Leaders from OpenAI and Samsara spoke of how fleets can start using AI to boost productivity, improve decision-making, and reduce operational costs, with some companies already reporting significant productivity gains — and at a fraction of the cost, as AI prices have dropped 100-fold in just two years.
Toki Sherbakov, head of solutions architecture at OpenAI, who spoke during Samsara’s Beyond conference in San Diego, Calif., said companies should begin using AI for low-complexity, high-value use cases and to enhance, not replace, your workforce. Then, eventually prepare for rapid advances as AI becomes more capable, affordable, and multimodal, and integrated into products.
Think of AI as an “Iron Man suit,” he said — a tool that helps employees do more, not something that takes their place.
Sherbakov said companies often overestimate how complex their AI strategy needs to be at the beginning. Instead, he recommended starting small — with use cases that are simple to implement but deliver immediate value. “Don’t overcomplicate things early. Don’t boil the ocean with AI. It’s like, first you start it on the lower complexity things, put it in the hands of your employees to build that AI literacy, enable everyone to know how to use it for the day-to-day [operations].”
Employees’ comfort with using AI tools daily is essential, he said, before scaling to more advanced applications. When addressing the concerns about AI replacing employees altogether, he said, “I very much don’t believe this. I actually think AI is a great accelerator to make people even more effective. We call it like putting on your Iron Man suit. It makes you much more impactful as an employee…We call [AI] a co-pilot, or it will be your assistant, hopefully. And then eventually, as you gain trust and confidence, it can move between the driver [and co-pilot seat].”
That impact of using AI is already being felt across different sectors, Sherbakov said, pointing to examples outside transportation to illustrate what’s already possible, adding that many companies in various sectors — from Apple and T-Mobile to Target and Spotify — use AI in their operations. For example, OpenAI’s technology helped John Deere reduce chemical use by up to 70% through AI-powered vision systems that identify the most efficient way to deploy crop treatments. John Deere also created an AI-powered maintenance assistant that responds to in-app customer queries based on their specific equipment, offering personalized troubleshooting and service recommendations.
Lowe’s took a different route — using AI to improve its retail experience in the home improvement space. Its customer-facing AI assistant, Mylow, can help users navigate purchases by recommending the right product and offering guidance tailored to their needs.
Fleets do not sell paint, but the same principles apply: start with tools that help your teams work smarter and solve problems faster, Sherbakov said. He pointed to Samsara’s own internal AI assistant, Samsara GPT, which is used by sales reps to retrieve technical product information, installation instructions, and case studies.
Zach Merritt, director of data science and AI at Samsara, said during the panel that quick access to company knowledge helps the team be more effective in customer interactions. Sales mechanisms have seen about a 20% productivity boost with the tool, he added, saying it’s also being used to speed up onboarding for new employees.
Merritt also stressed the importance of clean, high-quality documentation as the foundation for good AI outputs.
Evan Welbourne, vice-president of machine learning engineering at Samsara, echoed that internal enablement is key. His team uses AI for a range of internal workflows — automatically transcribing and summarizing meetings, automating routine bookkeeping and program management, and enabling non-technical staff to build and use simple tools.
The company has also adopted AI-assisted software development. “We’re using tools like Cursor or Codex, and they’re able to generate about 1% more code within the same amount of time when we’re using AI, and that’s an area where both the engineers and the models are getting better month over month.”
To encourage adoption, Samsara has implemented an internal champions network, hosted hackathons, and created learning events where employees share wins and tips with peers. “It’s very positive to see your colleagues accelerating,” Welbourne said.
Ultimately, OpenAI’s Sherbakov said AI is not a one-size-fits-all solution. Companies should avoid assuming that generative AI can replace more traditional models like machine learning entirely. Instead, the most effective strategies often combine both.
“They can actually inform each other really well,” he said.
He also noted that with the current generation of large language models, fleets don’t need to train their own systems from scratch. “Recently, we’re seeing it’s actually less needed, where the models are becoming quite generative… What’s really effective is you can just retrieve the right information, use that to inform the model and then give an output. It actually performs just as well, if not better, than building custom models.”
