Rivian is developing its own AI assistant

Rivian has spent nearly two years developing its own AI assistant, an effort separate from its multibillion-dollar technology joint venture with Volkswagen, TechCrunch has learned.

Rivian hasn’t said when it will put its AI assistant in the hands of consumers. However, in an interview earlier this year, Rivian’s head of software Wassym Bensaid told TechCrunch that the company is targeting the end of the year. The company may share more at its upcoming AI & Autonomy Day, which will be livestreamed on December 11th at 9am PT.

Rivian’s plans reflect a time when the pace of development of foundational AI companies (tech giants and startups like Anthropic, Google, Microsoft, Meta, and OpenAI that are building core models and infrastructure, such as Anthropic, Google, Microsoft, Meta, and OpenAI) is accelerating and the industry is racing to keep up.

But as Benside pointed out to TechCrunch earlier this year, this isn’t a haphazard effort to stay on trend. Nor is it just a chatbot built into an infotainment system. Benside said the company has put considerable thought, resources and time into this product, which is designed to integrate with all vehicle controls.

According to Bensaid, the company started with a basic philosophy of building an overall architecture that is model and platform independent. The Rivian AI Assistant team, based in the company’s Palo Alto office, quickly realized that they should focus their efforts and attention on developing not only the control logic that resolves conflicts, but also the software layers that help coordinate the various workflows.

“And that’s the in-vehicle platform we built,” Benside said. “We use something called the agent framework, which is now beloved in the industry, but we were thinking about its architecture very early on so that it could work with a variety of different models.”

The internal AI assistant program is consistent with Rivian’s efforts to become more vertically integrated. For 2024, Rivian has completely overhauled its flagship R1T truck and R1S SUV, changing everything from the battery pack and suspension system to the electrical architecture, sensor stack, and software user interface.

The company has also put significant resources into developing and improving its proprietary software stack, including everything related to the real-time operating system (RTOS) that manages the car, such as thermodynamics, ADAS, and safety systems, as well as another layer related to infotainment systems.

Benside didn’t provide further information about the AI ​​assistants, but said there will be a mix of models handling specific tasks. The result is a hybrid software stack that combines edge AI, where tasks are processed on the device, and cloud AI, where larger, compute-intensive models are processed on remote servers.

This means flexible, customized AI assistants that split workloads between the edge and the cloud.

Rivian developed much of its AI software stack in-house, including its own custom models and “orchestration layers” like conductors and traffic cops that ensure the various AI models work together. Rivian tapped other companies for certain agent AI capabilities.

Benside said his mission is to develop an AI assistant that increases customer trust and engagement.

For now, the AI ​​assistant will remain within Rivian. The company’s joint venture with Volkswagen focuses on software, but has nothing to do with AI assistants or autonomous driving.

The technology joint venture with Volkswagen, announced in 2024 and worth up to $5.8 billion, will focus on the underlying electrical architecture, zone computing and infotainment. The joint venture will officially start in November 2024 and is expected to supply electrical architecture and software to the Volkswagen Group as early as 2027.

While autonomy and AI are separated for now, “that doesn’t mean they might not be separated in the future,” Benside said.