@RISK Agent (Beta)

@RISK Agent (Beta) brings an AI collaborator directly into the @RISK Excel environment through Claude for Excel. It combines the speed and accessibility of AI with the rigor, precision, and auditability of @RISK Monte Carlo simulation, so you can build better models, interpret results instantly, and generate stakeholder-ready summaries without leaving Excel.

Rather than answering @RISK questions from general knowledge, @RISK Agent (Beta) routes them through a set of purpose-built tools that read the structure of your model and your simulation results. This keeps every response grounded in your actual workbook. @RISK Agent (Beta) provides four core capabilities:

  • Model Review & Analysis — audits the structure of your @RISK model, flags issues by severity, and gives prioritized fixes to improve your @RISK model.

  • Simulation Results Interpretation — explains what your simulation results mean, focus on output statistics, key drivers, and tail risk for executive decision making.

  • Distribution Suggestion — recommends a set of appropriate probability distributions based on your input of what you are modeling.

  • Function Help & Documentation Search — answers questions about @RISK functions and features using official @RISK documentation and Community Articles.

Please note: @RISK Agent is a Beta feature. Capabilities, prompts, and behavior may change between releases. It is an AI assistant — always review its output before relying on it for a decision, just as you would a colleague's analysis.

Before You Begin

@RISK Agent (Beta) runs through Claude for Excel or Claude Desktop. To use it you need:

  • The latest 8.14.0 @RISK
  • A My Lumivero account with a valid @RISK/DTS license
  • A connected Claude-Anthropic Account — @RISK Agent (Beta) works with Claude. You supply your own AI assistant; @RISK Agent (Beta) provides the @RISK intelligence behind it.
  • A Custom Connector setup with @RISK Agent-MCP. Check our community article for a step-by-step guide on how to connect.(https://community.lumivero.com/s/article/How-to-create-a-custom-RISK-Agent-Connector)

There is not a separate ribbon menu for @RISK Agent (Beta), enable the Custom Connector in Claude every time you want to use @RISK Agent (Beta)

Distribution Suggestion

When you are unsure which probability distribution best represents an uncertain variable, ask @RISK Agent (Beta)for a recommendation. Describe your industry, model domain and what you are modeling, it suggests a set of appropriate @RISK distribution function and explains why it fits.

Helpful details to include in your request:

  • What the variable represents — for example, a duration, a cost, a count, a percentage, or a probability.
  • What you know about it — such as a known minimum and maximum, a most-likely value, or historical data.
  • The nature of the uncertainty — whether values are bounded, can only be positive, are discrete counts, or are symmetric around an estimate.

Try asking:

“What distribution should I use for project duration?”

“I have a minimum, most likely, and maximum estimate for cost — which distribution fits?”

“Which distribution is best for modeling number of defects per batch?”

“What should I use for an oil reserve estimate that can't be negative?”

 

Distribution Suggestion is a great companion to Model Review. If a review flags a distribution choice as a concern, ask @RISK Agent-Beta what to use instead and why.

Model Review & Analysis

Use this capability to have @RISK Agent-Beta audit your model before you run a simulation — or any time you want a second opinion on model quality. The Agent reads the inputs, outputs, and structure of your workbook and returns an executive-focused report. This is the AI equivalent of a peer review by an experienced @RISK modeler.

Try asking:

“Review my @RISK model and tell me if it's set up correctly.”

“Audit this model and rate its quality.”

“Find any issues in my model before I run the simulation.”

“Are there any distributions in my model that don't make sense?”

 

Please note: Model Review analyzes how the model is built — distribution choices, parameters, structure, and documentation. It does not run a simulation.

Simulation Results Interpretation

After you run a simulation, use this capability to turn the @RISK simulation results into a clear, narrative explanation. @RISK Agent (Beta) reads the results @RISK produced — output statistics, percentiles, and sensitivity data — and explains what they mean for your decision.

Make sure to enable the General Preference to create the simulation results for Claude to read in your workbook.

Try asking:

“What do my simulation results mean?”

“Interpret the results for my Total Cost output.”

“What's the probability we stay under the $5M budget?”

“Which inputs are driving most of the risk in my output?”

“Summarize these results for an executive audience.”

 

Please note: This capability requires a completed simulation. If no simulation results are found, or if the results cannot be read, @RISK Agent (Beta) will ask you to run (or re-run) the simulation rather than guess. Re-run the simulation and ask again.

Function Help & Documentation Search

Use this capability to look up how a specific @RISK features and functions works. Answers are drawn from official @RISK documentation rather than generated from general knowledge, and include a link to the relevant help page.

Try asking:

“What does RiskPert do and what are its parameters?”

“When should I use RiskLognorm instead of RiskNormal?”

“How does Latin Hypercube sampling work in @RISK?”

“What is Smart Sensitivity Analysis?”

 

For the full reference, you can always browse the @RISK Help directly.

Tips for Getting the Best Results

  • Keep the right workbook open and active — the Agent analyzes the workbook in front of it. Make sure the model you mean is the active one.
  • Run a simulation before asking about results — interpretation needs completed results. Review and Distribution Suggestion do not.
  • Be specific — name the output, variable, or function you care about. “Interpret my NPV output” beats “what do you think?”
  • Iterate — treat @RISK Agent (Beta)like a colleague — ask follow-up questions, request a fix, then ask it to review the change.
  • Verify before you decide — @RISK Agent (Beta) is an assistant. Always confirm its findings against the model and your own judgment.