AI risk, rewards, and reality: Insights from our expert panel
Generative artificial intelligence (AI) tools have been taking industries (and business functions) by storm—and risk management is no exception. At Lumivero, we believe that effective risk management blends innovation with informed decision-making, and AI is becoming an integral part of that process.
To explore this developing dynamic, David Danielson, Senior Product Strategist at Lumivero, brought together a panel of risk management experts from a range of industries to discuss how AI tools are impacting the work they do now – and what changes could be in store for the future.
Panelists included:
- Manuel Carmona, Risk and Decision Analysis Specialist at EdyTraining, Ltd.
- Lachlan Hughson, Founder, 4-D Resources Advisory, LLC
- Glen Justis, Senior Partner at Experience on Demand, LLC
- Quinton van Eeden, Quantitative Project Risk Analyst & Planner, TPG GRC
In this lively discussion, the panel examined how AI is reshaping risk management—highlighting its benefits, limitations, and potential roadblocks to adoption. They explored challenges like bias and accuracy, as well as real-world examples of companies integrating AI to strengthen their operations and decision-making.
Watch the webinar or continue reading to hear what they had to say.
Artificial intelligence as a partner, not a replacement
David opened the discussion by asking the panel for their impressions of AI’s role in risk management today.
“I think of AI as a thought partner to the modeler, the risk manager, or the project manager,” said Manuel Carmona. He emphasized that AI can enhance predictive accuracy by detecting patterns in historical data that traditional probabilistic models or human experts might overlook. Purpose-built AI models, he explained, can quickly analyze risk registers, project plans, and financial models to identify emerging threats. However, he cautioned that AI should be seen as a “sidekick” rather than a leader in risk management. The panelists agreed—while AI enhances automation, it is not yet sophisticated enough to operate without human oversight.
“I think there's great opportunity here for AI in the risk space,” Quinton van Eeden added, “But let's not get stuck on a black box. Let's do the uniquely sentient human endeavor of applying our minds to the problem.”
Glen Justis pointed out that the fundamentals of the risk management function haven’t changed simply because a powerful new tool is available. “It's [still] all about identifying, evaluating and managing threats to business performance,” he said.
Modernizing financial institution risk management
For decades, financial risk management has relied heavily on spreadsheets, with Microsoft Excel serving as the default tool for modeling and analysis. While useful, using Excel alone often lacks the sophistication needed to capture the full complexity of financial risk. The panel agreed that AI and machine learning offer a path forward, providing finance teams with more advanced, dynamic risk analysis capabilities.
Lachlan Hughson, drawing on his experience in finance within the energy sector, argued that it’s time for finance professionals to move beyond outdated tools and embrace more powerful risk modeling techniques—methods that have been standard in science and engineering for years. AI-driven analytics, he noted, have the potential to add significant value by enhancing how organizations understand uncertainty and variability in financial risk.
“It’s time to get the finance function that currently uses a 30-year-old tool – Excel – to upgrade its contribution in a way that really can add significant dollar value.” Lachlan noted that he was encouraged to see so many attendees indicate that they were already using @RISK and other Monte Carlo simulation-based risk analysis software. “That gives us a much broader way to understand risk, to understand variability, to understand uncertainty.”
Manuel Carmona was able to give an example of how he had helped a company develop a more accurate financial risk model across multiple business units by using AI. After developing an initial probabilistic risk register based on consultation with each business unit leader, Manuel and his colleagues turned to AI.
“We fed [the risk estimates] into the AI system just as a second filter, providing lots of context,” Manuel explained. They then compared the AI’s outputs to those provided by the business unit leaders and used variances between the two to further refine their model. The result was a financial risk model far more sophisticated and dynamic than simple spreadsheets and basic linear regressions.
From static to smart: Using AI to manage risk with self-updating models
Updating and rebuilding risk models as data changes is a time-consuming task, often requiring hundreds of hours from analysts and programmers. AI can streamline this process, reducing manual effort and making model adjustments more efficient.
Quinton van Eeden was enthusiastic about the potential of neural networks to create operational risk models with automated update capabilities, describing the progress he had been able to make with a South African mining company. The company was continually running into issues with oversupplies of buffer stock. Fortunately, they had invested in plenty of sensors and other tracking devices that generated large amounts of data for every stage of production.
“We were able to pull that [data] and build a little Monte Carlo model initially,” van Eeden explained. “But then they decided to populate and train a neural network with the data based on the dependent variable. The mining company now has a model that constantly adjusts itself – again, subject to review by human experts – and is beginning to provide realistic projections they can use to make adjustments along the production line.”
David Danielson agreed with this, explaining that he has seen many companies find similar efficiencies. “AI does provide us with an opportunity to take real-time data and feed it back into the [risk] models and enhance the outcomes faster.”
Additionally, automating risk model adjustments would free up time for deeper analytical work, allowing professionals to focus on interpretation rather than manual recalibration. As Glen Justis explained, “With the proper use of AI, you can have the human spend more time in analysis and interpretation of information rather than going back and recalibrating the models manually.”
Effective AI adoption requires buy-in, time, and expertise
Several of the panelists indicated that many clients were experiencing hesitancy around incorporating AI into risk management processes, especially when making critical operational or financial decisions. Quinton van Eeden suggested starting by showing decision makers small wins in some low-criticality areas, such as document summarization.
Glen Justis explained that there is quite a lot of groundwork involved with getting AI to produce valuable outputs or time-saving opportunities. This has both positive and negative aspects. On the plus side, the fact that organizations need to spend time “basically teeing up the AI engine to give you reliable information,” as Glen put it, can reassure decision-makers that the AI is not producing results out of thin air, and that it can genuinely add value. On the negative side, mid- and senior-level risk decision-makers are incredibly busy. “There's so many pressures to just get the basic business governance work done that people don't have time to really scratch the surface of [new technology],” said Justis.
The panelists agreed that while newer risk management and finance professionals were learning how to incorporate AI during their education and training, senior-level professionals tended to be most resistant. Quinton van Eeden reasoned that companies would move ahead with AI adoption as competitive pressure to do so arises. “Frankly, the only way a lot of boards can really meet their fiduciary duty is by bringing a more dynamic approach [to risk management].”
The accuracy gap: A key challenge for AI in risk and compliance
A major concern shared by both the panelists and attendees was the accuracy of AI-generated data and outputs. Large language models like ChatGPT and Google Gemini are known to produce errors—or “hallucinations”—raising questions about their reliability in risk management.
IBM describes hallucinations as “misinterpretations [that] occur due to various factors, including overfitting, training data bias/inaccuracy, and high model complexity.” To users without expertise, hallucinations can seem plausible, leading to decisions based on bad information.
“Everyone is wary or scared of this black box effect,” said Manuel Carmona. He explained that he mitigates the risk of inaccurate data by rigorously reviewing AI outputs. “I filter through all the information that comes from the AI,” he said. “I compare it to the information that I get from other experts and consultants about the risks.” He then takes the AI output and cross-checks it against other LLMs. Finally, he uses Monte Carlo simulation in @RISK as a further check of how realistic the outputs are.
Quinton van Eeden again stressed the importance of well-informed human oversight when using AI tools. “There's no algorithm in the likes of ChatGPT to check [whether] whatever it produces is true,” he reminded attendees. “ChatGPT is only usable if you know the subject very, very well, because it does make embarrassing mistakes that only a true expert or a connoisseur can detect.”
Generative AI in risk management is just getting started
The future of AI in risk management isn’t just promising—it's full of possibilities. The panelists saw generative AI not as something new, but rather as a natural evolution in automation and noted that AI would lead to other types of innovations we can’t imagine yet.
“When electricity was [harnessed],” Manuel explained, “That led to the invention of the light bulb. Once people start figuring out how to use [AI] in real-world applications, we are going to see a massive change in all sorts of professional applications and life in general.”
The panelists’ discussion made one thing clear: AI is set to enhance risk management efforts, not replace its fundamentals. While new tools will improve efficiency and expand analytical capabilities, the core principles of risk management remain unchanged. Wrapping up the session, David Danielson reinforced this point, stating, “The best practices [in risk management] will stay the same. But I think the methods and the efficiency of getting to those will improve over the next couple of years.” As AI continues to evolve, its greatest impact will come from empowering professionals to make smarter, faster, and more informed decisions.
Rethink risk: Innovate with Lumivero
Ready to transform how you manage risk? Request a demo of @RISK today.
Glen Justis
Senior Partner at Experience on Demand, LLC
With over 30 years of experience in consulting and industry, Glen Justis has built an exceptional reputation for assisting clients at the intersection of strategy, economics, and risk management. He employs a goals-driven approach to address strategic and tactical issues, ensuring the implementation of optimal solutions tailored to client needs.
Manuel Carmona
Risk and Decision Analysis Specialist at EdyTraining Ltd
Manuel Carmona, MBA-RMP, is a specialist in risk and decision analysis with a focus on project risk management. He has extensive experience in managing risks in projects using ScheduleRiskAnalysis and is recognized for his contributions to leading risk management standards.
Quinton van Eeden
Quantitative Project Risk Analyst & Planner, TPG GRC
Quinton Van Eeden is a risk/decision analyst with more than 30 years’ experience in enterprise/project risk within mining and other industries. He is a lawyer by training and holds advanced level professional PMI certifications in project- and project risk management as well as a master’s degree in information and knowledge management.
He specializes in the application of quantitative modeling and analysis techniques to elucidate the effect of uncertainty on business decisions, project estimates, operations, and strategic investment decisions so as to achieve Decision Quality.
Lachlan Hughson
Founder, 4-D Resources Advisory LLC
Lachlan Hughson, the founder of 4-D Resources Advisory LLC, has over 30 years of experience in corporate finance, M&A, and capital markets across the oil/gas, renewables, and mining/metals industries, and as an investment banker and finance director – undertaking $30+ billion of M&A and $15+ billion of capital raising as an agent and principal. His education includes an MSc from Imperial College London and an MBA from the Kellogg School of Management. More information can be found at 4-dresourcesadvisory.com.