North American OEMs Are Rethinking Automotive AI: Here’s What’s Driving the Shift

As vehicles become increasingly software-defined, North American automakers are re-evaluating some of the most fundamental assumptions behind infotainment, voice assistants, and in-vehicle AI.

In our work with the world’s leading automakers, our discussions have shifted beyond features and interfaces toward deeper architectural questions: How should software scale across vehicle lines and brands? What needs to be embedded versus cloud-based? How do OEMs retain control over data, branding, and long-term evolution as vehicles become platforms?

These questions reflect broader patterns emerging across the North American market as OEMs adapt to longer software lifecycles, tighter cost constraints, and rising expectations around reliability and control.

To explore how those priorities are shaping technology decisions, we spoke with Robby Ilich, Director of Sales Engineering at Cerence AI. In this Q&A, Robby shares his perspective on why one-size-fits-all infotainment is losing relevance, how embedded, automotive-grade AI is being re-valued, and what OEMs are looking for in scalable, future-ready platforms.

Q: You spend a lot of time with OEM engineering and platform teams. What’s the biggest shift you’re seeing right now?

Robby Ilich:
The biggest change is where decisions are being made. More and more, we’re talking with centralized platform and software teams rather than individual feature owners.

As vehicles become software-defined, OEMs are thinking less about point solutions and more about long-term architecture. They want systems that can evolve, scale across programs, and support multiple vehicle segments without forcing a reset every model cycle. That mindset naturally challenges the idea of one-size-fits-all infotainment.

Q: What’s behind the move away from one-size-fits-all infotainment?

Robby:
A lot of it comes down to scale and reuse. OEMs are trying to avoid reinventing the wheel across vehicle lines and brands, but they also don’t want a “take it or leave it” experience that limits differentiation.

They’re looking for platforms that make it easier to build once and adapt intelligently without turning every program into a bespoke integration project. And as that happens, voice becomes more central because it reduces dependency on deep menu trees and visual interaction, especially as vehicles add more features over time.

Q: Why is there so much attention on embedded AI right now?

Robby:

It comes down to two factors. First, connectivity in the vehicle is inherently variable. Coverage changes, latency fluctuates, and systems still need to behave predictably regardless of network conditions.

Second, OEMs are becoming much more conscious of the rising costs associated with cloud‑based AI services.

Together, these two factors are driving a renewed focus on embedded AI. OEMs want confidence that core interactions like voice will work consistently without depending on always‑on connectivity or frequent cloud calls. Pushing more intelligence into the vehicle helps control operational costs and keeps bandwidth available for use cases that can only work using real‑time data or higher processing power.

Cerence’s embedded AI has reached a level of maturity where it can handle the majority of in‑vehicle use cases locally, without sacrificing quality or capability. With our embedded technology, OEMs can deliver responsive, reliable experiences while being far more selective and strategic about when and where cloud AI is used.

Q: How are software-defined vehicles changing the way OEMs evaluate vendors and platforms?

Robby:
There’s a much stronger emphasis on control and long-term optionality.

OEMs are thinking carefully about who owns the architecture, how data and branding are handled, and how easily their platform can evolve over time. They’re wary of tightly coupled stacks where intelligence, services, and providers are all bound together, because that can limit flexibility down the road.

What many OEMs are gravitating toward instead are platforms designed to be adaptable from the outset: flexible, technology-agnostic, and able to integrate with different models, services, and partners as requirements change.

That’s the thinking behind Cerence xUI. It’s built to support OEM choice (including which models, hardware and service providers to use) and designed to fit into OEM-defined environments rather than impose a fixed ecosystem.

Q: When OEMs say they want “a scalable platform,” what do they actually mean?

Robby:
Scalability isn’t just about deploying at volume. It’s about repeatability and reuse.

OEMs want platforms they can carry across brands, vehicle lines, and regions, while still allowing for differentiation where it matters. They’re trying to reduce fragmentation and avoid rebuilding the same capabilities over and over again.

There’s also a very practical hardware dimension. Many OEMs are focused on avoiding unnecessary hardware escalation. They’re looking for right-sized intelligence – systems that deliver strong performance on widely adopted automotive hardware, instead of requiring ever-higher specs just to keep moving forward.

Q: Are there misconceptions you see when teams start moving toward this SDV-style platform approach?

Robby:
One misconception is that speed comes from keeping things loosely defined early on. In reality, the OEMs making the most progress invest upfront in clear architectural boundaries and realistic assumptions about system behavior.

Another misconception is underestimating how valuable simplicity can be. Systems with clear reasoning tend to scale better over time than highly complex stacks, even if those stacks appear more powerful on paper.

Q: What’s the big takeaway you’re hearing from platform leaders and product executives?

Robby:
The SDV shift is forcing a reset in how OEMs think about long-term software ownership.

The platforms that succeed will scale across programs, behave predictably in real driving conditions, and preserve OEM control over the experience.

The most important signal in North America right now is that OEMs aren’t just choosing features; they’re choosing foundations. As vehicles become platforms, the bar rises for reliability, flexibility, and long-term control.

If you’re building for the next decade of software-defined vehicles, the question isn’t “Can this work?” It’s “Can this scale, evolve, and hold up in the real world, year after year?”