immediately highlight changes to compliance rules relevant to ongoing projects; and any detected non-compliance can be rectified in a timely manner before penalties apply.

“Advanced algorithms will likely soon be used to enable rapid analysis of documentation, plans, and sensor data from past and current projects, allowing correlation analysis to identify trends, causal relationships, and anomalies.  Using this approach, decisions will be assessed based on the impact they have on materials, vendors, and workforce allocation, among other factors. Monte Carlo simulations can then visualise the effects of each option on budgets, timelines, quality, and earnings, aiding decision-making by comparing alternatives for key performance indicators under hypothesised scenarios. If deviations from the critical path are detected, machine learning algorithms can then quickly re-evaluate options to adjust labour, materials, sequences, or forecasts, keeping activities on track.”

Risk Assessment

In project management, risk assessment is crucial for identifying and mitigating potential hazards.

“Quantitative methods have always existed, but traditional practice often relied heavily on qualitative analysis and institutional experience to evaluate uncertainties, which in many cases is subjective.”

Moataz highlights how this often led to events being missed or underestimated, known as bias.

“AI now enables far more data-driven, quantitative, and rapid risk assessment – allowing stakeholders to foresee uncertainties before they escalate. By combining predicted risks with real-time monitoring, AI systems can also provide an integrated, end-to-end view of evolving risks across projects for better resource allocation.”

He uses advanced weather prediction as an example of this.

“Unexpected weather often forces improvised changes on jobs, causing schedule disruptions. Predictive AI analytics can equip managers to plan around these climatic shifts, minimising schedule overruns. By processing volumes of historical meteorological data (far surpassing human capacity), AI-powered weather forecasting can be used to optimise operational scheduling and staff allocation. If you have high winds, or heavy rain, workers often can’t do their work. This causes changes in project schedules with the contractor repurposing staff to do other jobs, leading to lost productivity because of out-of-sequence work.

From Qualitative to Quantitative Approaches

Since AI weather forecasting models aren’t limited by the datasets they were trained on, they are able to produce incredibly accurate forecasts to guide operational adjustments and improve scheduling accuracy now, and in the future. This means they can be used to help model contingency plans to identify at-risk areas on a job, weeks, months or even years ahead.”

By combining AI and risk management algorithms, it is possible to look at thousands of probabilistic scenarios to enable rapid assessment of cost estimates, sequencing alternatives, resource contingencies, regulations, safety incidents, weather delays, and other variables to quantify the likelihood and impact of various risks materialising.

“This provides data-driven mitigation recommendations that rank the highest priority threats so that risks can be avoided before they cascade into major delays or cost overruns.”

Productive Workforces

Declining productivity is a considerable challenge for the construction sector, with cost overruns averaging near 30% on many of the largest megaprojects exceeding $1 billion. This has only been compounded by delays in adopting technological innovations.

Uncertainty, such as unexpected scope changes or time overruns, is a major cause of construction delays. However, artificial intelligence technologies offer promising solutions to mitigate these uncertainties by providing benchmarking capabilities, monitoring progress, and identifying potential risks and productivity issues.

“AI can enhance benchmarking by looking at historical data from completed projects, including scope, size, duration, and costs. This allows for the creation of reference points for future project estimates, improving accuracy and consistency and can help predict potential safety risks and causes of declining productivity.”

He points out how wearable sensors are now able to track risky worker behaviour, hazardous site conditions.

“These devices dynamically update individual risk ratings and can even detect out-of-sequence activities and pause-resume errors, significantly enhancing on-site safety protocols.”

On projects with fragmented subcontractors, AI technology can be used to match specific crews to tasks based on credentials, certifications, past performance, and availability to maximise work quality and sequencing and diminish accidents caused by inexperienced workers.

“AI-driven simulation and modelling platforms can even create digital twin replicas of projects to identify high risk variables through virtual prototyping.”

Supply chain optimisation

Supply chain delays and material/equipment shortages are major contributors to the high risk of cost overruns and missed deadlines in infrastructure projects. Using AI and machine learning models to continuously predict, detect and mitigate external risk events, projects can be shielded from supply and logistics uncertainties.

Internet of Things (IoT) sensors on material shipments are now able to provide real-time tracking data to the AI systems regarding delays – and then recompute better ongoing supply chain strategies for the future.

“Automated tracking of real-time inventory can be used to ensure that as assets get deployed on site, supplies are dynamically recalculated to prevent overstocking of some materials while running out of others. Aside from the obvious safety benefits of these controls, expensive and time-consuming lawsuits from accidents or material workmanship can be prevented by demonstrating how effective monitoring was in place – even lower insurance premiums could result from well-documented AI safety efforts.”

Cost Estimation

Inaccurate cost estimations during the bidding stage of an infrastructure project can carry significant financial risks throughout development. Overruns can be as high as 45 percent for rail projects, 34 percent for bridges, and 20 percent for road projects, according to industry research.

“With advanced machine learning algorithms processing vast sets of parameters around past project budgets, geography, regulations, chosen technologies, construction materials as well as macroeconomic factors – AI models can estimate total costs with a high level of accuracy for new projects based on preliminary specifications.”

Schedule Optimisation

AI-based schedule optimisation offers a promising solution to mitigate the risks of time overruns, which can escalate costs and impede project momentum.

“By analysing data on activity sequencing, dependency constraints, resource availability, and variability in task durations – rooted in historical performance – AI may be able to significantly enhance project scheduling. Although currently, many market-available platforms primarily serve as comparison tools rather than optimisation solutions, the potential for AI in this domain is vast.”

Asset Management

Worksites contain valuable equipment and vehicles that are prone to overuse, damage, or theft.

Machine learning algorithms can be used to detect signs of attrition and predict failure likelihood for assets due to overuse so that preventive measures can be taken before a complete breakdown occurs.

“Computer vision powered by deep learning also helps review asset conditions from images to detect safety issues or missing parts requiring repairs – while autonomous drones (much like in a Sci Fi film) can be used to scan sites to automatically audit asset stocks and send anomaly alerts if any misuse or theft is detected.”

The future of AI in infrastructure risk management

Another potential application that Moataz believes might useful could be standardising regulatory procedures in countries like Australia who have their own distinct state-by-state worker qualifications – which in its current shape fragments the workforce market and restricts mobility (and productivity) of employees who are qualified in one state, but ineligible for similar jobs across state borderlines.

“The magnitude of systematising qualifications on an industry-wide basis has slowed progress toward standardisation goals, and digital systems like project management software and building information modelling tools also remain siloed, unable to share data insights across states.  This hampers benchmarking, progress tracking, and continuity since each state’s dataset remains disconnected from the other, therefore project managers are often unable to share insights between projects for benchmarking.”

Moataz envisions a national AI data hub that could be used to catalogue construction initiatives using standardised metrics to allow diagnosis of why some underperform through quantitative modelling. “By scrutinising differences in project types, fiscal environments, precise qualification criteria, and regional labour constraints – this kind of system would then be able to determine their impact on outcome inconsistencies.”

Summing up

Though the transition may seem daunting at first, the potential benefits of AI make it an exciting evolution to help modernise and maximise the profitability of infrastructure projects through enhanced risk management.

In essence, AI adoption can help teams uncover what they “don’t know they don’t know” by connecting data dots that would otherwise remain scattered. This lays the foundation for more proactive management, higher performing contractors, and ultimately the knowledge reinvestment vital for advancing civil infrastructure as we know it.

About Moataz Mahmoud

Moataz Mahmoud In his role as TBH’s head of risk services, Moataz Mahmoud helps clients establish, reform and improve their risk and contingency management functions for some of the most complex and high-risk portfolios in the infrastructure, defence, utilities, ICT, and building sectors of the economy.

For more information on TBH’s Risk services, click here.

Speak to Our Experts