The automotive industry stands at a fascinating crossroads. While much attention focuses on the shift from internal combustion engines to electric vehicles (EVs), a more profound transformation is quietly taking shape inside our vehicles: the emergence of AI agent ecosystems.
At IAA 2025, Mercedes-Benz unveiled something remarkable in their new GLC model - not just another voice assistant, but an orchestration layer. The "Hey Mercedes" system now acts as a conductor, calling upon specialised AI agents from Microsoft, Google and other providers based on the task at hand. Ask for a fine dining restaurant nearby, and a Google agent springs into action. Need to capture a sudden flash of inspiration whilst driving? A Microsoft note-taking agent is ready.
In January 2025, Mercedes-Benz and Google Cloud announced an expanded partnership to introduce conversational capabilities powered by Google's Automotive AI Agent, built using Gemini on Vertex AI. This integration gives drivers access to information about 250 million places worldwide through Google Maps Platform, with the map updated nearly in real time through over 100 million daily updates.
This isn't simply about adding more features. It's about fundamentally rethinking what a car can become.
The car as a creative space
There's a psychological phenomenon many drivers know well: autopilot mode, when our brains shift focus during repetitive tasks like driving. It's why some of our best thinking happens behind the wheel - our conscious mind handles the familiar patterns of the road whilst our subconscious works through complex problems.
But until now, capturing those insights meant fumbling for your phone or hoping you'd remember later. Mercedes' integration of note-taking agents changes this dynamic. Drivers can now brain-dump directly whilst maintaining focus on the road, with their thoughts automatically synced to their company's SharePoint or OneDrive, secured by the same enterprise policies that protect their laptops.
The implications extend beyond productivity. Cars already possess contextual awareness that smartphones cannot match: driving speed, heart rate, facial expressions via interior cameras, even emotional state. When you're stressed after a difficult call, your car knows. When you're fatigued, it can detect the subtle signs. This rich context makes the vehicle an ideal platform for personalised AI assistance.
Context awareness is essential for autonomous AI systems, referring to a system's ability to collect and process environmental information to make decisions. The global autonomous AI and agents market is expected to reach $70.53 billion by 2030, registering a compound annual growth rate of 42.8%. Agentic AI represents a new era where autonomous agents manage processes, freeing people to focus on true value creation and revolutionising industries.
Beyond walled gardens
Yet here's where the automotive industry risks repeating the mobile phone industry's early mistakes. Remember when phone manufacturers tried to tightly control which applications could run on their devices? That approach failed spectacularly once Apple and Google opened their platforms to millions of developers.
The same principle applies to vehicles. No single automaker - regardless of resources - can anticipate and develop AI agents for every possible use case. Why would you use a limited in-car system when CarPlay or Android Auto offers the full power of your smartphone's AI capabilities?
The answer isn't to compete with smartphones. It's to embrace an ecosystem mindset.
Consider the diversity of human needs. One driver might want a fitness coaching agent that uses the car's sensors to monitor health metrics. Another might want a language learning agent for their daily commute - say, someone with a Polish spouse wanting to finally master the language. A third might want a meditation guide that responds to stress indicators during heavy traffic.
These hyper-personalised experiences mirror what we've achieved with smartphone app stores. Apple's App Store ecosystem alone facilitated $1.3 trillion in developer billings and sales in 2024, with over 90% commission-free for developers. That scale of innovation - millions of developers creating solutions for billions of users - is only possible through open platforms.
The question for automakers is whether they'll enable this level of personalisation or resist it.
The architecture of trust
Opening up a vehicle platform to third-party AI agents isn't straightforward, and the challenges are both technical and philosophical.
From a governance perspective, automakers selling vehicles for €85,000 or more cannot allow just anything to run in their carefully engineered environments. Some content and capabilities must remain restricted to maintain brand standards and, more importantly, driver safety.
The architectural challenge is equally complex. When multiple AI agents might respond to a single request, the voice assistant must determine which responses are relevant, synthesise them appropriately and present information in a way that doesn't overwhelm or distract the driver. It needs to understand intent - not just keywords - and route requests to the most appropriate specialist agent.
At Endava, we have developed an AI orchestration layer as part of a proof of concept, demonstrating how multiple specialised agents can interact seamlessly while maintaining safety and contextual integrity.
This orchestration layer is the technical moat. Much like how Alexa skills work with Echo devices (though more sophisticated), the system must manage:
- Intent routing: Determining which agents should handle each request
- Context management: Sharing relevant vehicle and driver state with agents whilst maintaining privacy
- Response synthesis: Combining insights from multiple agents into coherent, actionable information
- Safety filtering: Ensuring agent outputs don't compromise driver attention or vehicle safety
Equally critical is security and privacy. When third-party agents access vehicle sensors, driver behaviour data and personal information, robust security frameworks become non-negotiable. This means end-to-end encryption, secure authentication protocols, granular permission controls and transparent data usage policies. Drivers must understand what data each agent can access and retain the ability to revoke permissions at any time.
The ecosystem imperative
The automakers I've spoken with are grappling with this transition. Some still operate with a closed thinking mindset, believing they can develop everything in-house.
But the mathematics are stark. Even the largest OEM cannot compete with the collective innovation power of millions of developers. The smartphone revolution proved that open platforms, properly governed, win.
The automotive AI market is projected to grow from $18.83 billion in 2025 to $38.45 billion by 2030, at a compound annual growth rate of 15.3%. This growth is driven by autonomous vehicles, in-vehicle data processing needs and the demand for intelligent systems. But much of this value will accrue to those who enable ecosystems, not those who build walls around their platforms.
This doesn't mean chaos. It means creating the rails within which innovation can safely run. Establishing an app store model for vehicle AI agents where:
- Developers can create and monetise specialised agents
- Automakers maintain quality control and safety standards
- Drivers enjoy the hyper-personalisation they've come to expect from other digital platforms
- Business models work for all parties - OEMs, developers and end users
What this means for automakers
The transition to EVs represents a change in powertrain. The transition to AI agent ecosystems represents a change in business model and strategic positioning.
The software-defined vehicle paradigm - where physical and digital components are decoupled and features are defined through software - is already reshaping industry architecture. In this new model, vehicles become platforms for continuous innovation rather than fixed products.
Automakers can choose to be guardians of closed systems, attempting to be all things to all drivers. Or they can evolve into platform providers, enabling an ecosystem whilst maintaining the architectural integrity and safety that premium vehicles demand.
Those who embrace the latter will find themselves in a stronger competitive position. Not because they've built the single best AI agent, but because they've enabled thousands of others to build agents that collectively serve every conceivable driver need.
The future of in-vehicle AI isn't about which automaker has the smartest assistant. It's about which creates the most vibrant ecosystem - where your car becomes as personalised as your phone, whilst leveraging context that no phone can match.
The orchestration has already begun. The question is: will the full orchestra be invited to play?
Ready to explore more about AI transformation in automotive? Dive deeper with our e-book where we unpack the revenue opportunities for automakers.