Mercedes-Benz taps Google's AI agent to give in-car navigation real memory

Anyone who has tried talking to an in-car assistant knows the pain: ask for a quick coffee stop along your route, follow up with "which one has drive-thru?", and the car instantly forgets the entire context.

Mercedes-Benz is trying to fix that frustrating back-and-forth. At the NRF 2025 retail show in New York, the luxury automaker revealed that it is bringing conversational search to its navigation via Google's AI agent technology.

Powered by Google Cloud's Automotive AI Agent, the feature runs on Gemini models through Vertex AI. It will officially debut later this year inside the new Mercedes-Benz CLA, arriving alongside the automaker's revamped MB.OS operating system.

Natural banter with live map data

The marquee upgrade for the MBUX Virtual Assistant is multi-turn conversation memory. The system can retain your spoken dialogue and recall details later in the trip, meaning you can reference earlier topics without starting from scratch.

Whether that conversational data remains stored once the engine is turned off is still unclear.

Beyond raw conversation, the integration taps directly into Google Maps, pulling from a database of roughly 250 million locations worldwide that sees over 100 million updates every day.

Drivers can ask complex, compound questions on the fly. For instance, you could search for Italian restaurants along a specific highway route and ask the system to cross-check real-world customer reviews before picking one.

Google CEO Sundar Pichai framed the partnership as the beginning of how autonomous AI agents will transform the automotive sector. Other luxury players like Porsche already support conversational interfaces for navigation, but they lack this direct tie-in to Google Maps reviews and rich real-time data.

Still, it might not be bulletproof yet. As Near Media cofounder Greg Sterling pointed out regarding Google's conversational map features, asking highly layered questions sometimes returns a generic warning that there is not enough local information to answer.

Why it matters

For drivers, in-car voice control is finally evolving from an awkward command-line interface into a genuine co-pilot that can plan stops dynamically. If you commute often or take road trips, having an assistant that actually remembers what you said two minutes ago makes in-car interaction far less tedious.

For the auto industry, Mercedes is acknowledging that building world-class AI and maintaining global mapping data from scratch is a losing battle for a legacy automaker. Tapping Google Cloud gives Mercedes instant access to cutting-edge models and massive POI databases without reinventing the wheel.

We think this partnership is a sharp tactical move to keep the upcoming CLA competitive against tech-heavy rivals. But the trade-off is clear: by outsourcing the brains behind navigation and ambient search to Google, Mercedes risks surrendering a piece of its software independence and customer data loop to Big Tech.