The Biggest AI Deployment Nobody Talks About | Samsara CEO Sanjit Biswas

Sanjit Biswas, co-founder and CEO of Samsara, explains that physical AI represents the application of AI to real-world infrastructure like roads, construction sites, electrical grids, and supply chains—areas that have historically been underdigitized compared to the digital world. Samsara operates what may be the largest AI deployment in the physical world, collecting 25 trillion data points annually from millions of vehicles and frontline workers.

The company's architecture comprises three layers: hardware (GPS trackers, dash cameras, asset tags, vehicle gateways), cloud infrastructure for data ingestion and processing, and AI models for reasoning and action. Samsara's data is uniquely defensible because it cannot be found online—it comes from actual operational experiences on construction sites, roads, and utility grids. The company has helped prevent approximately 380,000 car crashes in the last year through real-time driver alerts, fatigue detection, and behavioral coaching.

Regarding AI implementation, Samsara runs inference at the edge for low-latency driver alerts and detection, while performing training, video reasoning, and complex analysis in the cloud. The company uses multiple model families including convolutional neural networks, vision language models (VLMs), and generative models for video coaching. Recently, they launched Agent Studio to enable longer-horizon autonomous workflows—such as a warranty agent that can correlate fault codes with OEM warranty agreements to automatically open work orders.

Biswas emphasizes that successful physical AI deployment requires change management, transparency with frontline workers, and careful consideration of safety. The company's approach to camera systems focuses on positive reinforcement (exoneration, safety improvement) rather than surveillance, which has driven adoption and trust among drivers. The combination of operational context (workflows, guardrails) with agentic reasoning capabilities appears essential—neither works well in isolation.

Looking forward, Samsara sees mixed fleets of human workers and robots, with humans handling exceptions and judgment calls while robots perform repetitive tasks. The energy sector exemplifies this trend: one customer noted they will triple grid capacity in 5 years compared to what was built over 125 years, with 90% of that demand driven by data centers. This infrastructure buildout creates massive demand for trades workers, even as automation increases. The company positions itself as the orchestration layer across diverse equipment, labor types, and autonomous systems.