I interviewed the vice president of corporate security, asset management, and housing at Grand Isle Shipyard. Derek Champagne discussed how he has used Samsara Agent Studio to automate fleet management and safety processes. He builds agents that automate work by speaking to the Agent Studio in plain language and telling it what he wants to accomplish. He has no background in IT.

At the Samsara user event, Champagne told the audience how he uses Agent Studio. “I had a gentleman sitting next to me who was a software guy, and I told the crowd, ‘I am not him. I’m just an everyday guy who likes to solve problems. If I can do it, you can do it.’”

Samsara is a public company with $1.6 billion in revenue in the last fiscal year. The company serves over 20,000 customers. Its platform integrates telematics, dash cameras, and AI to monitor and improve safety, fuel efficiency, and operational efficiency.

Grand Isle Shipyard services the oil and gas, power, renewables, and critical infrastructure industries. With nearly 3,000 employees, 21 facilities, and 20 service lines, GIS aligns its services with customer operations across the lifecycle of major construction projects. That lifecycle spans development, construction, commissioning, and ultimately decommissioning.

GIS has over 1,000 tagged assets. This includes over 700 trucks and roughly 300 other valuable pieces of equipment, including backhoes, bulldozers, and barges. They also track smaller items, like toolboxes used at offshore oil rigs. The fleet mainly consists of standard pickup trucks, such as Ford F-150s and Chevy Silverados, but also includes a few larger trucks, like 18-wheeler semis.

Prior to automating work with agents, a customer still needs to implement the Samsara solution. The Samsara platform can then operationalize the data associated with the organization’s assets, such as vehicles and equipment. The rich data buried in this application makes the agentic journey so valuable.

Historically, creating new capabilities by accessing core data buried deep in an enterprise application has required a specialist, often someone in IT or operations, to laboriously configure the dashboards, reports, and workflows everyone else relied on. Now, with the Samsara Assistant, an operator can describe what they want in plain language, like ‘show me my top 10 safest drivers,’ and the platform produces an answer, a custom dashboard, or an agent that automates the work. Samsara is clearly ahead of its peers in its agentic AI journey.

Mr. Champagne spent over 14 years in law enforcement, primarily engaged in narcotics investigations, before joining GIS. Before that, he served in the Marines. He has been at GIS for three years. He has no background in IT.

When GIS hired him, they initially wanted him to focus on many of the firm’s vehicle issues. GIS had another system in place. “I wasn’t really pleased with it.” That solution provider didn’t want to support some of the innovations Champagne was looking for. So, he sought another supplier. He started at GIS in December 2023; by April of 2024, he had decided to move ahead with Samsara.

The main goal of implementing Samsara was to prevent accidents by correcting a variety of driver behaviors. An accident, even if settled out of court, can cost over a million dollars. “Things were getting out of control for a while,” Champagne exclaimed.

“We’ve had several instances where a law enforcement guy on the scene tries to rule us at fault,” Champagne explained. “Then I get on the phone and say, ‘Hey, you want to check out this video!’ I show them this video, and then that automatically changes things right then and there. So, just having that visibility has been a no-brainer.”

Fewer accidents mean lower insurance premiums. Their insurance company has full access to the Samsara dashboard. GIS receives a substantial discount for allowing that.

The main barrier to driving benefits with the Samsara solution “was getting people to buy into it. Everybody looks at it and says, ‘Oh, I have Big Brother in the truck with me,’” Champagne lamented. He has six big TVs on his office wall used to monitor operations. “Everybody thinks, because of my background, that all I do all day is watch them in their trucks!”

GIS executives take road trips to meet with their dispersed workforce and talk about different aspects of the company. To support the Samsara implementation, Derek’s team gave asset-management PowerPoint presentations with examples of how Samsara protects the company. “And I always let them know that this is not a punishment tool. Samsara helps us. We want to see them be great in the things they do.” The goal is to use the application to coach, not to punish.

GIS and the Samsara Studio

Champagne was an early adopter of Samsara’s Agent Studio. He builds agents that automate work by speaking to the Agent Studio in plain language and telling it what he wants to accomplish.

For example, you can say, ‘I need to build an agent to give me my weekly KPI for mileage use.’ Once you get the initial applet built, then you can go back and expand on it. Champagne verifies the agent is pulling the right data out of the system by drilling down into the system and checking the output. Once verification has occurred, he says to Agent Studio, “All right, let’s go.”

Champagne is an early adopter. Sometimes he tells the system what to do, but it doesn’t give him what he is looking for. He reports that to Samsara, and they fix it.

The first application he built eliminated the tedious task of ensuring inspection reports are filled out correctly. Every driver must inspect their vehicle every month. This involves taking pictures of the vehicle, confirming their registration is valid, reporting new damage, and ensuring their inspection stickers are up to date.

As a result, one team member reviewed 730 reports, entered data into an Excel spreadsheet, and created a report card for all managers. That well-paid team member spent a full three weeks on this task. All this data lived in the Samsara system, so an AI agent could be used to automate the task and free the team member to do more productive work.

This process was automated and, unsurprisingly, delivered a solid ROI. That will not always be the case. Agentic AI is built using large language models. The tokens paid for the use of an LLM are not insignificant. Samsara customers use an AI license based on a credit system that looks at AI consumption per vehicle, similar to consumer AI models.

Samsara supports an open approach to LLMs by using MCP (Model Context Protocol) to allow customers to access Samsara data through external tools like ChatGPT, Microsoft Copilot, or Claude. This open strategy lets a company choose the least expensive LLM tool to accomplish a task. The fleet management software company is developing outcome-based deals in which agents are measured against specific objectives, such as increasing seatbelt compliance, to enable easier ROI calculations.

Since June, Grand Isle Shipyard’s vice president of security has created 20 agents. A Driver ID agent calls truck cabs and asks drivers to identify themselves if they fail to sign into their vehicles correctly. A Crash agent detects accidents, checks driver safety via the cab speaker, and automatically initiates the formal incident management process. A Maintenance and Vendor Status agent tracks assets in third-party shops to ensure drivers don’t leave GPS units unplugged after service. A KPI reporting agent integrates compliance and fuel utilization data into visual dashboards so managers can control costs effectively.

“Whatever you’re thinking in your head,” Mr. Champagne concluded, “you can make it happen. You just have to have a vision of what you want. It’s almost criminal not to use.”