Next-Gen VEU Analysis
Three ways AI can enhance Vehicle Equivalent Unit calculations.
In August, I attended the Fleet Pro annual conference in Denver, and facilitated a class on Vehicle Equivalent Units (VEUs). The room was packed, and attendees had lots of questions. One that generated a lot of interest was how to use artificial intelligence (AI) to enhance VEU calculations and results.
For decades, fleet managers have relied on VEUs to translate a mixed fleet of sedans, dump trucks, and specialty equipment into a common measure of maintenance workload. The math behind VEUs hasn't changed, but the inputs feeding it are getting smarter. Here are three ways that AI is changing how organizations calculate and apply VEUs.
"The real value of AI shows up in speed and scale."
1. Data-Driven Weighting Factors
Most organizations use historic VEU tables that were created decades ago and are periodically adjusted. This approach has had reliable results, although it tends to treat every fleet as if it behaves like the average fleet.
AI changes this by deriving weighting factors directly from an agency's own maintenance history. Models can process years of work order data, including labour hours, parts costs, and downtime, and calculate VEU factors that reflect how a specific fleet's vehicles actually perform.
This is especially useful for specialized pieces of public works equipment that don't fit neatly into a standard classification. You can weight them based on real maintenance requirements instead of shoehorning them into the nearest category.
The result is a VEU table that's defensible, because it's adjusted based on the organization's own numbers.
2. Cleaner Inputs, Fewer Distortions
A VEU calculation is only as reliable as the data behind it, and fleet management data is often messy. Miscoded vehicle classes, duplicate work orders, and unexplained labour-hour outliers are common. These errors skew VEU averages and the staffing recommendations built on them.
AI-assisted anomaly detection can flag inconsistencies, catching outlier records that a manual review would likely miss. Natural language processing takes this further by extracting structured information such as failure type, repair category, and root cause from technician narrative fields that are typically too unstructured to analyze. This expands the usable dataset considerably, often revealing patterns in the data that were previously invisible simply because they were buried in free text.
3. Scenario Modelling for a Changing Fleet
Once VEU factors and staffing formulas are established, the real value of AI shows up in speed and scale. Instead of manually rebuilding spreadsheet models for every "what if," planners can run dozens of scenarios in the time it used to take to run one. They can calculate current VEUs and compare them to scenarios involving fleet growth, outsourcing shifts, or a phased transition to zero-emission vehicles (ZEV).
ZEV transition is a popular scenario right now, and many organizations want to know how electrification affects shop staffing. Electric vehicles typically carry a different VEU profile than their combustion counterparts. Predictive modelling can help agencies estimate how VEU totals, and the staffing levels tied to them, will shift as fleets electrify, giving planners a data-backed answer well before the transition is complete.
None of this replaces judgment. Just for fun, I entered my presentation slides into Claude.ai and uploaded a fleet inventory for a Canadian municipality. I asked Claude to follow the methodology and calculate the VEUs. I had results in minutes, but they were without context and failed to fully account for the location and climate, seniority of staff, or age of the fleet.
AI can validate inputs and speed up the analysis, but the nuances, interpretation, and recommendations still belong to humans.
