Pair large language models with @RISK to broaden risk identification, propose distributions, and tighten assumptions—but always keep Microsoft Excel “in the loop” for fitting, verification, and expert review. The greatest value comes when results move beyond a single workbook. Connect @RISK simulations to Predict! to centralize registers, feed dashboards, and align decisions to governance—so model insights turn into enterprise action.
New techniques for refining probabilistic risk modeling are always emerging—and right now, AI is at the front. Some of the most recent such techniques involve the use of large language models (LLMs) such as ChatGPT, Microsoft Copilot, and Anthropic’s Claude—tools commonly referred to as artificial intelligence or AI.
In a recent Lumivero webinar, Manuel Carmona, Risk and Decision Analysis Specialist at EdyTraining Ltd, described some strategies for using LLMs to compliment the building methods of probabilistic models in Lumivero’s quantitative risk analysis tools—@RISK, PrecisionTree, and RiskOptimizer.
Continue reading for the highlights of the presentation or watch the webinar on-demand.
Monte Carlo simulation is a method of sampling statistical distributions to view all possible outcomes. It helps decision makers understand the probability of how uncertainty and risk may impact budgets, timelines, and performance.
@RISK puts powerful Monte Carlo simulation tools right into Microsoft Excel, making it easy to assign input distributions and simulate thousands of scenarios for robust, decision-ready analysis. This allows you to model uncertainties in cost estimation as well as financial, operational, or project-related scenarios. With help from LLMs, you can refine these models for even better forecasting.
Probabilistic cost estimations created with Monte Carlo simulation in @RISK provide more information on the range of possible outcomes.
LLMs don’t replace quantitative risk analysis—they accelerate the “thinking-around-the-model.” Two patterns from the session stood out:
Example 1: Pressure-testing a risk register
In this first example, a risk manager collected departmental input on risks but faced uncertainty around certain probabilities. The team used a LLM as a “sounding board” to help surface overlooked risks and propose initial ranges by prompting the AI to review both the department leaders’ perceptions of risks and the risk analysis model. The output requested was a set of probabilities for the risks the department leaders were uncertain about.
The output from the LLM was then reviewed with department leaders and other experts. “In the vast majority of cases,” said Manuel, “the information provided back from the LLM tallied with the information provided by the experts.”
Example 2: Identifying optimal distributions for risk models
The next example shows how LLMs can scan historical data or parameter ranges and propose plausible distributions with rationale.
“Believe it or not, the LLM knows a lot about @RISK,” said Manuel. In his example, the LLM suggested specific functions in addition to statistical distribution options. However, the LLM also demonstrated why you must use care with AI in project risk analysis—the LLM declared that @RISK does not include the option to use a Fréchet distribution (also known as an F-distribution). This isn’t correct. “It's very, very important at all times that you . . . assess that all the outputs the LLM is giving you are right,” Manuel said.
There are many practical handoffs between LLMs and additional tool’s in Lumivero’s risk analysis toolkit like PrecisionTree, NeuralTools, and RISKOptimizer:
Giving an LLM good instructions—called “prompts”—is critical for effective use of AI in risk analysis.
“If you want to create probabilistic model-writing prompts, it’s not just writing two lines, then copy and paste,” explained Manuel. “It requires thinking about the problem and providing a lot of context to the LLM.”
The basic building blocks of a prompt include:
One example of an AI prompt for risk analysis was:
“You are an expert consultant in cost estimation for oil and gas projects. Analyze this project’s dataset to determine how likely it is that costs will exceed $10 million. Produce a report of your findings that is based on industry best practices ensuring easy interpretation by financial planners.”
MIT’s Sloan School of Management’s “Effective Prompts for AI: The Essentials” offers additional insight into writing effective prompts and follow-up prompts.
Finally, proceed with care when using LLMs to enhance your probabilistic risk modeling. Be sure to:
Ready to build more powerful risk models? Request a demo of @RISK to see how Monte Carlo simulations and decision analysis come together in Microsoft Excel—augmented by practical AI workflows.
