Hertz adopts AI for fleet and workforce management

Hertz got its start renting out a dozen Model T Fords more than 100 years ago. Today, Hertz operates in 160 countries and has more than 20,000 employees and 500,000 vehicles in its fleet. To streamline an operation with so many moving parts, the company has deployed Hertz Connected Fleet OS, an AI-enabled operating system for fleet management.

“This is all around our purpose for our customers, which is making sure we have the right car, at the right place, at the right time,” said EVP and CIO Tim Langley-Hawthorne at Palantir’s AIPC last month, before stepping down from the role soon after. “It orchestrates customers, vehicles, and workforce. Those are the three critical components for businesses like ours on the ground.”

Hertz has leveraged Palantir Foundry and Palantir Artificial Intelligence Platform (AIP) to create a set of AI-powered applications to help it increase efficiencies in vehicle turnaround, reduce maintenance expenses, predictively allocate its workforce across field locations, and match the best car to customers.

At the moment, when customers return vehicles to one of Hertz’s 11,000 locations, the company must go through a process before it can be rented to someone else, all held together by relatively low tech. First, it goes through intake, then it’s moved to maintenance staging. Maintenance workers go over the vehicle and certify multi checks are complete before it’s moved to dirty staging. Workers then clean the vehicle before it’s moved on to clean staging. From there, it’s ready to be rented again.

Delays anywhere in this process have the potential to cascade, jeopardizing the company’s ability to put customers in the vehicles they want as swiftly as possible.

“Historically, this operation was run on two-way radios and dispatchers,” Langley-Hawthorne said. “It was run on people running around the lots, talking to people.”

Now, though, Hertz has a lightweight app built on Foundry and deployed on Android devices. All employees involved in the process — from maintenance workers and cleaners, to transporters and customer service agents — can log where they are in the process and receive intelligence on their next best action. With that data, AI can monitor activity and provide real-time recommendations for smoothing out bottlenecks.

For instance, a weather event in Atlanta might delay flights and snarl traffic, which might make it harder for customers to return cars on time, leading to longer cleaning times as well. Foundry might suggest moving some employees from front desk, the main counter, and exit gates to returns, because it predicts there’ll be a surge in returns later in the day as a result. It might also determine that overtime might be necessary to cope with some of those issues, and send out overtime requests to employees via their Android devices. Employees can accept or reject the request, and their responses are worked back into the model for optimizing the operation.

Building a case for data

Hertz, Langley-Hawthorne said, isn’t a company that’s here to do math projects and science experiments. Rather, the IT team is relentlessly focused on enabling the business. But when he joined the company in November 2021, it didn’t have much of a data analytics team in place.

“Every technology leader before me in the last 10 years started something but never finished,” he told CIO.com.

The upshot of that was the company didn’t have a reference data architecture in place. While conventional wisdom says it’s essential to get your data architecture and data governance in order before diving into the deep end with AI, Langley-Hawthorne’s plan was to identify areas where data could make a big difference to the business and build out those areas as quickly as possible. He also didn’t ask for extra funding to put data architecture and data governance in place. Instead, he was in the process of steadily retiring legacy data lakes and technologies, finding savings there, and then reinvesting in data governance, for example.

“The trick for us was don’t try to perfect all of it,” he said. “Narrow it into the areas where we think using tools like Foundry will drive the most value and get that piece. I wouldn’t say our data architecture and our data governance is fully complete at this point, but we’re in a pretty good spot in the fleet area. It’s not perfect, but we narrowed it down to the areas we thought were most value.”