AI agents are coming to Rivian's cars

My take: Rivian CEO RJ Scaringe said at TechCrunch Disrupt that the company's next vehicles will run AI agents from outside companies, like Anthropic. Today Rivian's voice assistant already adjusts the temperature or turns on seat warmers; the plan is to add models from frontier labs for more complex tasks, like planning dinner using something you mentioned on another device.

What interests me most here isn't the technology, it's the line they drew: Scaringe was clear that safety-critical functions like braking or steering will never be handed to an agent. That's exactly the conversation every company should have before putting agentic AI into anything: which tasks get delegated for convenience, and which ones stay protected because the cost of a mistake is too high.

For anyone running a business or building a product, this is a good template: it's not about handing AI full control all at once, it's about releasing specific tasks, from lower to higher risk, as you build trust in how it responds.

If you were designing a product with agentic AI, which tasks would you delegate first, and which would you keep completely out of its reach?